# Artificial Poets Artificial Poets builds AI and ML products for media and publishers — autonomous agents, content intelligence, and revenue optimization. ## Posts ### The Real Future of AI for Media & Publishers https://www.artificialpoets.com/blog/the-real-future-of-ai-for-media-publishers/ DISCLAIMER: No AI was usted to write this article. It's been over 20 months since we started working on Artificial Poets. We wrote the very first line of code in December 2024, is not that far away but the world changed a lot since then. We started with a idea in mind: to be an AI-Native company, something that is still true today, however the whole initial idea we had-- got transformed many times until come to what we have today. Which I think it's the natural process of developing any kind of new idea: you need to keep working and discovering and adapt as you keep making progress. Our initial thesis was related to be able to leverage LLM models to do work inside WordPress, like generate images, suggest titles, improve SEO, very generative AI intensive. We worked the first two quarters making improvements and handling system prompts to be able to output great results. For our own luck we moved away from this because we could not find any value there. Later the models will become so good that our system prompts would become obsolete. The real problem But there was a problem that keep resonating with us: something was not right in the way millions of Media & Publishers companies operate. Not because they did something wrong or choose the wrong technology, simply because there were other companies, specially social media like Meta, TikTok, YouTube, Spotify or Netflix that captured billions of users and offered a better experience. And the thing is once you experience that, you do not want to go back. May 2025 was the month were Google released AI Overviews or AI Mode, basically a new way for Google Search to give you answers, this caused and keep causing a HUGE problem to Media & Publishers, people search in Google and now they see the answer in front of their eyes, this is causing severe problems for them, around 40% of the organic search traffic is gone, there are not more clicks, you search what you want to know, you get the answer, and you keep navigating. Where the attention ($) is going? Traditionally the chosen business model for online publishers were ads, even if in the last decade some companies started with paid tiers like subscriptions or memberships and gate some of their content or features under it, ads represent a huge part of the Internet Economy, if you think on Internet giants as Meta or Google, their make most of their revenue because of their ads. 68% of Google's revenue is because of ads. So, does Google AI Mode impact negatively in their own business model? No, it's the opposite, now the supply of click is lower, so each click is increasingly appreciated by the company willing to purchase it. Before you could have 70% of your revenue coming from organic traffic and 30% via ads, after May 2025 that 70% with the 40% decrease becomes 42%. This has a HUGE impact in the economics of their business model, not only for the ads, but for conversions, affiliates, upgrades, etc. In a different direction the attention gap of the people is decreasing year over year, and the consumers are putting all their attention on social platforms: Instagram, TikTok and YouTube. So the scenario for anyone building a content business over the Internet seems to be getting attacked in multiple directions. AI joins the room During the 2010s most digital publishers were falling behind in terms of technology. Let's face it, WordPress powers a big chunk of the Internet, including most of media & publishers that you can name. WordPress is a great tool and piece of technology, with huge community, a lot of controls and customizations. I truly love it. However WordPress was not designed for the AI era, and that is totally fine. WordPress play it roles pretty well, it's the default tool for any publisher to publish content and have a digital presence, it works well for small sites or for massive sites. There are multiple providers like WPVIP or Kinsta. There are plans from $5 dollars to the $100,000+ so you get the point. However it was not designed to serve or understand the audience. Because of this last premise many companies born during the decade of 2010s to solve that problem. However most of these companies do not rely on a method that you can prove to understand and serve better the audience. Most of this tools are related to attribution and adding extra layers of information you can use for a monetization mainly, like demographics, firmographics, PII and so. The closest thing to user behavior we had during then was heat maps. Where I think Media & Publishers should be with AI I am not going to get technical here, just let me give you some context. For the nerds like me 12 of June, 2017 is a important date: a paper titled "Attention Is All You Need" was published. This paper played a huge role into the development of modern AI. Introduced the transformer architecture, so important that the T in ChatGPT means transformer- a little tribute from OpenAI. The full circle comes from the attention part. Media & Publishers are losing attention, and not the kind of attention coming from the transformer, real attention from the humans: which translates directly into revenue, the fuel that power any business. Since attention is getting scarce overtime, we need to adapt and come with new ways to generate value. Back in 2008 you could have a good business out of impressions alone just running Google AdSense without even negotiate with any advertiser. If we look back we will always find inflection points where this industry been evolving and changing drastically because the attention of the consumers changes. And I strongly believe this will be the constant and something the new generations will have to deal with it too. However, we are in another huge inflection point in front of us, for the first time ever the technology and resources to be able to operate with a high level of understanding what our audience wants based on behavior and strong pattern recognition, and based on that take actions to serve them better, with personalization at scale, and being able to maximize our revenue, a full circle where it compounds overtime. It was extremely complicated to be able to run this machine learning a few years ago, not only the innovation in the transformer but also the advancement in GPU to take fully advantage of this make it possible today. Companies like Meta, Netflix or TikTok spent billions developing systems with neuronal networks and attaching their economical value into paths so later they can optimize it. When you do that, you have a machine that improves overtime towards a clear objetive that you can measure it. You have a flywheel. Media & Publishers powered by AI Now it's time to convert that vision into a reality. As I mentioned earlier, we started to work on this around 20 months ago and since then we been working day at night to make this happen. But finally it's here! Our way to see how Media & Publishers should operate with AI is what we define as the AI-Native Publisher, and this has nothing to do with generate the content using AI, that is not part of our vision, very little people wants to consume AI generated content, even with the highest level of taste on it. Some others could argue that as long as there's value in the content is fine. How should Media & Publishers operate with AI? We strongly believe that personalization IS the business model. Do you think Meta or TikTok would make the same amount of money and have the same amount of growth if they show the same feed to every single user? It's evident that no, right? So what you do it yourself? Why you are treating in the same way a visitor that been coming over and over for the past years in the same way that just a regular visitor? Being a AI-Native Publisher means that you could serve your audience in orders of magnitude better. This means they will be able to find what's relevant for them, your platform should understand what they want, how they want and when they want. This is possible today. And with this, you are not only serving your audience better, which is great, you are also unlocking flywheels related to revenue. Let's put it in this way, if a visitor keep finding what they like, they will spend more time in your site, they will be more engage, if you make it easy for them to share the content, they will attract more people. Google will notice the people engage more with your content and will help you to rank better in Google Search and the AI Overviews. You will know your visitor so well that you will know when it's the right moment to ask them to convert into a paid subscriber, and they will be so engage and see so much value in your content that they will do it frictionless. All of this happening in real time, the same machine learning that if you go to Instagram or TikTok and likes cats you start seeing cats, but applied to your publishing business, in a way you can control. On top of that you will have a mechanism that knows exactly if you are offering a great user experience for them, because maybe you ask them to convert into a paid subscriber but there are other components that are affecting that conversion. The system should be smart to know and find patterns. Now this idea of flywheel it makes sense, right? Let me present you something. Talk about the vision of Artificial Poets. Italy, October 2025. Introducing Artificial Poets Artificial Poets helps Media & Publishers to become AI-Native in matter of weeks. Instead of spend 20+ months and build this yourself, you can just use our Platform and become AI-Native and start operating at orders of magnitud, discovering and building your own flywheels, that understand your audience, your content and your business model. Tailored to your business, responding to you interested and without compromising the labor of your editorial team. The way we build it do not require you to change anything in your website. We have a proven way to implement our Platform in any site, and we have a mathematical methodology to measure and guarantee results. You can read more about this in our website, see the different case studies we have, read our research and get in touch too. Artificial Poets mission's to democratize the access to neuronal networks and artificial intelligence and make it accesible to everyone. I am extremely exited and confident of the future we are building. Starting with making accesible advanced machine learning and recommendations engine used by elite companies like Netflix, Spotify or Meta and making it accesible to all. If you are confident about the future of AI in Media & Publishing and think this way of operate would be beneficial for your organization, feel free to reach out. ## Pages ### We Help Media Companies Become AI-Native https://www.artificialpoets.com/ The AI that decides what people read, watch and listen to next — the kind Netflix, Meta and Spotify built for themselves, brought to publishers and media networks. ### Terms & Conditions https://www.artificialpoets.com/terms/ These Terms and Conditions (the "Terms") are an agreement between Artificial Poets, Inc., a Delaware corporation ("Artificial Poets," "we," "us"), and you. They govern your use of our websites, products, and services, including PoeHost, the Artificial Poets Platform, a13s, InboundMode, OutboundMode, Intent Analysis, our documentation sites, and our APIs (together, the "Services"). By creating an account, clicking to accept, or using the Services, you agree to these Terms. If you are using the Services for an organization, you represent that you have authority to bind that organization, and "you" means the organization. If your organization has a signed agreement or order form with us (for example, an enterprise subscription), that agreement controls where it conflicts with these Terms. The short version You keep ownership of your content. We need a limited license to host and process it, purely to run the Services for you.You own the AI outputs we generate for you, and you are responsible for reviewing them before relying on them.Fees are your subscription plus any metered usage (such as AI tokens), billed as described in your plan.Use the Services lawfully; do not abuse them or send spam through them.Open source components we publish are governed by their own licenses, not these Terms.The Services are provided as is, our liability is capped, and disputes go to individual arbitration under Delaware law (you can opt out of arbitration within 30 days). The full Terms below are what govern. 1. Eligibility and accounts You must be at least 18 and able to form a binding contract. The Services are built for business and professional use. You are responsible for your account: keep your credentials confidential, keep your information accurate, and tell us promptly at legal@artificialpoets.com if you suspect unauthorized use. Organizations control their own workspaces, including inviting members, assigning roles, and removing access; members act on the organization's behalf. 2. The Services; changes; beta features We provide the Services described in your plan or order form. We improve the Services continuously and may add, change, or remove features; if a change materially reduces the core functionality you pay for, you may cancel and receive a pro-rated refund of prepaid, unused fees for the affected Service. Features labeled alpha, beta, preview, demo, or early access are provided for evaluation, may change or be withdrawn at any time, may be incomplete, and are excluded from any support or service commitments. 3. Fees, billing, and taxes Subscriptions. Plans are billed in advance on the cycle you choose. Except where these Terms or your order form say otherwise, fees are non-refundable. Metered usage. Some Services meter usage, such as AI token consumption, and bill it in arrears at the rates shown in your plan. To protect you from runaway costs, plans include usage allowances and spend limits; when a limit is reached, the metered feature pauses until the period resets or you raise the limit. Payment. Payments are processed by our third-party payment processor; invoices are payable per their stated terms. If an amount is past due, we may suspend the affected Services (including AI features) until the balance clears, after notice. You are responsible for applicable taxes other than taxes on our income. Price changes. We may change pricing with at least 30 days' notice; changes take effect at your next renewal. Cancellation. You can cancel any time, effective at the end of the current billing period. Access continues through the period already paid. 4. Your content "Customer Content" is everything you or your users submit to the Services: websites, databases, posts, media, campaigns, conversations, files, and data from accounts you connect. You own it. We claim no ownership of Customer Content. License to us. You grant us a worldwide, non-exclusive license to host, store, copy, process, transmit, display, and back up Customer Content, solely as needed to provide, secure, and support the Services, to meter usage for billing, and as described in our Privacy Policy. This license ends when the content is deleted from the Services, except for backups held for their scheduled window and records we must keep by law. Your responsibilities. You are responsible for Customer Content: that you have the rights to use it, that it is lawful, and that your own collection of data from your users and visitors complies with the laws that apply to you, including having your own privacy policy where required. Our access. Our staff access Customer Content only to provide, secure, and support the Services, or with your permission. Data use and model training. Our data practices, including the default described here, are set out in the Privacy Policy: by default we collect usage data and analytics and may use de-identified, aggregated data to improve the Services and train our own models; we do not use Customer Content or your end users' personal information to train models unless your organization opts in. 5. AI features and outputs The Services generate content and take actions using artificial intelligence, including third-party models. Outputs are yours. As between you and us, and to the extent we hold any rights in them, we assign to you our rights in the outputs the Services generate for you ("Outputs"). Because models can produce similar results for similar requests, we cannot promise Outputs are unique, and other customers may receive similar Outputs. Verify before you rely. AI can be wrong, incomplete, or biased. Outputs are not professional advice. You are responsible for reviewing Outputs before publishing them or acting on them, and for how you use them. Agent actions. Some features can take actions in connected accounts you authorize (for example, updating an analytics property or managing a campaign). You control what is connected and what is permitted, and you are responsible for actions taken with the permissions you grant. Revoke access at any time by disconnecting the account. 6. Acceptable use Do not, and do not allow others to: Break the law, or infringe anyone's intellectual property, privacy, or other rightsUpload or distribute malware, or attempt to probe, breach, or disrupt the Services or anyone else's systemsSend spam or any email that violates applicable law (including CAN-SPAM) through the Services; email you send must go to recipients who have consented where consent is required, and honor unsubscribesMisrepresent the origin of content, or impersonate any person or organizationCircumvent usage limits, metering, or access controls, or share credentialsResell, sublicense, or provide the Services to third parties as your own offering, except as your plan or an agreement with us allowsScrape or bulk-extract data from the Services outside the documented APIsUse the Services to build a competing product by systematically copying their features or dataUse the Services for high-risk activities where failure could lead to death, personal injury, or severe damage We may suspend or limit access immediately where we reasonably believe an account threatens the security or integrity of the Services or others, and we will restore it once the threat is resolved. 7. Third-party services and connected accounts The Services interoperate with third-party platforms (for example Google, Meta, Slack, Microsoft, and WhatsApp). Your use of those platforms is governed by their terms, and you are responsible for complying with them. We are not responsible for third-party platforms, their availability, or changes they make. Connecting an account authorizes us to access it as needed to provide the features you use. 8. Hosting-specific terms (PoeHost) You are responsible for your website and everything on it: content, themes, plugins you install, and its compliance with law, including toward your own visitors.Backups are taken and retained per your plan. Restores are available through the Services; test what matters to you.Plans include resource allocations (visits, storage, bandwidth). We will not silently degrade your site: if you consistently exceed your plan, we will contact you about upgrading.We may apply protective measures (for example caching, security rules, or malware remediation) to keep your site and the platform healthy.Software we install on hosted sites to provide the Services is licensed to you for use with the Services and remains ours or our licensors'. 9. Domain registration Where you purchase a domain through the Services, we register it through our registrar account on your behalf. You are the beneficial owner of the domain: you direct its use, renewal, and disposition. On your request and subject to registry rules (including ICANN transfer windows), we will transfer the domain to a registrar account of your choosing or update the registrant to you. Keep your renewal payments current; expired domains can be lost or incur redemption fees set by registries. 10. Open source and trademarks Some of our software is published under open source licenses (for example Apache 2.0, or GPLv2 or later for WordPress plugins). Those licenses, not these Terms, govern your use of that code, and nothing in these Terms limits rights they grant you. No trademark rights are granted by these Terms or by any open source license. Artificial Poets, PoeHost, a13s, InboundMode, OutboundMode, Intent Analysis, PoeticUI, and our logos are marks of Artificial Poets, Inc. 11. Our intellectual property; feedback We and our licensors own the Services, including software, models, interfaces, and documentation. We grant you a limited, non-exclusive, non-transferable right to use them per these Terms and your plan. If you send us feedback or suggestions, we may use them without restriction or obligation. 12. Confidentiality Each of us may learn non-public information of the other through the Services ("Confidential Information"). Each of us will protect the other's Confidential Information with reasonable care, use it only as needed under these Terms, and not disclose it except to personnel and advisers under confidentiality obligations, or where required by law with notice where permitted. 13. Privacy and data protection Our Privacy Policy (artificialpoets.com/privacy-policy/) describes how we handle personal information and is part of these Terms. Where you need a data processing agreement covering personal data we process on your behalf, contact legal@artificialpoets.com and we will put one in place. 14. Publicity We may identify you by name and logo as a customer in our marketing. If you prefer we do not, email legal@artificialpoets.com and we will stop. 15. Term, suspension, and termination These Terms apply while you use the Services. Either party may terminate for material breach if the breach is not cured within 30 days of notice. We may terminate or suspend immediately for violations of section 6, for non-payment after notice, or where required by law. After termination. Your access ends. For 30 days after termination, we will make your Customer Content available for export in a reasonable format on request. After that we delete it in the normal course, per the retention windows in the Privacy Policy. Sections that by their nature should survive (including 4 ownership, 10 through 12, 16 through 18, and 19) survive. 16. Disclaimers The Services are provided "as is" and "as available." To the fullest extent permitted by law, we disclaim all warranties, express or implied, including merchantability, fitness for a particular purpose, non-infringement, and any warranty that the Services will be uninterrupted, error-free, or that Outputs will be accurate or reliable. Service level commitments exist only where an order form grants them. 17. Limitation of liability To the fullest extent permitted by law: neither party will be liable for indirect, incidental, special, consequential, or punitive damages, or lost profits, revenue, data, or goodwill, even if advised of the possibility; and each party's total liability under these Terms is capped at the greater of $100 and the amounts you paid us for the Services in the 12 months before the claim. These limits do not apply to your payment obligations, your breach of section 6, either party's breach of section 12, or your indemnification obligations, or to liability that cannot be limited by law. 18. Indemnification You will defend and indemnify Artificial Poets against third-party claims arising from Customer Content, your use of the Services in violation of these Terms or law, or your violation of third-party rights, and pay resulting damages, settlements, and reasonable costs. We will notify you promptly of any claim and reasonably cooperate at your expense. 19. Governing law and dispute resolution These Terms are governed by the laws of the State of Delaware, without regard to conflict of laws rules. Talk first. Before filing a claim, either party will send notice to the other (for us: legal@artificialpoets.com) describing the dispute, and both will try in good faith to resolve it within 30 days. Arbitration. Any dispute not resolved informally will be finally settled by binding arbitration administered by the American Arbitration Association under its rules, by a single arbitrator, in English. Judgment on the award may be entered in any court with jurisdiction. Either party may instead bring an individual claim in small claims court, and either party may seek injunctive relief in the state or federal courts located in Delaware for infringement or misuse of intellectual property or Confidential Information. No class actions. Disputes will be resolved only on an individual basis. Neither party may participate in a class, consolidated, or representative action, and the arbitrator may not consolidate claims. Your right to opt out. You may opt out of this arbitration agreement by emailing legal@artificialpoets.com within 30 days of first accepting these Terms, with your name and organization. If you opt out, disputes will be resolved in the state or federal courts located in Delaware, and both parties consent to their jurisdiction. 20. Changes to these Terms We may update these Terms. For material changes we will give at least 30 days' notice by email or through the Services before they take effect; other changes take effect when posted. If you do not agree to a change, cancel before it takes effect; continued use after the effective date is acceptance. The current version always lives at artificialpoets.com/terms/. 21. General Notices. To us: legal@artificialpoets.com, or Artificial Poets, Inc., 131 Continental Drive, Suite 305, Newark, DE 19713, USA. To you: the email on your account.Assignment. You may not assign these Terms without our consent, except to a successor in a merger or sale of substantially all assets, with notice. We may assign them to an affiliate or successor.Entire agreement. These Terms, your plan or order form, and the policies they reference are the entire agreement and supersede prior discussions.Severability and waiver. If a provision is unenforceable, the rest stands. Not enforcing a provision is not a waiver.Force majeure. Neither party is liable for delay or failure caused by events beyond its reasonable control.Export and sanctions. You will comply with United States export control and sanctions laws and represent that you are not barred from receiving the Services under them.Independent parties. Nothing here creates a partnership, agency, or employment relationship. Questions about these Terms: legal@artificialpoets.com. ### Privacy Policy https://www.artificialpoets.com/privacy-policy/ Artificial Poets, Inc. ("Artificial Poets," "we," "us") builds publishing technology: managed WordPress hosting, analytics, AI assistants, and the tools around them. This policy explains what personal information we collect, why we collect it, and the choices you have. It covers artificialpoets.com and our product surfaces and brands, including PoeHost, the Artificial Poets Platform, a13s, InboundMode, OutboundMode, Intent Analysis, and our documentation sites (together, the "Services"). The short version We collect what we need to run the Services: account details, lead form submissions, billing records, usage analytics, and support conversations.We do not sell personal information, and we do not share it for cross-context behavioral advertising.By default we collect usage data and analytics, and we may use de-identified, aggregated data to improve the Services and train our own models. We do not use your content, or your end users' personal information, to train models unless your organization chooses to opt in.The AI model providers we work with process data under terms that do not allow them to train their models on it.Everything runs in the United States.When we process data on behalf of a customer (for example, visitors to a website we host or measure), that customer controls the data and their privacy policy applies.You can ask us to access, correct, delete, or export your information at any time: privacy@artificialpoets.com. The full policy below is what governs. 1. Our two roles We handle personal information in two capacities, and your rights run to the right party: As the business (controller). For our own websites, marketing, lead forms, newsletters, sales outreach, accounts, and billing, we decide how and why data is used. This policy applies directly. As a service provider (processor). When a customer uses our Services to run their business (hosting their website, measuring their audience, sending their email, powering their AI features), we process their data, including their visitors' and end users' personal information, on their instructions. The customer controls that data and their privacy policy governs it. If you are a visitor or user of a site or service operated by one of our customers, please direct privacy requests to that customer; we will assist them in responding. Section 8 describes this processing. 2. Information we collect Information you give us. Account information. Name, email address, profile photo, organization name, and role, collected when you or your organization creates an account. Authentication is handled by our identity provider; we do not see or store your password.Lead and newsletter forms. When you subscribe to a newsletter, download a guide, or request a demo on our sites, we collect what the form asks for: email address, and depending on the form, first and last name, company, and phone number.Billing information. Billing contact, company details, and billing address. Payment cards are entered directly with our payment processor on its hosted pages; full card numbers never touch our systems. We keep invoices and payment status.Communications. Messages you send us for support, sales, or anything else, and, for conversational products such as a13s, the conversations you have with the assistant. Information we collect automatically. Website analytics. Our sites use a product analytics service (hosted in the United States) configured so that anonymous visitors are not individually profiled. If you submit a lead form, we connect your activity to a pseudonymous identifier derived from your email address, rather than to the address itself, so we can see which pages brought you to us. We also use Google Tag Manager to manage measurement tags; our sites push their own events into it and do not embed advertising pixels.Product usage and session activity. Within our applications we record product events (pages viewed, features used, API calls, AI token consumption and cost) tied to your account, along with the interactions that make up a visit, such as clicks and navigation. We also capture session replays: reconstructions of how a page was used, which can include on-page interactions and browser console output. We use them to find and fix problems, understand where people get stuck, and improve the experience. Where an email address is used for analytics, it is replaced with a pseudonymous identifier first. You can ask us to switch non-essential analytics off for your account or organization at any time; see section 4.Performance and error monitoring. We collect performance timings and error reports from our applications to keep them working. Inputs are masked and monitoring is tied to internal identifiers, not names or emails.Consent and anti-abuse records. When you submit a lead form we record the consent wording version you saw, the time, and your IP address, user agent, and country. This is the evidence that you actually signed up, and it also powers our spam and abuse protections.Logs. Standard server logs (IP address, request details, timestamps), retained briefly for security and operations. Information from other sources. Connected accounts. If your organization connects third-party accounts to the Services (for example Google Analytics, Google Ads, Search Console, or Meta Ads), we access that data with your authorization to provide the features you asked for. OAuth tokens are held by our integration broker; we fetch credentials at request time rather than storing them.Business contact information. Described in section 9. 3. How we use information We use personal information to: Provide, operate, and secure the ServicesCreate and manage accounts, organizations, and permissionsProcess payments, send invoices, and manage subscriptionsDeliver the newsletters, guides, and demos you ask for, with proof of consentRespond to support requests and communicate about the ServicesMeasure and improve the Services, including product analytics and performance monitoringTrain and improve our own machine learning models, using only the data described in section 4Conduct business-to-business outreach as described in section 9Detect and prevent spam, fraud, and abuseComply with law and enforce our agreements 4. AI features, analytics, and model training Several of our products use large language models and machine learning. Here is exactly what that means for your data. Processing to provide features. When you use an AI feature, the relevant content (a prompt, an article to classify, a page to optimize, data from a connected account) is processed by AI models to produce the result you asked for. We route these requests through our own infrastructure to established commercial model providers. Our agreements with those providers do not permit them to use your data to train their models. Usage metering. For billing and cost controls we record metadata about AI usage: model, token counts, and cost, attributed to your organization. We do not use the content of your prompts or conversations for billing records. What we use to improve the Services (the default). By default, your use of the Services is opted in to our collection of usage data and analytics as described in this policy, and we may use data that has been de-identified or aggregated so it no longer identifies any person or customer to improve the Services, develop new features, and train our own models, such as our content and intent classifiers. What we do not use (unless you opt in). We do not use your content (your sites, posts, media, conversations, campaigns, or data from your connected accounts), or personal information belonging to you or your end users, to train our models, unless your organization expressly opts in. Enterprise agreements can further restrict our data use, and where they do, the agreement wins. Opting out. If you want to opt out of non-essential analytics for your account or organization, email privacy@artificialpoets.com and we will configure it. 5. Cookies and similar technologies We use a small set of cookies and browser storage: Strictly necessary. Session and security cookies from our identity provider so you can sign in, and infrastructure cookies from our content delivery network.Analytics. Cookies set by our analytics service to distinguish visits, configured as described in section 2. Some interface preferences are kept in your browser's local storage and never leave your device. We do not use advertising cookies and we do not run third-party ad trackers on our sites. Our current approach is notice-based: we tell you here, and you can control cookies through your browser settings, including blocking or deleting them. We are building region-aware consent controls for visitors from jurisdictions that require opt-in consent, and this policy will be updated when they ship. Our sites are operated from the United States and are directed to a United States audience. 6. How we share information We do not sell personal information, and we do not share it for cross-context behavioral advertising. We share personal information only with: Service providers. Vendors that process data on our behalf and on our instructions: cloud infrastructure and hosting, content delivery and security, identity and authentication, payment and billing processing (including our payment processor's hosted checkout), email delivery, product analytics, performance monitoring, integration brokering, and AI model providers. We limit each provider to the data it needs to do its job.Professional advisers such as lawyers and accountants, under confidentiality.Legal and safety. When required by law, legal process, or to protect the rights, safety, and property of Artificial Poets, our customers, or others.Business transfers. If we go through a merger, acquisition, financing, or sale of assets, personal information may transfer as part of that transaction, subject to this policy. A list of the specific vendors we use to process customer data is available on request at privacy@artificialpoets.com. 7. Lead capture, newsletters, and email Our lead and newsletter system is built for consent you can verify: Each subscription is stored per list, with the exact consent wording version you agreed to and when.Lists may use confirmed (double) opt-in; where they do, you are not subscribed until you click the confirmation link.Every marketing email includes a working one-click unsubscribe that takes effect immediately for that list.Unsubscribing from one list does not touch your other subscriptions; bounce and complaint suppressions apply account-wide.If you unsubscribe, we keep a minimal suppression record (a hashed identifier) so we do not email you again. We do not send marketing SMS. 8. When we process data for our customers Customers use our Services to run their own websites and audiences. In that processing, the customer is the controller and we act on their instructions. Depending on what the customer has enabled, this can include: Hosting their website and its database, media, and backupsCollecting analytics events from their site's visitors, including page views, performance metrics, searches, and (if the customer enables it) session recordings, into the customer's analytics projectComputing audience insights for the customer, such as content and visitor intent classificationsRelaying transactional email their site sends (password resets, form notifications)Storing their uploaded media and files We do not use this data for our own purposes except as needed to provide and secure the Services, to meter usage for billing, and as described in section 4 (de-identified, aggregated improvement data). We do not pool one customer's data with another's, and each customer's data is segregated per organization. If you are a visitor to a customer's website, the customer's privacy policy applies to you. Contact the site owner to exercise your rights; we support our customers in honoring those requests. 9. Business contact and prospect information Because we sell to businesses, we maintain contact information about people in professional roles at companies we believe could use the Services. This includes name, job title, company, business email address, business phone, and links to public professional profiles. We gather this from publicly available professional sources and from licensed business data providers, and we use it for one purpose: relevant, business-to-business outreach about the Services. Every outreach email identifies us and includes a working unsubscribe. If you opt out, we add you to a permanent suppression list and stop contacting you. You can also email privacy@artificialpoets.com at any time to ask what we hold about you, correct it, or have it deleted; section 11 explains how. 10. Data retention We keep personal information only as long as we need it for the purposes above, and we use specific windows where we can: Lead and subscriber records: for as long as the list is active, plus a minimal suppression record after you unsubscribe. Historical capture event copies are deleted after 2 years.Product analytics events and AI usage records: 13 months.Assistant conversations (a13s): deleted after 12 months of inactivity, unless your plan includes longer or unlimited retention.Hosting backups: per your plan's backup window (7 to 365 days).Operational logs: around 30 days.Account, billing, and tax records: for the life of the account and as long as law requires afterward. When retention ends, we delete or de-identify the data. 11. Your rights and choices Wherever you live, you can ask us to: Access the personal information we hold about youCorrect it if it is wrongDelete itExport it in a portable formatOpt out of marketing email (every message has an unsubscribe link, or just ask) Send requests to privacy@artificialpoets.com. We will verify the request (usually by confirming control of the email address involved) and respond within 45 days. We will not discriminate against you for exercising any right. If we cannot fully honor a request (for example, records we must keep for tax or security reasons), we will tell you what we kept and why. We do not sell personal information or share it for cross-context behavioral advertising, so opt-out-of-sale signals such as Global Privacy Control have nothing to switch off today; if our practices ever change, we will honor them. If your data reached us through one of our customers, we will refer your request to that customer and help them respond. 12. Security We run the Services on established United States cloud infrastructure with encryption in transit, encryption at rest, and access limited by role and scoped credentials (for example, the key our websites use to submit leads can write a lead and do nothing else). Payment card data is handled entirely by our payment processor. No system is perfectly secure, but if a breach affects your personal information we will notify you and the authorities as the law requires. 13. Where your data lives We are a United States company and the Services are hosted and operated in the United States. If you use the Services from outside the United States, your information will be transferred to, stored, and processed in the United States, where privacy laws may differ from those in your jurisdiction. 14. Children The Services are for business use and are not directed to children under 16. We do not knowingly collect personal information from children. If you believe a child has given us personal information, contact us and we will delete it. 15. Changes to this policy When we change this policy, we will update the date at the top and, for material changes, give notice on our sites or by email. The current version always lives at artificialpoets.com/privacy-policy/. 16. Contact us Artificial Poets, Inc. 131 Continental Drive, Suite 305 Newark, DE 19713, USA Privacy requests: privacy@artificialpoets.com Everything else: legal@artificialpoets.com ### Coming Soon https://www.artificialpoets.com/coming-soon/ ### Contact https://www.artificialpoets.com/contact/ Tell us what you run and what you are trying to move. We will come back with a range and what a pilot would cover. ### Every Episode Findable https://www.artificialpoets.com/use-cases/podcast-discovery/ "We make great shows nobody finds — how does a listener discover the next episode, or the next show?" ### How We Measure https://www.artificialpoets.com/platform/measurement/ "How do I know any of this is real — and how do I prove it to my board?" ### One Audience. Every Property. https://www.artificialpoets.com/solutions/media-networks/ "We own a dozen properties — why does our audience behave like it belongs to strangers?" ### Keep the Readers Who Never Stay https://www.artificialpoets.com/solutions/mobile-social/ "Two-thirds of my traffic is mobile and social one-and-dones — can those readers actually be kept?" ### Defend the Ad Line https://www.artificialpoets.com/solutions/ad-revenue/ "Impressions are down and CPMs are soft — how do I defend ad revenue without more traffic?" ### Build the Audience That Comes Back https://www.artificialpoets.com/solutions/direct-audience/ "My subscriber targets keep rising while my funnel shrinks — where does direct audience actually come from?" ### Make Every Visit Worth More https://www.artificialpoets.com/solutions/session-depth/ "Traffic keeps falling and I can't stop it — how do I make the visits I still get worth more?" ### Get Them to the Second Episode https://www.artificialpoets.com/use-cases/video-recommendations/ "Subscribers churn before a second episode — how do I get them watching in the first session?" ### New Impressions on Inventory You Already Own https://www.artificialpoets.com/use-cases/ad-inventory/ "If sessions get deeper, how does that become sellable impressions in my existing ad stack?" ### See What Readers Want Before You Commission It https://www.artificialpoets.com/use-cases/editorial-intelligence/ "What should we publish next — and what does our audience want that we've never given them?" ### Ask at the Right Article https://www.artificialpoets.com/use-cases/depth-triggered-conversion/ "I'm showing newsletter popups to readers who were leaving anyway — how do I ask the ones who are ready?" ### Search That Understands the Question https://www.artificialpoets.com/use-cases/intent-search/ "Readers search my site and leave empty-handed — can search understand what they're actually asking?" ### Fix the Second Click https://www.artificialpoets.com/use-cases/next-article-recommendations/ "My related-articles widget shows everyone the same five links — what would a smart second click look like?" ### Serve the Next Article Before They Leave https://www.artificialpoets.com/use-cases/infinite-article-feed/ "What happens after a reader finishes an article — can the next one just be there?" ### You Could Build It. Should You? https://www.artificialpoets.com/platform/recommendation-engine/ "Our engineers say they could build this in a year. Are they wrong?" ### The Artificial Poets Platform https://www.artificialpoets.com/platform/ The Platform learns what every story is about and what every reader came for, then acts on it: what to serve next, what search should answer, when to ask, what to commission. ### Solutions https://www.artificialpoets.com/solutions/ One engine behind every problem here: readers who leave after one page, an audience you rent, ad revenue under pressure. For one title or a whole network, every solution ends in a number you can defend. ### Use Cases https://www.artificialpoets.com/use-cases/ One engine across articles, streaming video and podcasts, for a single title or a whole network. Every use case here keeps the visit going after the first click, and turns that into more impressions and signups. ### Resources https://www.artificialpoets.com/resources/ White papers with the method attached, guides for auditing any vendor claim including ours, and tools that place your own numbers against a measured publisher panel. ### Customers https://www.artificialpoets.com/customers/ Case studies from live publisher deployments: the window, the comparison group and the method, with the segments that fell printed next to the ones that rose. ### Blog https://www.artificialpoets.com/blog/ Working notes on measurement, recirculation and what we are finding in publisher data. Written for people who have to defend the numbers internally. ## Resources ### State of AI in Media & Publishing https://www.artificialpoets.com/resources/state-of-ai-in-media-and-publishing/ A free monthly report on how AI is reshaping the publishing web: crawler policy, licensing, the AI already inside publisher infrastructure, and web performance. The download is always the newest edition. ### Pilot Design Template https://www.artificialpoets.com/resources/pilot-design-template/ The four decisions that make a pilot measurable — as a worksheet you complete with the vendor and attach to the agreement. Any section left blank is where the post-hoc storytelling will enter. Section 1 · The split Treatment group. Properties or sections where the product will be enabled.Held-back group. Comparable properties or sections that stay unchanged — the counterfactual.Selection rule. Why these — ideally "entire eligible set" or a stated criterion, never chosen after results. Comparability check: the two groups' weekly trends must move together during the baseline, or the comparison is repaired before the post-period counts. Section 2 · Metric freeze Primary metric and exact definition. e.g. multi-page session share = sessions with >1 pageview ÷ sessions; distinct-article counting stated.Secondary metrics.Counting rules. Bot filter, de-duplication, distinct-content units.Freeze date. Before enablement, in writing. Section 3 · Baseline Instrumentation live from (day zero).Measurement-only period (weeks). Floor: 4.Enablement date — the intervention anchor. A named day. Every later claim hangs on it.Instrumentation faults found and fixed during baseline. Section 4 · Windows and integrity bounds Post-period length. Minimum: one full quarter past launch — long enough to see past the launch bump.Launch window vs steady-state window. Both will be reported.Week-invalidation rule. Tracking gap, migration or retag on either group ⇒ the week is dropped. Panel balance: every property present every counted week, or the week is dropped. Analysis ends at the last clean date for the whole panel — not the calendar. Section 5 · Decision gates Success condition (pre-agreed). Treatment exceeds held-back group by ___ points on the primary metric over the steady-state window.Failure condition — the vendor completes this sentence. "If ___ by ___, it isn't working."Off-test (optional, powerful). A scheduled pause late in the pilot; expected: metric returns toward baseline, recovers on resume.Decision date and owners. ### You Can’t Get the Traffic Back. You Can Get the Session Back. https://www.artificialpoets.com/resources/traffic-back/ The economics of session depth, and the causal evidence behind it. Executive summary Media companies are losing audience to a change they did not cause and cannot reverse. AI-mediated discovery has pushed zero-click searches up roughly 60%, cutting publisher traffic by about 26%. [1] [2] Pages holding the top organic position can lose as much as 60% of their traffic when an AI Overview appears above them. [3] CPMs fell 15–20% over the same period, even as total digital ad investment grew. [4] Most responses to this focus on replacing the lost traffic. This paper argues the opposite: the traffic is not coming back, and the remaining lever is what happens after a reader arrives. The obstacle is structural. Most publishers run static content architectures organised by keyword and category, unable to read the intent behind an individual visit. What elite consumer platforms have done for a decade — predict what a person wants next and serve it immediately — has not been available to publishers at a workable cost. The Artificial Poets Platform closes that gap, combining semantic understanding of content, anonymous interest profiles and real-time intent inference to decide what each reader sees next — without storing personally identifiable information. Deployed on a live publisher property, the Platform produced a causally estimated +53% increase in pageviews per user, with a 95% credible interval of [+48%, +58%] and a posterior probability of a true effect above 99.9%. Downstream engagement improved as a direct consequence. Deployment takes eight weeks. What changed, and why the usual answer doesn't work Publishers have spent two years watching search referrals fall. That conversation is exhausted, and most audience teams have concluded — correctly — that there is little they can do about the top of the funnel. May 2024 marked the turn. Google AI Overviews began answering queries directly, interrupting the organic traffic flow that had sustained media businesses for two decades. [1] Zero-click searches rose sharply, [2] and a page holding the number-one position can now lose up to 60% of its traffic when an AI-generated answer appears above it. [3] Referrals from AI assistants have grown — ChatGPT referrals to publishers rose 25% year over year [1] — but they arrive at a fraction of the volume being lost. Reported figures often understate the damage. Chartbeat has described traffic to large publishers as broadly stable, [5] but that figure depends on including Google Discover, which masks the decline in traditional search. The commercial pressure compounds it. Publishers and ad-tech firms reported CPM declines of 15–20%, concentrated in the fourth quarter, despite aggregate growth in 2024. [4] The result is an environment where every remaining visitor is both scarcer and more valuable. The second problem, which gets less attention When a reader arrives from a search result that now competes with an AI-generated answer, they arrive with a narrower question and less appetite. The path that once produced three or four pageviews produces one or two. Publishers are losing visitors and losing depth from the visitors who remain. The first loss is outside their control. The second is not. Why personalization is the growth engine The economic case is already demonstrated. Sophi, the system built by The Globe and Mail, produced a 222% increase in user registrations and a 51% increase in subscription conversions. [6] Research on modern recommendation systems shows they materially increase engagement, [7] and that correctly applied personalization raises revenue per user by 5–15%, reaching 25% in the strongest cases. [8] Related work finds that when analytics genuinely inform editorial decisions, publishers see measurable gains in both engagement and monetization. [9] Maximising revenue, however, requires more than technique. The experience has to feel relevant and uninvasive rather than mechanical. [10] A reader who notices they are being optimised is a reader you have lost. How the Platform works The Artificial Poets Platform does not bring more traffic. It makes each visit more valuable, which decouples revenue growth from traffic volume. It operates in four stages. There is no redesign, no migration, and no change to how a site is authored. The reader does not click anything. They keep scrolling, and there is something else there. The value cascade The primary mechanism is pageview growth per user. Everything else follows from it. The table below is an illustrative model, not measured customer data. It shows how the mechanism translates at a hypothetical one-million-user scale, at an assumed $10 CPM. Because the model is multiplicative, a publisher with higher-value inventory sees proportionally more. The figure that matters is the pageview lift — the revenue column is arithmetic applied to it. Beyond the revenue line Improvements appear across the digital-experience metrics editorial and product teams already track: bounce rate falls as initial content relevance improves, pages per session rise as journey depth increases, and session duration extends as attention is sustained by dynamically prioritised content. Two organisational benefits arrive alongside them. Data-driven editorial intelligence — insight replaces guesswork, and editorial teams act on real signals rather than intuition. And operational efficiency — automated analytics reduce manual effort in curation, tagging and prioritisation. The evidence Causal impact on pageviews per user A Bayesian causal impact model estimated the isolated effect of deployment by contrasting observed outcomes against a counterfactual scenario in which the intervention did not occur. The analysis used pageviews per user rather than raw pageviews, which controls for traffic volume and isolates genuine engagement change from ordinary fluctuation in visits. The uplift is statistically robust: an average increase of +53% in pageviews per user, and a posterior probability of a true causal effect above 99.9%. The increase is highly unlikely to be random variation. Practical relevance A statistically significant result is not automatically a commercially meaningful one, so a practical relevance threshold was defined in advance — the minimum uplift required to justify deployment economically. Combining the model's posterior effect trajectories with actual daily user counts yields the full posterior distribution of the causal effect, and with it the probability that the effect clears that economic bar. It does, with 99.99% certainty, across every scenario evaluated. The measured uplift exceeds the profitability threshold by roughly seven times at the lower bound of the credible interval. The cascade is causal, not correlational A hierarchical Bayesian framework with full posterior uncertainty propagation was used to estimate the Platform's indirect effect on secondary engagement metrics — effects entirely mediated by its impact on pageviews per user. The natural indirect effect for each metric was estimated by Monte Carlo simulation, quantifying the causal lift attributable exclusively to that indirect route. This is the part that separates the result from a correlation. The secondary metrics did not merely move alongside the deployment — they moved because pageviews per user moved, and the model quantifies how much of each is attributable to that path. Implementation Eight weeks from authorisation to measurable effect. The publisher's contribution is access and a decision. Authorisation to proceed, CMS access, and infrastructure access. That is the list. The Platform is a JavaScript integration. The specifics vary by content management system, but nothing is rebuilt, migrated or re-authored. Integration is carried out by our team using documented standard implementations and is included in the pilot — the publisher's engineering team is not asked for sprint capacity. Where a deployment has genuinely non-standard requirements, forward-deployed engineering is scoped separately. Ad serving is handled the same way. The Platform integrates with Google Ad Manager or an equivalent stack, and can render additional inventory and manage refresh behaviour within it. Legal documentation is prepared per customer, against that customer's own requirements and contract terms. Weeks 1–4 · Instrument and index. The Platform goes live in collection mode on day zero and does two things at once. It builds the behavioural baseline against which every later claim will be measured, and it indexes the site's full content library into embeddings. Indexing runs in the background and does not affect site performance. No AI feature is serving during this period. Readers see no change. Week 4 · Enablement. AI features are switched on, configured against the data collected rather than against defaults. This is the intervention date, and it is the reason the effect can be isolated afterwards. Weeks 5–8 · Tune. The models continue adjusting through reinforcement learning against live outcomes. Effect becomes measurable within this window. Ongoing. Results are available through a dashboard covering engagement, session depth, content performance and infrastructure health. Why the baseline comes first Most vendors switch on and start reporting. Four weeks of measurement before anything changes is what makes the result attributable when it arrives — it is the difference between a number that went up and a number that went up because of something. It also means the enablement date is unambiguous. Every result in this paper, and in the case studies that accompany it, rests on knowing exactly when the intervention happened. Setting expectations You can start small, and we would recommend it. A publisher running several titles can enable on a subset and hold the rest as a comparison group. One portfolio in our data did exactly that — two titles enabled, the rest unchanged — which is why the effect on those titles could be isolated at all. It costs a few months of delay on the titles held back, and it buys a counterfactual. The first quarter is not the run rate. Launch quarters overstate steady state, in this product as in every other. On one measured property the first-quarter effect settled to roughly a fifth of its peak over the following six months, and it is the settled figure that should carry a business case. And it is reversible without residue. On one property the Platform stopped serving for four weeks. Engagement returned to its pre-enablement baseline and recovered within a fortnight of serving resuming — no degradation, no relearning period, no lasting effect on the weeks either side. What this means The new era of digital attention has redefined how audiences discover and consume content, and the consequences for media businesses are unambiguous: declining CPMs, sustained traffic loss, and a rising share of searches that end without a click. Knowing precisely what each reader wants next is no longer a differentiator. It is a requirement. What was until recently exclusive to elite technology platforms is now available to publishers, supported by empirical and causal evidence rather than assertion. You cannot make Google send you the visitors it used to. You can decide what happens to the ones who arrive. Appendix — modelling assumptions The validity of the causal inference rests on four assumptions, stated plainly so they can be challenged. Stable relationship. The correlation between predictors and pageviews held consistently across the pre- and post-intervention periods. No unobserved external shocks. No major external event — traffic spike, platform change or seasonal anomaly — coincided with the intervention. Accurate counterfactual. The pre-intervention model captured baseline dynamics including seasonality and autoregressive structure (AR(1)). Independent predictors. Control variables were not themselves influenced by the intervention, which is what keeps the effect estimate unbiased. References Similarweb, “The Impact of Generative AI on Publishers” (2025).D. Goodwin, “Nearly 60% of Google searches end without a click in 2024,” Search Engine Land (2024).Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results” (July 2025) — the study behind The Guardian’s AI Overviews coverage.IAB / PwC, “Internet Advertising Revenue Report: Full Year 2024” (April 2025).J. Schwartz / Chartbeat data, “Publisher traffic sources 2019–2025 analysed,” Press Gazette (2025).A. Hermida & F. M. Simon, “AI in the Newsroom: Lessons from the Adoption of The Globe and Mail’s Sophi,” Journalism Practice 19(10) (2025).Liu et al. (ByteDance / TikTok), “Monolith: Real Time Recommendation System With Collisionless Embedding Table,” ORSUM @ ACM RecSys (2022).McKinsey & Company, “The value of getting personalization right — or wrong — is multiplying” (November 2021).M. K. Maurya & A. Kumar, “Audience Data and Editorial Decision-Making: Evolution and Applications of Web Analytics in Newsrooms,” Journal of Media Ethics 39(4) (2024).J. Yin, X. Qiu & Y. Wang, “The Impact of AI-Personalized Recommendations on Clicking Intentions,” JTAER 20(1) (2025). Results are from a live publisher deployment; the customer is not named. Property-level results are published separately as named case studies. ### Session ROI Calculator https://www.artificialpoets.com/resources/session-roi-calculator/ Your pageviews, sessions and RPM against the lifts we have measured. ## Case Studies ### Horse & Rider Grew Pages Per Session 19% Without Growing Traffic https://www.artificialpoets.com/customers/horse-and-rider/ In the first quarter after enabling the Artificial Poets Platform, one in six Horse & Rider readers went past the first page, up from one in ten. Three comparable titles in the same network moved less than a quarter of a point over the identical weeks. In brief Horse & Rider enabled the Artificial Poets Platform on 18 December 2025. 63% more readers got past the opening article. The share of sessions reaching more than one page went from 10.4% to 16.9%. Distinct pages per session increase 19.1% in the first quarter. Three sister titles in the same network never switched it on. Over the very same weeks they barely moved — less than a quarter of a percentage point — so the jump is not a seasonal swing or a search-algorithm change. Sessions and users did not grow. The entire gain came from depth per visit. The pressure Every publisher in this market has already had the conversation about traffic. Google search referrals to publishers fell roughly a third in the year to December 2025, and the decline was worst for exactly the kind of title Horse & Rider is: sites in the 1,000–10,000 daily pageview band lost around 60%.[1] Fewer than four in ten news executives say they are confident about the industry they work in, and they expect search referrals to fall another 43% within three years.[2] What gets less attention is the second problem, which is quieter and more expensive. The visits that still arrive are worth less than they used to be. When a reader lands from a search result that now competes with an AI-generated answer, the path that once produced three or four pageviews produces one or two.[3] Publishers are losing visitors and losing depth from the visitors who remain. That second loss is the one still inside a publisher's control. Chartbeat's own data puts it plainly: a first-time reader who views a single page has an 8% chance of returning within the week. A reader who views one more page has a 22% chance.[4] One additional article roughly triples the odds of building a habit, and it does it on an audience the publisher has already paid to acquire. What was enabled On 18 December 2025, Horse & Rider switched on the Artificial Poets Platform: as a reader moves down an article, the Artificial Poets Platform's AI selects and loads related articles beneath it, chosen against the reader's current context rather than a fixed "most popular" list. There is no redesign, no migration, and no change to how the site is authored. The reader does not click anything. They keep scrolling, and there is something else there. Adoption was immediate and then settled, which is the normal shape. In the first full week the Artificial Poets Platform was involved in 14.3% of consecutive pageviews on the site, against a pre-launch baseline of 0.7%. It runs at 3–4% today. What moved Distinct pages per session increase 19.1%, from 1.11 to 1.33, across the first quarter after enablement, measured against a 20-week pre-launch baseline. Weekly, Horse & Rider climbed from around 1.09 the week before enablement to a peak of 1.57, and stayed elevated through March, while the average of three comparison titles held between 1.08 and 1.14 throughout. 63% more sessions reached more than one page. In plain terms: the share went from 10.4% to 16.9% — before the Artificial Poets Platform, roughly one reader in ten went beyond the article they landed on; after it, roughly one in six did. Those readers enter the part of the session where a publisher's economics actually work: every additional page load is a fresh ad opportunity on audience already acquired, and Playwire's session data puts publishers above the median on page depth at many times the revenue per session of those below it.[5] Traffic did not grow. Sessions and users were flat to slightly down across the same period. This is a depth result, not an acquisition result, and it is worth being precise about that: Horse & Rider did not get more readers. It got more out of the readers it already had. How we know The obvious objection to any result like this is that something else changed: a seasonal swing, an algorithm update, a content push. So the comparison was built to answer it. Equine Network runs a portfolio of equestrian titles. The Artificial Poets Platform was enabled on two of them and not on the others, which leaves a set of comparison titles serving an overlapping audience, in the same category, over the same weeks, under the same market conditions. Three independent titles, all effectively flat, while the two enabled titles moved 19.1% and 12.2%. Their multi-page session share moved by an average of half a point across the same weeks. Metrics are counted per site by domain, bots excluded, over identical calendar weeks. Pages per session counts distinct article paths, so the figure reflects readers reaching genuinely different articles. References Chartbeat data reported in Press Gazette, "Global publisher Google traffic dropped by a third in 2025." pressgazette.co.uk Reuters Institute for the Study of Journalism, "Journalism, Media and Technology Trends and Predictions 2026." reutersinstitute.politics.ox.ac.uk Nitropay, "Protecting Ad Revenue Per Session in the Age of AI Search." blog.nitropay.com Chartbeat, "Encourage multiple page visits, use social video for successful 2026 publishing strategy." chartbeat.com Playwire, "Content Architecture and Publisher Revenue: Why Page Depth Is Your Most Underrated Monetization Lever." playwire.com Method: PostHog, all devices, bot-filtered. Baseline 4 Aug – 15 Dec 2025; first-quarter window 22 Dec 2025 – 30 Mar 2026. Multi-page session share compares 4 Aug – 17 Dec 2025 with 22 Dec 2025 – 10 May 2026. Comparison titles are Equine Network properties without the Artificial Poets Platform enabled. ### Mobile Multi-Page Sessions Rose 48%. The Comparison Titles Fell 23%. https://www.artificialpoets.com/customers/equine-network-segments/ We compared titles from the same network running the Artificial Poets Platform against sister titles that were not. Mobile is two-thirds of this network’s traffic. In brief 46% more sessions went past the first page across the enabled Equine Network titles — multi-page share rose from 10.3% to 15.1%. This piece asks where that came from. Mobile rose 48%, from 9.2% to 13.6% — while the comparison group's mobile fell 23% over the same nine months. Social arrivals almost doubled their multi-page rate, up 99% — from 8.6% to 17.2%. On the comparison titles, social fell. The lift held on every device, and on every acquisition channel except direct. Direct traffic declined on the enabled titles, but it declined further on the comparison titles, so we do not attribute it to the Artificial Poets Platform either way. The pressure Any publisher reading a case study asks the same private question: will this work on my audience? It is a fair thing to worry about, because most engagement products work best exactly where you need them least. They lift the readers who were already going to stay (the loyal, the direct, the desktop) and do nothing for the traffic that actually costs you money. The two segments publishers most want moved are the two hardest to move. Mobile. Roughly two-thirds of traffic on these titles, with the smallest screen, the least patience and the worst monetisation per session in the business. It is also where the erosion everyone is describing runs fastest: across the comparison titles, the share of mobile sessions reaching more than one page fell from 11.5% to 8.8% in nine months, with nothing changed on any of them. Social arrivals. They came for one specific thing they saw in a feed, and they leave. Nearly a quarter of one enabled title's traffic arrives this way. So we broke the result apart and looked at where it actually landed. What was enabled On 18 December 2025, two Equine Network titles switched on the Artificial Poets Platform: as a reader moves down an article, the Artificial Poets Platform's AI selects and loads related articles beneath it, chosen against what that reader is actually reading rather than a fixed "most popular" list. Three comparable titles in the same portfolio were left unchanged and measured alongside them across identical weeks. Where the lift landed By device Mobile is the commercially important row. It is about two-thirds of these titles' traffic, and it is where the comparison group lost the most. The enabled titles went the other way by 48%. That is a 71-point swing on the surface where most publishers make most of their impressions. Desktop improved least in relative terms, which is what you would expect: desktop readers already went deeper, so there was less headroom. The pattern holds across all three devices, and it is largest exactly where the headroom was. By acquisition channel Social roughly doubled. A reader arriving from a feed used to go past the first article about one time in twelve. Now it is closer to one in six. That is the segment conventional wisdom says you cannot do anything with, and it moved further than any other channel on the site. Organic search rose 59%. This is the traffic being squeezed hardest by AI answers: referrals to publishers fell roughly a third in the year to December 2025,[1] and the reader who does land arrives with a narrower question and less appetite than they used to have.[2] The Artificial Poets Platform does not bring that reader back, but it does get more out of them once they arrive. Referral rose most in percentage terms, but it is the smallest channel here and the enabled and comparison titles clearly draw referrals from different places. Note the very different starting points. We would not build a case on this row. The exception Direct traffic went down on the enabled titles From 12.7% to 10.3%. We are not going to explain that away. What we can say is that direct fell further on the comparison titles over the same weeks, from 7.0% to 4.6%, a 35% decline against our 19%. Something is happening to direct-arriving readers across this whole category, and it is happening to the enabled titles slightly less. We do not claim the Artificial Poets Platform caused the smaller decline, and we would not want anyone to read it that way. It is one channel out of four, and it is the one where readers arrive already knowing where they are going. How we know Metrics are counted per site by domain, bots excluded, over identical calendar weeks: 4 August to 17 December 2025 against 22 December 2025 to 10 May 2026. Multi-page session share is the proportion of sessions recording more than one pageview. Device is taken from the user agent; channel is assigned at session level. The comparison titles are the Equine Network properties that report continuously across the whole window and were never enabled. The entire eligible set, not a flattering subset. Every segment we measured is shown above, including the one that went against us. References Chartbeat data reported in Press Gazette, "Global publisher Google traffic dropped by a third in 2025." pressgazette.co.uk Nitropay, "Protecting Ad Revenue Per Session in the Age of AI Search." blog.nitropay.com Method: PostHog, all devices, bot-filtered. Pre-period 4 Aug – 17 Dec 2025; post-period 22 Dec 2025 – 10 May 2026. Multi-page session share is the proportion of sessions recording more than one pageview. Segment figures are pooled across the two enabled titles and the three comparison titles respectively, session-weighted. Comparison titles are Equine Network properties without the Artificial Poets Platform enabled. ### Equus Magazine Tripled the Sessions That Reach Five Articles https://www.artificialpoets.com/customers/equus-magazine-deep-readers/ Most publishers measure whether a reader gets to a second article. This is about what happens after that. On Equus Magazine, sessions reaching three articles nearly doubled and sessions reaching five tripled, while four comparable titles in the same network went backwards at every depth. In brief Equus Magazine enabled the Artificial Poets Platform on 18 December 2025. 221% more sessions reached five or more articles, and 91% more reached three or more. In share terms: three-deep went from 1.94% to 3.71%, five-deep from 0.32% to 1.04%. The effect gets larger the deeper you look: +62% at two articles, +221% at five, +443% at ten. Four sister titles in the same portfolio never switched it on. All four went backwards at every depth over the same nine months, down 13% to 22% — so this is not something the whole category was getting anyway. Serving was switched off for four weeks in April. Depth fell straight back to its pre-launch level, then recovered when serving resumed — the result reproduced itself on the same site. The pressure Every publisher in this market is trying to build a direct audience, because the search referrals that used to do that job are going away: down roughly a third in the year to December 2025,[1] with another 43% expected within three years.[2] The problem is where a direct audience actually comes from. Newsletter signups, registrations and subscriptions are not drawn evenly from your traffic. They come from the small population of readers who were already going deep: the ones who read three, five, ten articles in a sitting. That group is the top of every loyalty funnel a publisher has. On most sites it is vanishingly small, and almost nobody measures it. The industry benchmark stops at the second pageview: Chartbeat's data shows a reader who views one more page is 2.75× more likely to return within the week.[3] That is a real and useful finding, and it is also where the measurement ends. Nobody reports what share of their sessions reach five articles, which means nobody knows whether it is growing or shrinking. On the four comparison titles here, it was shrinking — by 13% to 22%, at every depth — over nine months with nothing changed on any of them. What was enabled On 18 December 2025, Equus Magazine switched on the Artificial Poets Platform: as a reader moves down an article, the Artificial Poets Platform's AI selects and loads related articles beneath it, chosen against what that reader is actually reading rather than a fixed "most popular" list. No redesign, no migration, no change to how the site is authored. The reader does not click anything. They keep scrolling, and there is something else there. Equus runs the Artificial Poets Platform's recirculation feature on its own, without the AI recommendation feature its sister title also uses, which makes it a clean read on what recirculation does by itself. What moved The headline is not that more readers reached a second article. It is the shape of what happened after that. The lift compounds with depth. A 62% gain at two articles becomes 221% at five and 443% at ten. That is the signature of a platform that keeps working: a reader who accepts the second article is being offered a third, and a fourth, and the effect accumulates rather than fading after the first hand-off. The comparison titles fell at every single depth, in a tight band between 13% and 22%. Whatever is eroding session depth across this category did not stop at the second pageview either. Be clear about the base. Even after tripling, fewer than 2% of Equus sessions reach five articles. This is a small population getting substantially larger, not a majority behaviour. The three-article measure, 1.94% to 3.71%, is the one to plan against, because it sits on a base big enough to move revenue on its own. Traffic did not grow. Sessions and users were flat to slightly down. This is entirely a composition change in how deep the existing audience goes. How long they spend there Reaching more articles only matters if readers actually read them, so the Artificial Poets Platform measures attention on each article it surfaces directly rather than inferring it. The mean is 76 seconds, and each served article holds a reader for around 30 seconds. Not a glance, and not a bounce. Why it compounds The same telemetry shows what happens at each step of a scroll sequence, and it explains the shape of the depth result. The hard part is the second article. Only about a third of readers who get one served article accept another. But the readers who reach a fourth go on to a fifth 70% of the time, and past that it runs between 75% and 82% at every step. Persuading someone to start is difficult; persuading someone already reading their fifth article is close to automatic. Attention does not decay to zero. It settles. Time per article drops sharply from the first to the third, then flattens at around 18 to 19 seconds and stays there. The eighth served article holds a reader almost exactly as long as the fourth does. Together those two patterns are why the lift compounds with depth. The Artificial Poets Platform's work is front-loaded, and everything after it is a reader who has already decided to keep going. How we know The comparison titles. The Artificial Poets Platform was enabled on two Equine Network titles and not on the others, leaving four comparable properties serving an overlapping audience, in the same category, over identical weeks. All four declined at every depth measured. They are the entire eligible set: every Equine Network property that reports continuously across the window and was never enabled. Not a selected subset. A pause in serving. For four weeks from 13 April, the Artificial Poets Platform did not serve on Equus. Over exactly those weeks, session depth returned to its pre-enablement baseline. Serving resumed on 11 May and depth recovered within a fortnight. Two independent signals date the pause identically: the share of sessions the Artificial Poets Platform served in, which ran at 14.7% before, 0.0% for four consecutive weeks, and 8.4% after; and a behavioural fingerprint in pageview timing that collapsed and returned over the same weeks. That combination, a control group and a within-site pause, is not evidence most vendors can put in front of you. One thing we did not expect Over the same period, mobile page experience on Equus improved: largest contentful paint at the 75th percentile fell 24%, against a 10% improvement on the largest comparison title. We are not claiming the Artificial Poets Platform caused that, and we would not want it read that way. The site may have had separate performance work in flight. What we will say is that the common objection to infinite-scroll products, that they cost you page speed and your search rankings with it, did not happen here. Desktop was flat. References Chartbeat data reported in Press Gazette, "Global publisher Google traffic dropped by a third in 2025." pressgazette.co.uk Reuters Institute for the Study of Journalism, "Journalism, Media and Technology Trends and Predictions 2026." reutersinstitute.politics.ox.ac.uk Chartbeat, "Encourage multiple page visits, use social video for successful 2026 publishing strategy." chartbeat.com Method: PostHog, all devices, bot-filtered. Pre-period 4 Aug – 17 Dec 2025; post-period 22 Dec 2025 – 10 May 2026. Depth is counted as distinct article paths within a session. Comparison figures are pooled across four Equine Network titles without the Artificial Poets Platform enabled, session-weighted. Serving pause 13 Apr – 10 May 2026. Web vitals are 75th-percentile values from the same windows. ### How Equine Network Grew Multi-Page Sessions 46% https://www.artificialpoets.com/customers/equine-network-rollout/ Equine Network enabled the Artificial Poets Platform on two of its titles and left the rest running unchanged. Over the following nine months the enabled titles grew multi-page sessions 46%. Across the comparison titles, the same measure fell 16%. In brief Equine Network enabled the Artificial Poets Platform on two of its titles on 18 December 2025, leaving the rest of the portfolio unchanged. Across the enabled titles, the share of sessions reaching more than one page had a 46% increase. Across three comparable titles in the same portfolio, the same measure fell ~16% over identical weeks. Distinct pages per session: +14.9% on the enabled titles, −0.3% on the comparison group. The decline in the comparison group was not uniform. One of the three edged up. We show all of it below. The pressure Publishers have spent two years watching search referrals fall: roughly a third in the year to December 2025,[1] with another 43% expected within three years.[2] That conversation is exhausted. Most audience teams have concluded there is not much they can do about the top of the funnel. The erosion further down is less discussed and more actionable. When a reader arrives from a search result that now competes with an AI-generated answer, they read one or two pages where they used to read three or four.[3] Publishers are losing visitors and losing session depth from the visitors who remain. Most publishers have to take that second trend on faith, because they have nothing to compare themselves against. Equine Network is unusual: it runs a portfolio of titles serving an overlapping audience in a single category, which means the erosion is measurable inside its own walls. It is. Across three of its titles, the share of sessions reaching more than one page fell 16% between August 2025 and May 2026 — a nine-month decline with no change in strategy, staffing or content operation on any of them. That is the counterfactual. It is what the enabled titles would most likely have done. The problem with rolling anything across a portfolio A network operator has a harder version of every vendor decision. Enabling something across nine titles means nine integrations, nine sets of ad-stack interactions, and nine audiences exposed to a change that cannot be quietly reversed if it disappoints. The claim on the slide is never the issue. The issue is that by the time you can evaluate it, you have already committed. Equine Network took the other route. It enabled Artificial Poets Platform on two titles and left the rest alone. That decision is the reason this case study contains anything worth reading. It produced a measurement design most vendors can never offer, because most customers roll out everywhere at once and then have nothing to compare against. What was enabled On 18 December 2025, two Equine Network titles switched on the Artificial Poets Platform: as a reader moves down an article, the Artificial Poets Platform's AI selects and loads related articles beneath it, chosen against what that reader is actually reading rather than a fixed "most popular" list. No redesign, no migration, no change to how either site is authored. What moved Across the enabled titles, multi-page session share rose 46%. Just over one session in ten used to go past the article it landed on. Now it is roughly one in six and a half. Across the comparison titles, the same measure fell 16%. That is the erosion the whole industry is describing, visible inside a single portfolio over the same weeks. The gap between those two trajectories is the result. Not 46% on its own, and not −16% on its own: the fact that two groups of titles in one company, serving one audience, in one category, moved in opposite directions after exactly one thing changed on two of them. Distinct pages per session tells the same story with less volatility: +14.9% across the enabled titles, −0.3% across the comparison group. Traffic did not grow. Sessions and users were flat to slightly down on the enabled titles. This is a depth result, not an acquisition result. The comparison group's decline is not uniform We would rather show that than average it away. One comparison title accounts for most of it; the second is essentially flat; the third edged up three quarters of a point. The pooled figure falls 16% because the declining title is much the largest of the three by session volume. What is uniform is the separation. Both enabled titles gained more than any comparison title, by a margin of at least four times. How we know Metrics are counted per site by domain, bots excluded, over identical calendar weeks: 4 August to 17 December 2025 against 22 December 2025 to 10 May 2026. Distinct pages per session counts unique article paths, so it reflects readers reaching genuinely different articles rather than any artefact of how pageviews are recorded. The comparison titles were not selected after the fact. They are the Equine Network properties that report continuously across the whole window and were never enabled. The entire eligible set, not a flattering subset. References Chartbeat data reported in Press Gazette, "Global publisher Google traffic dropped by a third in 2025." pressgazette.co.uk Reuters Institute for the Study of Journalism, "Journalism, Media and Technology Trends and Predictions 2026." reutersinstitute.politics.ox.ac.uk Nitropay, "Protecting Ad Revenue Per Session in the Age of AI Search." blog.nitropay.com Method: PostHog, all devices, bot-filtered. Pre-period 4 Aug – 17 Dec 2025; post-period 22 Dec 2025 – 10 May 2026. Multi-page session share is the proportion of sessions recording more than one pageview. Distinct pages per session counts unique article paths, measured 4 Aug – 15 Dec 2025 against 22 Dec 2025 – 30 Mar 2026. Pooled figures are session-weighted. Comparison titles are Equine Network properties without the Artificial Poets Platform enabled. ## Optional - [Sitemap](https://www.artificialpoets.com/wp-sitemap.xml) - [REST API](https://www.artificialpoets.com/wp-json/)