--- title: One engine that knows every story you've published — and every reader who shows up. description: "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." url: https://preview.artificialpoets.com/platform/ site: Artificial Poets type: page date: 2026-08-09T06:31:06+00:00 modified: 2026-08-16T02:43:53+00:00 --- # The Artificial Poets Platform New [The nine-month rollout read: +46% vs −16%](/blog/) **For publishers and media networks losing search traffic** # One engine that knows every story you've published — and every reader who shows up. That is what becoming AI-native actually means: the decisions stop being guesses. The Platform learns what each piece of your archive is about and what each visitor came for, then acts on it — the next story to put in front of them, the search result that answers the real question, the right moment to ask for the signup, the gap in your coverage worth commissioning. It runs inside your site, your CMS and your ad stack. On the titles running it, multi-page sessions grew 46% while comparable titles fell 16%. [Book a session-depth read](/request-a-demo/) ## The traffic is a market condition. The session is a product decision. How it works Google search traffic to publishers fell a third in the year to November 2025 (Chartbeat, via Press Gazette), and surveyed media leaders expect a further decline over the next three years (Reuters Institute, January 2026). What is still yours is what happens after a reader lands: whether the visit ends on the page it started on, or becomes a session. That is the number we move. Four stages, running inside your CMS and your ad stack. This is where the AI is — in the mechanism, not the adjectives. Getting started asks almost nothing of your team. You give us the go-ahead and access to your site; we do the work. For the first four weeks nothing is served to readers — we only watch, so we know what your numbers looked like before. Then we switch it on, on a date you pick. - **Content vectorization** — Language models embed the intent and context of every article in your library — matching on meaning, not keywords or tags. - **Reader fingerprinting** — An anonymous, privacy-first interest profile per reader, built in real time. No PII stored. GDPR/CCPA-compatible. - **Intent matching** — The Platform infers what this visit is for — informational, navigational, transactional — from behavioral signals. - **Content orchestration** — The next article is selected and served as the reader scrolls. Served pages render your ad slots, inside your existing stack. ## What your readers actually see One engine, showing up in the places that decide whether a visit ends or continues: the story that appears when they finish reading, the recommendations under the article, the answer their search should have found, the signup asked at the right moment — and, for your team, a clear read on what your audience wants next. - **Use case: The infinite article feed** — The next article is simply there — the mechanism behind the +46%. - **Use case: Recommendations that re-render per reader** — Three readers, one article, different next steps. Try the demo. - **Use case: Search that understands intent** — The query is a reader telling you what they want. Serve it. - **Use case: Ask at the right depth** — Newsletter and registration asks timed to reading behavior. - **Solution: Defend the ad line** — Depth is the one revenue lever still entirely yours. - **Use case: See what readers want before you commission** — The interest graph as an editorial instrument. [All seven solutions](/solutions/) · [All nine use cases](/blog/) ## Does it actually work? Search sends fewer readers every month, and the ones who arrive leave after one page. These are the numbers from titles that ran the Platform against comparable titles that didn't — including what the lift settled at once the novelty wore off, because that is the number you would budget against. - **+46% vs −16%** Multi-page sessions, enabled vs comparison titles, nine months - **+19.1% → +4.0%** Pages per session on the largest title: first quarter, then steady state — plan on the second - **+48% vs −23%** Multi-page sessions on mobile — the hardest segment — enabled vs comparison titles, nine months ## What we can't claim The rest of the record — including the segment that fell on enabled titles — is in the published case studies. - **This won't fix your search traffic.** — On the measured network, sessions and users were flat — the result is depth per visit, and we say so. - **The launch quarter is not the run rate.** — +19.1% decayed to +4.0% at steady state on the largest title. Plan on the second number; we print both. - **Deep-session growth starts from a small base.** — Sessions of five or more articles grew +221% over the nine months — and still fewer than 2% of sessions reach five articles. - **Not every channel rose.** — Direct visits fell on the enabled titles too: −19%, against −35% on comparison titles. Six of seven segments improved — every segment is printed. - **Not every title fits.** — Pre-monetization properties and single-article utility traffic won't see these numbers. Start with the assessment instead. ## Who this is running on A publisher with nine special-interest titles. Two of them switched the Platform on; the rest carried on exactly as before, which is what makes the comparison fair. We have been measuring all of them, every week, since August 2025 — and the case studies show the numbers title by title. - [Case study: Pages per session: +19.1% peak, +4.0% steady state](/blog/) — The depth flagship, with windows. - [Case study: +46% vs −16% — the rollout](/blog/) — Nine months against every unchanged, continuously-reporting title. - [Case study: Sessions reaching 3+ articles: +91%, nine months](/blog/) — Small base, printed — and the attention curve behind it. - [Case study: Mobile multi-page sessions: +48% vs −23%](/blog/) — Where readers are hardest to keep — every segment printed, including the one that fell. ## Tools, guides and the research Use them without talking to us. The methodology is public because checking our numbers is the fastest way to trust them. - [Tool: Session Depth Assessment](/blog/) — Your single-page rate against the published bands, in thirty seconds. - [Tool: Session RPM Calculator](/blog/) — The measured lift range, on your pageviews and your RPM. - [White paper: You Can't Get the Traffic Back. You Can Get the Session Back.](/blog/) — The thesis, the mechanism, and the causal analysis. - [Guide: The Vendor Claim Audit Checklist](/blog/) — 15 checks to run on any engagement vendor's numbers. Including ours. - [Article: The Second Pageview Is the Last Metric You Still Control](/blog/) — The full market argument, sourced. - [Article: How to Check an AI Vendor's Numbers — Starting With Ours](/blog/) — The six rules we hold ourselves to, as a buyer's checklist. ## See it running on a live title A working session on a real publisher site: the reader scrolls, the Platform chooses what loads beneath the article, and you watch session depth move with them. Live product, no deck. [Book my demo](/request-a-demo/)