---
title: "Content Recommendation Engine for Publishers: Fix the Second Click"
description: Related-articles widgets show everyone the same links. A content recommendation engine that picks per reader — measured +46% vs −16% on multi-page sessions.
url: https://preview.artificialpoets.com/use-cases/next-article-recommendations/
site: Artificial Poets
type: page
date: 2026-08-09T19:46:31+00:00
modified: 2026-08-18T02:10:14+00:00
---
# Fix the Second Click

**Measured on live deployments**

"My related-articles widget shows everyone the same five links — what would a smart second click look like?" A smart second click is a different set of links for each reader of the same article. The Platform fills end-of-article and in-page modules per reader, matched on meaning and intent rather than last week's popularity: +46% multi-page sessions on enabled titles while comparable titles fell 16%.

[See it on my numbers](/request-a-demo/)

Horse & Rider · Ride TV · Equine Network · Equus Magazine · The Score · Equine Network Lockup

## The problem, in your numbers

Internal links already carry more of the load — internal traffic rose from 38% to 41% of publisher pageviews in a year (Chartbeat). The question is whether the module choosing the second click knows anything about the reader in front of it. A most-popular widget treats every reader as the average reader, and the average reader leaves.

- **3–4 → 1–2** Pageviews per arriving visit under AI search (Nitropay)
- **6.2%** Sessions that reached a second article before enablement, on the network we measure
- **−16%** Multi-page sessions over nine months on titles that changed nothing (our panel)
- **8% → 22%** Week-one return rate when the visit includes one extra page (Chartbeat)

## Three readers, one article, two widgets

What each widget serves under the same article. The static widget is the same for all three people.

- **Most popular (static)** — This week's most-read story, the piece everyone shared last month, the year's most popular article — ranked by pageviews, refreshed weekly, identical for every reader.
- **Researching a purchase** — Budget or premium: what the price gap actually buys. The 12-point checklist before you commit. Five options compared, after a season of use — decision-stage content for a reader weighing options.
- **Deep in a how-to** — The step most guides skip, explained. Three mistakes that show up at step four — and the fixes. Once the basics hold: the advanced version — the next step in the task this reader is mid-way through.
- **First-time visitor from social** — Start here: the guide our readers begin with. The one story to read if you only read one. What this site covers, in five minutes — cold start, matched on the arrival article's meaning.

## The evidence, at its honest tier

- **Multi-page sessions** — +46% on enabled titles (10.3% → 15.1%) vs −16% on comparison titles (12.0% → 10.1%) over nine months
- **Largest single-title lift** — +19.1% in the first quarter, +4.0% at steady state — both printed; plan on the second
- **During a four-week pause in serving** — the measured lift fell from +12.2% to −2.7%, then returned to +6.9% when serving resumed — the effect follows the serving
- **Readers stay by choice** — ~30 seconds median added reading per served article, ~50 seconds per engaged session
- **Traffic did not grow.** — Sessions and users were flat — this is a depth result, and we say so

The single-title version of this, measured: [Horse & Rider grew pages per session 19% without growing traffic](/customers/horse-and-rider/).

**Introducing**

## Artificial Poets Platform

One engine behind every solution on this site. It learns your archive and your readers, then acts inside your CMS, your templates and your ad stack.

- It learns your archive: **Every story you have published, current again** — The engine understands each piece by what it is about, not when it ran or where it was filed. A feature from 2019 competes for the next slot on merit with one from this morning.
- It reads the visit: **What a reader wants, without asking** — Interest builds from what someone actually does in the session. No login, no third-party cookies, nothing leaving your domain. Useful on the second pageview, not the tenth visit.
- It chooses: **The right next read, not the popular one** — Someone comparing products and someone following a running story want different things. A most-read list gives both the same five links and serves neither.
- It serves: **There before the reader leaves** — The feed, the recommendations, the search answer and the signup ask all run on the same engine, in your templates and your ad stack. Any slot that arrives with them is yours to sell.
- It proves: **A lift you can defend, or we say so** — Every deployment runs beside titles that did not get it, plus a serving pause. That is how a result becomes a number you can take to a board instead of a vendor claim.

## FAQ

### Our content-discovery network pays a guaranteed check — why trade contracted revenue for a projection?

Don't trade on a projection; measure the gap. Recommendations here draw only from your own library — recirculation, not made-for-advertising inventory, so brand suitability is structural. The engagement side is measured against a comparison group, not read off a correlational dashboard; the revenue side stays your own arithmetic — your session RPM against the depth change the pilot measures — and the baseline runs before any contract decision.

### How is this different from the related-posts module our CMS already ships?

A related-posts block matches metadata: same category, same tags, newest first, identical for every reader. This matches what each article is about against what this reader has shown interest in during this session. Same slot, different mechanism, and the difference is measurable rather than assumed.

### Most of our visitors are first-time arrivals. What do you know about them?

Nothing when they land, and enough by the second pageview. Interest builds from what the reader does in the session, so the first recommendation is matched on meaning from your library and sharpens from there. There is no login, no third-party cookie and no waiting for a tenth visit.

### What does the widget weigh — will it tax Core Web Vitals?

It is a JavaScript integration, and vitals are part of the measurement plan, not an externality. On the measured deployments, mobile page experience improved over the same window — not claimed as caused; claimed as not harmed. If vitals degrade during baseline or after enablement, the dashboard says so.

### Can editors override it — and what keeps off-brand stories out of the module?

Editors pin, exclude, and set adjacency rules per section; sensitive or legally reviewed stories can be blocked from serving entirely. Nothing arrives from outside — the candidate pool is your archive. On breaking news, selection stays contextual until a reader's profile earns weight.

### Does anything learned from our readers serve a competitor's pages?

No. Models are not pooled across customers. What the engine learns from your audience is used to serve your titles, and your behavioural data and content stay yours.

### What happens when we leave?

Your content and behavioural data remain yours, and nothing learned from your readers survives the relationship as someone else's advantage. The integration is a JavaScript tag, so removal is removing a tag rather than unpicking a migration.

## Related

- [Two Titles Grew Multi-Page Sessions 46%. The Comparison Group Fell 16%.](https://preview.artificialpoets.com/customers/equine-network-rollout/) — Equine Network enabled the Platform on two titles and left the rest unchanged. Over nine months the enabled titles grew multi-page sessions 46%; across the comparison titles the same measure fell 16%.
- [The Second Pageview Is the Last Metric You Still Control](https://preview.artificialpoets.com/blog/second-pageview-metric/) — Session depth is the share of your sessions that reach more than one page — and in 2026 it is the only major publisher metric whose fate is still decided on your own site. You cannot make search send the readers it used to send. You cannot reprice a programmatic market that is shedding volume. What you can decide is what happens in the thirty seconds after a reader lands.
- [Serve the Next Article Before They Leave](https://preview.artificialpoets.com/use-cases/infinite-article-feed/) — "What happens after a reader finishes an article — can the next one just be there?"
- [Session Depth Assessment](https://preview.artificialpoets.com/resources/session-depth-assessment/) — Two numbers from your analytics place your single-page rate and depth distribution against a continuously measured publisher panel.

## See how this works on your titles

Thirty minutes on your analytics. We tell you what share of your sessions stop at the first page, and what this would realistically move first for a setup like yours. You leave with the annotated read. No deck.

[See it on my numbers](/request-a-demo/)

How these numbers are made: [the measurement method](/solutions/measurement/). Figures from a multi-title publisher network measured continuously, August 2025 to May 2026. Last updated August 15, 2026.

## Questions this page answers

### What are personalized next-article recommendations?

They are end-of-article and in-page modules that show a different set of links to each reader of the same article, matched on meaning and in-session intent rather than on popularity or shared tags. On the network we measure, multi-page sessions rose 46% on enabled titles while comparable titles fell 16%.

### How is this different from a related-articles widget?

A related-articles module matches metadata: same category, same tags, newest first, identical for everyone. This matches what each article is actually about against what this reader has shown interest in during this session. The behavioural difference is measurable: continuation past the fourth served article runs 70 to 82%.

### How much does a recommendation module cost?

There is no public price list; pricing follows network size and formats. It replaces a module you already run rather than adding a system to maintain, deploys as a JavaScript integration handled by our team, and starts with a four-week measurement-only baseline so you can judge it against your own before.

### Who is it for?

Publishers running a related-posts module that has never been evaluated: the block at the end of every article that shows the same five links to everyone. If you cannot say what your current module contributes, this is the surface where that becomes measurable, and where the second click is either won or lost.

### What are the alternatives?

The CMS related-posts block is free and matches on tags. Content-recommendation networks pay for the slot but send readers off your site. Building per-reader matching in-house means an ML team and 18 to 24 months. This serves only your library, on your domain, with your behavioural data staying yours.

### How do I get started?

The modules go where your current ones sit, in your own templates. Authorization, CMS access and infrastructure access are the requirements, integration is carried out by our team, and four weeks of measurement-only baseline precede enablement on a named day. Eight weeks to a measurable effect.