Google has been quietly scaling a pilot that pays publishers when their content “significantly” contributes to AI-generated responses across Gemini, AI Overviews and AI Mode, Digiday reported on 14 September. Publishers that opt into the “AI contribution” pilot see a monthly earnings figure in a new Search Console widget — with no detail on how payouts are calculated. Google confirmed it is an early-stage learning pilot and pointed to testing it first described in a June policy blog post.
At least dozens of publishers have been approached, skewing small and mid-sized and extending well beyond news. Participants describe weekly calls with Google and a black-box formula; one source put early returns at “peanuts” relative to advertising revenue, and one executive said offers fell below the threshold needed to opt in, warning that participation could weaken publishers’ leverage for larger terms.
Google’s pay-per-value pilot is the first time a dominant platform has conceded, in product form, that a content contribution to an AI answer is a billable event — and it has done so on terms only Google can see. The precedent matters more than the payout.
This reasoning traces back to years spent building broadcast attribution systems, where an entire royalty economy depended on an independent record of what aired, when and where; rights holders who relied on the broadcaster’s own logs were paid on the broadcaster’s arithmetic. A Search Console widget showing a single monthly number is the same structure: the platform measures, the platform values, the platform pays. The same pattern appears in readiness assessments across media organizations — the ones with their own telemetry negotiate on evidence, and the ones without it negotiate on hope.
The publishers building durable positions are those figuring out how to deliver locally-contextualized intelligence at platform scale — precisely the content a value-weighted model should reward most, if anyone other than Google could verify the weighting.
The organizations handling this well are treating the pilot as a data-gathering exercise rather than a pricing benchmark, and pairing any opt-in with independent logging of AI crawler and grounding activity. Whether opting in makes sense for a given publisher depends on where it stands on measurement — that is the conversation worth having.
The open question is whether ‘value’ will ever be defined anywhere other than inside the payer’s own model — and whether a black-box number accepted now becomes the market rate later.
What this signals: AI grounding is being priced as a usage event, but the measurement, valuation and payment all sit on one side of the table.
For any publisher in active licensing discussions, the window to make independently verifiable usage data a condition of participation is now — before a platform-set number hardens into the benchmark.
Axios will take Axios Direct to market this month: licensed feeds of its reporting built for AI models and agents, its chief revenue officer said at the Digiday Publishing Summit on 16 September. The first feed targets investment firms, priced annually by company size and assets under management and sold on two-year terms; a second will feed corporate internal AI models; a third, for subscribers’ personal agents, is slated for 2027. Axios says it hit its 2026 revenue goal in early September, with roughly 52–54% of revenue sold client-direct.
The Operator’s Take: Axios is pricing AI access by the buyer’s economic stake, not by tokens or crawls — a direct-to-agent structure that keeps both the telemetry and the customer relationship with the publisher rather than a marketplace.
Commercial · Economic Model Benchmark direct AI-feed pricing on customer value, not crawl volume.
TIME has deployed two agentic applications from Kana across its advertising and sales organizations, the vendor announced on 22 September. A Campaign Orchestrator monitors delivery, pacing and performance on every live campaign to catch under-delivery before it becomes a make-good; a Media Proposal Generator, rolled out to the full sales team, turns inbound RFPs into inventory-aware proposals in minutes rather than days. Both run inside guardrails TIME controls directly — budget limits, pricing guidance and approval thresholds that decide when an agent acts alone and when it escalates to a human.
The Operator’s Take: The detail that matters is the guardrail layer: approval thresholds and escalation rules are what make agentic ad operations auditable. Deploying agents without that governance layer automates risk, not revenue.
Build · Economic Model Define approval thresholds and escalation rules before deploying sales or ad-ops agents.
Filings unsealed on 17 September in the New York Times’ case against OpenAI and Microsoft go further than the executive quotes first reported: OpenAI’s head of ChatGPT described its products as “largely substitutive” for publishers, Microsoft recorded 83–93% drops in click-through rates for the Times’ and Daily News’ domains on Copilot versus Bing, and the record references a workaround for the Times’ paywall. Microsoft’s CEO testified that chatbots substitute for visiting the source. Lawyers told Digiday the evidence strikes at fair use’s market-harm factor; both sides have moved for summary judgment.
The Operator’s Take: The 83–93% click-through collapse was measured by the defendant’s own instrumentation. Publishers who want that class of evidence in their negotiations need to generate it themselves rather than wait for discovery.
Legal · Industry Positioning Brief counsel on the substitution and paywall evidence as it bears on your own licensing and litigation posture.
Three weeks after the court’s 2 September ruling rejected a breakup of Google’s ad tech business in favor of behavioral remedies, seven senior publisher revenue and sell-side executives scored the now-public remedies for Digiday, averaging 5 out of 10. Separate contracts for AdX and DFP were judged closer to a billing change than a competition reset; data portability was questioned because few competing publisher ad servers remain; and the remedies stop at open-web display, leaving CTV, in-app and retail media publishers with nothing.
The Operator’s Take: Remedies confined to open-web display address the last war. The next intermediation fight — who controls the pipe between publisher content and AI answer engines — has no equivalent conduct rules yet.
Advocacy · Industry Positioning Press trade bodies to extend intermediary conduct rules to AI answer engines.