Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Google Pays About 100 Publishers for AI Overviews, Some Get Under $1,000

Google is paying roughly one hundred publishers, blogs and websites for content in its AI Overviews. The sums range from less than $1,000 to more than $1 million per outlet, according to The Information, summarised by heise online on 1 October.

Media & internetNewsGrace OkonkwoPublished: 1 October 20265 min readSources 15
Google Pays About 100 Publishers for AI Overviews, Some Get Under $1,000

The pilot has no published price list, eligibility rules, or total budget. Per The Information, the spread is wide enough to look like a set of individual deals rather than a rate card. One publisher told the outlet it had received about $50,000 to $60,000 so far. It described the sum as immaterial to its overall business. An earlier participant said it had cleared more than $1 million, a much larger share of its revenue.

Smaller sites did worse. Over several months, some received less than $1,000. For several small and mid-sized publishers, the payments worked out to less than 0.1% of their advertising revenue, according to The Information's research. Heise online published its summary of that reporting on 1 October.

Some publishers are holding out

The low end of the range has reportedly kept some larger media houses out of the programme altogether. By staying on the sidelines, they hope to push Google and other AI vendors toward higher payments, heise reports.

Other publishers treat the pilot as reconnaissance. They use it to learn which pages Google's AI values most, then reshape their editorial strategy around that signal. Google is not currently sharing that data, heise says. Participants who track earnings through the Google Search Console often do not know how individual amounts are calculated, according to Digiday. No public formula explains the gap between what a large publisher and a small one receive.

No public formula explains the gap between what a large publisher and a small one receive.

The context matters. Google's historical position was that publishers did not need to be paid for search visibility because the engine sent traffic back to their pages. Since AI Overviews began appearing at scale, that traffic has visibly shrunk. The pilot runs alongside growing regulatory pressure. Google currently faces several lawsuits on the matter, heise notes.

Google also appears to be pressing publishers on AI licensing. According to heise, anyone wanting to join the new AI pilot programme for Google News is expected to grant the company extensive rights to use their content. In some cases, this is a condition for continuing existing payments.

The courts have moved too

On 1 October, the same day as the German summary, a US federal judge dismissed antitrust lawsuits brought by Chegg and Penske Media over Google's AI Overviews, according to Reuters. Search Engine Journal reported that the judge acknowledged publisher harm but still dismissed the claims. Press Gazette framed the ruling around the absence of a formal bargain between the parties. Search Engine Roundtable put it as: the lawsuit was dismissed because there was no formal agreement with publishers.

That distinction cuts to the heart of the pilot. If Google has no formal bargain with publishers, it can set payments unilaterally, vary them by outlet, and decline to explain the method. The Information's figures, which surfaced on 29 September and were widely picked up the following day, are the closest thing to a public accounting so far. Ars Technica reported on 30 September that the early attempt to pay websites for AI answers is struggling. The Verge covered the same test on 30 September.

Publishers are not waiting for the experiment to mature. Digiday reported on 1 October that publishers are turning generative engine optimisation, GEO, from an experiment into a business line. They are hiring for it and selling it as a service. That is a hedge. If referral traffic keeps falling, being cited inside an AI answer becomes a channel in its own right, with its own metrics and sales pitch.

The traffic numbers behind that hedge are ugly. ADWEEK reported on 28 September that Chartbeat data shows Google referrals down 36% year over year. Publishers are pushing audiences toward owned channels instead. Search Engine Roundtable flagged on 25 September that Google's September 2026 spam update had landed, alongside changes to Discover traffic and new multimodal search reporting. On 1 October, TechJuice reported that AI Overviews are replacing brands in search results even more often than before.

What the payments actually buy

For the publishers inside the pilot, the practical question is not the headline figure but the contractual one. Heise describes a programme in which participants may not know how their individual amounts are calculated. They cannot see the valuation Google places on specific content. In the Google News case, they may face broad content rights as the price of continued payment. That is a weak negotiating position. The reported refusal of some large publishers to join looks like an attempt to improve it collectively rather than individually.

Google has not published the number of participants, the total spend, or the terms. The Information's reporting puts the participant count at roughly one hundred. Everything else, including the gap between the $1 million publisher and the sub-$1,000 ones, comes from individual accounts given to that outlet or relayed by heise online, Investing.com, PPC Land and others on 30 September and 1 October.

A second track runs in parallel. Mashable SEA reported on 30 September that Condé Nast, Hearst and around 300 other publishers are pressing Congress to ban bot scraping. That is a legislative route to the same goal the pilot is meant to address: compensation for content that feeds AI answers. It is also a slower one.

For now the pilot remains a test with an unknown denominator. The Information's figures describe a distribution, not a rate. Some outlets clear seven figures; others clear less than a thousand dollars over months. Google has not said which is which, or why. The judge who dismissed the Chegg and Penske claims on 1 October did so without requiring it to.

Comments 0

Sources

15
  1. 01KI-Übersichten: Google vergütet Publisher – Spanne offenbar enormEN
  2. 02OpenAI and Synopsys team up to build an AI model that designs chips like a seasoned engineerEN
  3. 03The Robot Report parent Arrowfly launches AI for Engineers platform, events for engineers navigating AIEN
  4. 04Agencies publish resolution plan feedback letters for 15 banking organizationsEN
  5. 05Valor, Atreides, and Sequoia back AI startup Flow Engineering at $750M valuationEN
  6. 06Canada Rocket Company plans first large-scale engine test site in London, OntarioEN
  7. 07Micron lawsuit claims Chinese memory maker YMTC poached its engineers, then sued it using its own stolen techEN
  8. 08Gears of War E-Day is an uncharacteristically CPU-heavy Unreal Engine 5 gameEN
  9. 09Trace.Space launches Trinity to tackle hardware engineering's next bottleneckEN
  10. 10Launching Vespper DOCX MCP: 3× faster, 2× cheaper, more accurateEN
  11. 11How to Develop Software Engineering Skills in the Age of AIEN
  12. 12Anatomy of a (Postgres) Search EngineEN
  13. 13SereneDB launches superfast S3 data search engineEN
  14. 14AI Security Is an Engineering Problem — How to Solve It at Every Layer of the Agent StackEN
  15. 15Engineering the Substation Exit for Reliability, Capacity, and ExpansionEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.