News & Media

Meta writes AI scraping rights into its business terms as Google widens its paid news pilot

Written by Paul E. Wright | Oct 19, 2026, 1:25:21 PM

Meta’s new business terms let it crawl and train on news sites with no opt-out, while Google widens its licensed news pilot to more than 400 publications.

AI Content Licensing, Provenance & Deal Intelligence

Meta's new commercial terms will allow the scraping of news sites, legal experts say

Meta's new commercial terms, effective October 30, let it send crawlers, bots and spiders to any site "associated with" a business account, then collect that content to provide, develop, improve, research and train its products. The terms define "AI at Meta" as one of those products, and they also let Meta create "test credentials" to get into these sites. Business accounts include anyone who manages Facebook pages or groups, runs ads or uses Meta's business dashboard, which Nieman Lab says covers virtually every news organization.

Meta alerted users in early October, and business accounts cannot opt out before the deadline. Those who disagree are told to stop all business use of Meta's products. Legal experts told Nieman Lab a court may not enforce the terms, but publishers should assume it would. Meta declined to answer the outlet's questions.

Read the full article →

The Operator’s Take

Meta did not ask for a license; it wrote one into the terms every newsroom already accepts to reach its audience.

  • Consent by dependency.To keep reaching readers on Facebook and Instagram, a publisher must accept scraping and training rights; the only exit is to give up the channel.
  • Tying.One law professor compared the structure to antitrust tying, the same argument aimed at Google for linking AI use to appearing in regular search.
  • No evidence trail.With test credentials and periodic pulls permitted, a publisher working from standard logs may not be able to tell which visits are Meta's or what was taken.

I learned this building an oral-history video archive where people's voices were the content: consent and attribution are the product, not an afterthought. Publishers have until October 30 to decide what they are consenting to, and the first step is knowing which of their sites and accounts the terms reach.

  • LegalIndustry Positioning
    Have counsel map every business account and site the October 30 terms would reach, and decide before the deadline whether any account can be retired.
  • CommercialDeal Execution
    Raise Meta's crawl-and-train permission in any current platform talks; a free grant in the terms weakens the case for paying for the same rights.
  • BuildDetection & Telemetry
    Start logging Meta-identified crawler traffic and any test-credential sign-ups now, to have a baseline before and after October 30.
AI Content Licensing, Provenance & Deal Intelligence · 15 October 2026

Google's News AI pilot expands to more than 400 publications, adding CNN, the FT, Axios and Forbes

Google's News AI pilot, which began in December, licenses content from select publishers to test experimental AI features in Google News. It now includes more than 400 publications worldwide. New partners include CNN, The Financial Times, Axios and Forbes, plus U.S. local outlets such as The Post and Courier, Oil City News, Shaw Media and The Texas Tribune. Google declined to disclose terms and said the pilot is not pay-per-value, so publishers are paid for providing content that appears in experimental features, not by how much it contributes to an AI answer. Features tested include AI story previews and a customizable audio briefing.

The Operator’s Take: The pilot pays for access to content during testing, not for the value an AI answer draws from it, which is the gap a separate pay-per-value program is meant to cover. I learned this running hyperlocal news platforms inside a cable and broadcast company: local content keeps its value only when the operator keeps the audience data around it.

Commercial · Economic Model Ask Google in writing which program each deal sits in, and what happens to the payment basis if a publisher later moves into pay-per-value.

Read more →

Product & Platform Strategy · 15 October 2026

Google adds inline links, source sections and article carousels to AI Mode and AI Overviews

Google says it has substantially increased links in AI Overview and AI Mode answers, embedding them in words and phrases and showing them higher and more often. When only a few official sites exist for a task, a source section at the top explains why they are useful. Article carousels for timely queries now appear in more places, languages and query types and highlight preferred sources, whose usage Google says has grown 7x this year. Search Engine Land's Barry Schwartz doubts searchers will click the links, and notes Search Console still gives no click data for them in its AI performance reports.

The Operator’s Take: More links are a design change, not a payment, and until click data exists a publisher cannot say what they are worth.

Build · Detection & Telemetry Track referral traffic arriving from Google's AI surfaces in your own analytics, and compare it with Search Console's AI performance reports.

Read more →

Product & Platform Strategy · 15 October 2026

Wirecutter builds sponsorable franchises as AI Overviews cut clicks on commercial searches

Wirecutter, owned by The New York Times, is adding verticals, story formats and franchises that advertisers can sponsor as AI changes search. Its Hall of Fame Awards launched October 1 with multiple sponsors, and that coverage drew over 3 million page views; Best New Picks, launched in March, has drawn over 2 million. "Hybrid deals" bundling display, licensing, brand partnerships and affiliate have grown over 300% year over year, per the Times, from a small base. An Advanced Web Ranking study from September found an 11.91% click-through rate on commercial-intent Google searches overall, against 3.51% when AI Overviews appear.

The Operator’s Take: The response to lost search clicks is owning franchises a brand can buy, rather than waiting for the click to return; the 300% growth starts from a small base.

Commercial · Economic Model Model what share of affiliate revenue depends on commercial-intent search, then price one sponsorable franchise as a hedge.

Read more →

AI Content Licensing, Provenance & Deal Intelligence · 15 October 2026

Publishers at Digiday's London town hall weigh blocking AI crawlers as licensing talks slow

At Digiday's London town hall on October 13, publishers weighed control against visibility when blocking AI crawlers. One worried that restricting LLM access could undermine topical authority; another said scarcity could help establish a market for the content. The Financial Times' Fiona Spooner said a trial account tried to access "about a million different articles" in roughly eight hours. Attendees said LLMs send little traffic and many AI licensing talks have slowed, leaving a licensing deal as potentially the main extra revenue. SPUR co-founder David Buttle said the fair market for content to fuel AI systems "doesn't exist. But it urgently needs to exist."

The Operator’s Take: Blocking is only a negotiating tactic if a publisher can show what it stopped and what got through; a trial account pulling a million articles in eight hours is the kind of evidence that backs the position.

Build · Detection & Telemetry Run a trial-account test on your own paywall to measure how fast automated access can pull articles, and keep the logs.

Read more →