Copy article

CMA’s Google search crackdown is really about AI attribution: prove who gets paid when summaries replace clicks

ended 05. March 2026

The CMA’s proposed rules for Google search are being framed as a competition story. It is also an AI accountability story for the media economy.

When AI Overviews answer the question directly, the reader never reaches the publisher. That shifts value from reporting to summarising. It also creates a murky chain of responsibility: whose words are being used, who can opt out, and what counts as proper attribution when the output is a stitched-together paraphrase.

This is the next version of platform power. It is not just about ranking links. It is about whether AI features can quietly turn journalism into training data and raw material, while keeping the revenue on the platform.

The real test is enforceability. A publisher needs a clean way to say yes to search indexing but no to AI feature use and model training. Users need transparency on what is an answer versus a summary of someone else’s work. And regulators need evidence that attribution and opt-outs actually work at scale, including when content is re-hosted, syndicated, or scraped.

If this stays vague, the result will be predictable: the biggest publishers negotiate, the smallest get ignored, and the public loses local reporting without noticing until it is gone.

We'd like your views:

  • Should AI summaries require a stronger form of attribution than a small link?
  • Can an opt-out ever be meaningful if the platform controls discovery?
  • What would a fair deal look like: licensing, revenue share, or traffic guarantees?
  • Should regulators treat AI Overviews as a new product with separate conduct rules?
  • How should smaller publishers avoid being squeezed out of the bargaining table?

2 responses from the Newspage community

Copy all

Star Quote
Copy

Rohit Parmar-Mistry
Founder at Pattrn Data
The CMA is not just picking a fight with Google. It is forcing a hard question: when AI summaries replace clicks, who gets paid and who gets hollowed out.

Attribution is not a tiny link. It is provenance you can prove. Publishers need a clean control panel: yes to indexing, no to AI Overviews, no to model training, and they need to know it actually works at scale, including rehosted and syndicated content. Regulators should treat AI summaries as a separate product with separate conduct rules, plus auditing rights. Otherwise the biggest publishers cut bespoke deals and everyone else subsidises the platform.

In our AI audits, the killer gap is evidence. If Google cannot show which sources shaped a summary, and what share of value flowed back, it is not innovation. It is extraction with better UX.
Copy

Editor
Kate Underwood
Founder & Chief People Strategist at Kate Underwood HR and Training
AI Overviews aren’t ‘helpful’. They’re Google, nicking the click and calling it progress.
Attribution: a tiny link is not credit, it’s camouflage. If the answer is built from journalism, name the publisher clearly, up front, and make the source unavoidable.
Opt-out: if Google controls discovery, opt-out only works if there’s no punishment. Publishers must be able to say “index yes, AI summaries no, training no” and keep fair visibility. If traffic tanks, that’s not choice, it’s coercion.
Fair deal: licensing for AI use plus revenue share when AI answers replace visits. Traffic “guarantees” are fairy dust. You can’t pay reporters in vibes.
New product, new rules: AI answers are publishing, not search. Treat them separately with audits, evidence, and real penalties when attribution and opt-outs fail.
Small publishers: set a baseline deal everyone gets, and allow collective bargaining. Otherwise, the big brands negotiate, and local news gets quietly mugged.