What the CMA changed on September 23
The UK's Competition and Markets Authority published a strengthened search choice-screen proposal on September 23, requiring Google to let Android and Chrome users pick ChatGPT, Perplexity or another AI assistant as their default search service, not only a rival search engine.
Google must show the screen the first time a device or browser is set up and prompt users again once a year, and an AI assistant only qualifies if it meets technical and security criteria the CMA has not yet published in detail. CMA chief executive Sarah Cardell said "the way people search for information online is changing, with new services creating fresh opportunities for innovation and competition."
The publisher ask this proposal says no to
In August, three UK publisher trade bodies, the Professional Publishers Association, the Independent Media Association and the Movement for an Open Web, filed submissions asking the CMA to keep AI chatbots off that same choice screen, arguing that an AI answer which skips a publisher's page entirely is not the same kind of competitor as a rival search engine.
The CMA's September proposal answers that request by rejecting the exclusion, but it does not ignore it either: every AI assistant on the screen now carries a mandatory attribution duty that a plain search engine listing never needed before.
An antitrust case is now writing AI's citation rules
That attribution duty is not a copyright rule and not part of any AI-specific law: it is a condition attached to a competition remedy under the Digital Markets, Competition and Consumers Act, so a chatbot that will not credit a publisher's page risks losing its place on the one screen that sets Google's default in the UK, not a court judgment over AI training data.
No equivalent lever exists yet in the EU's own AI Act or its copyright rules for generative systems, which means a UK competition case, built to settle a search engine argument, may end up doing more to define how an AI assistant must cite its sources than either.
What publishers asked for versus what they got
Set side by side, the August request and the September outcome do not match on a single point.
| Question | Publishers asked for | The CMA proposed |
|---|---|---|
| AI assistants on the choice screen | Excluded entirely | Included, if they meet technical and security criteria |
| Publisher attribution | Not addressed by the screen's design | A mandatory attribution duty attached to every listed assistant |
| Enforcement route | CMA discretion on eligibility | Tied to choice-screen eligibility under a DMCCA conduct requirement |
| Decision timing | Filed August 17 | Consultation closes October 9, final decision by end of 2026 |
Publishers lost the exclusion they asked for and gained an enforcement lever they did not name in their own filing.
What happens between now and the CMA's final decision
Anyone running a UK publisher site, a search or AI product or a regulator's office elsewhere in Europe watching this as a precedent has one date to put in the calendar: the consultation closes October 9, and the CMA has said it expects a binding decision by the end of 2026.
For UK publishers specifically, the concrete question is no longer whether the exclusion request succeeded, since it did not, but whether "fairly attribute" turns into a standard the CMA can actually measure, or stays a phrase it never defines, because that choice decides whether the remedy protects referral traffic or only appears to.
Read next: Your AI Vendor Just Got a Letter From Brussels | A Safety Essay Became Evidence in an Antitrust Suit
