Britain's competition regulator proposes that Android and Chrome ask people to pick a search service, and counts AI assistants among the options

On 23 September 2026 the Competition and Markets Authority reopened its proposal to make Google ask people which search service they want. The further consultation on a user choice conduct requirement, published alongside a report of user research on choice screens, would require Google to present Android and Chrome users with a choice of search service at first use and to prompt them again once a year. It closes on 9 October.

The change from the earlier draft is that AI assistants are now to be eligible for the ballot, provided they meet technical and security requirements. The CMA frames this as keeping pace with how people in the United Kingdom search. A third element requires Google to ensure search providers can fairly attribute publisher content, which is the publisher side of the same argument reappearing inside a consumer choice instrument.

The legal machinery matters here, because it is new. The CMA is acting under the strategic market status regime created by the Digital Markets, Competition and Consumers Act 2024, which lets it designate a firm in respect of specific digital activities and then impose tailored, binding conduct requirements. Designation is not a finding of wrongdoing. On Google's general search and search advertising activities the CMA has already imposed a publisher conduct requirement on 3 June 2026, and fair ranking and data portability conduct requirements on 17 June 2026. This would be the fourth.

Why a decentralisation publication should care about a search choice screen: a default is a decision that somebody else has already taken on your behalf, and the search default is the most valuable such decision in consumer technology. Everything about the regime under which it is being challenged is an attempt to move a decision from the centre of a platform to its edge, one requirement at a time. The honest question is whether the instrument works. A choice screen relocates the decision formally while leaving placement, ordering and habit in the hands of the party that designs the screen, and the history of remedies of this shape is not encouraging enough to assume the answer.

The AI assistant clause is where this stops being a rerun. Admitting assistants to the choice screen is the regulator conceding that the activity it designated is already turning into a different activity. It also raises a problem the consultation will have to answer: the set of firms capable of fielding a general purpose assistant is smaller and better capitalised than the set of firms that can run a search index, so a remedy aimed at concentration could end up seeding the next one. Whether the technical and security requirements are drawn to admit a small or self hosted provider, or only the three or four obvious names, is the part worth reading when the document is opened.

What is not established: the user research document was published but has not been read, so nothing here rests on what the CMA found about how people behave in front of a choice screen. The closing date is given as 9 October without a year on the page that was read. And the October 2025 designation date, while widely reported, was taken from a secondary summary rather than from the CMA's own case page.