Not Just Search Anymore: UK CMA Wants ChatGPT and Perplexity on Android's Choice Screens

For a decade, “choice screens” on smartphones were about picking between Google and a handful of lesser-known search engines. On Wednesday, September 23, Britain’s Competition and Markets Authority (CMA) proposed rules that would turn them into something far more consequential: a front door for AI assistants.

Under the revised proposals, Google would be required to offer Android and Chrome users in the UK a choice of search services when they first set up a device — and that choice must now include AI assistants such as OpenAI’s ChatGPT and Perplexity. A second requirement would prompt users annually to re-select their default, a cadence Google has already objected to. A third rule would force search providers, including AI-native ones, to clearly credit and link to the publisher content their answers rely on.

“The way people search for information online is changing,” CMA Chief Executive Sarah Cardell said as the regulator unveiled the package, acknowledging that the rise of AI assistants has overtaken the assumptions baked into the original case.

The road to this moment

The proposals are the final stage of a remedy package first initiated in January 2026, and they rest on a legal foundation laid over a year earlier. In October 2025, the CMA designated Google with Strategic Market Status (SMS) in general search and search advertising under the Digital Markets, Competition and Consumers Act 2024 — its first such designation under the new regime.

That designation was notable for how it drew the map. The CMA explicitly brought Google’s AI-based search features — AI Overviews, AI Mode, and Discover — within the scope of the regulated market, treating them as extensions of Google’s search dominance rather than separate products. But it excluded Google’s Gemini assistant, at least for the time being, while signalling that a future review could pull assistants into scope as the technology and usage matured.

Wednesday’s proposals show that future arriving faster than expected. The CMA’s own investigation files had found that AI assistants such as Gemini Apps and ChatGPT already exert competitive constraint on general search — but also that Google’s deep integration of Gemini into the Android operating system gives it a distribution moat that standalone assistants struggle to cross. The choice-screen remedy is a direct answer to that asymmetry.

What the rules would actually require

Three elements form the core of the package:

Choice screens at setup. Google and Android OEMs would present UK users with a choice of search services on first use of an Android phone or the Chrome browser, rather than defaulting silently to Google Search. Critically, the definition of eligible services now extends to qualifying AI assistants that meet technical and security bars — precisely the integration point where Gemini’s OS-level placement had insulated it from rivals.

Annual re-selection. Users would get a recurring prompt, at least once a year, to reconsider their default. Google has pushed back hardest on this element, arguing that repeated interruptions degrade the user experience. The CMA’s counter-logic is that defaults set once are defaults forever, and that a one-time choice cannot keep pace with a market where the competitive landscape shifts every few months.

Fair attribution. Search and AI providers would be required to clearly credit and link to publisher content used in their answers. This element responds to sustained lobbying from the Professional Publishers Association and News Media Association, which have warned that AI-generated answers divert traffic away from the news organizations whose material underwrites them.

Consultation on the revised plan runs until October 9, 2026, after which the CMA can move to impose binding conduct requirements.

Why AI assistants became the battleground

The CMA’s escalation reflects a structural shift in how people find information. Assistants like ChatGPT, Perplexity, and Gemini now answer many queries that would previously have flowed through a search results page — with links, ads, and publisher clicks attached. Every such shift erodes the search-advertising duopoly economics that made Google one of the world’s most profitable companies, and it does so faster than traditional antitrust timelines can accommodate.

For OpenAI and Perplexity, appearing on an Android setup screen in the UK would be distribution they could never buy. Default placement on the world’s dominant mobile OS is the scarcest real estate in consumer software; Android ships on the overwhelming majority of smartphones globally. A regulator-mandated slot turns the assistant market from a land-grab decided by marketing budgets into one decided by whether users, given a genuine choice, actually prefer the rival.

For Google, the stakes run deeper than one market. The company has faced a decade of choice-screen mandates in Europe — under DMA rules it already offers search choice screens in the EU, and in July 2026 the European Commission issued specification measures requiring Google to let competitors’ AI services compete with its own on Android. The UK rules would extend the same logic with a sharper AI focus and the added friction of annual prompts. Google’s objection to those yearly interruptions is not merely UX grumbling: every re-prompt is a moment a user could defect.

The parallel with the EU is instructive. Brussels’ July 2026 measures oblige Google to share anonymised search data — queries, rankings, clicks, and views — with eligible third-party search engines and AI chatbots from January 2027, and to open key Android handset features to rival AI voice assistants. If the CMA finalizes its proposals, Google will operate under two overlapping AI-era competition regimes in two of its most important markets, setting a template other regulators are likely to copy.

The publishers’ ambivalence

Not everyone cheering the AI-assistant competition welcomes every element. UK trade associations warned in August that conduct requirements pushing Google toward more AI in search could end up harming publishers — injecting even more AI-generated content between readers and the original sources, while attribution rules alone may not restore the traffic that AI answers displace.

That tension — between breaking Google’s dominance and handing AI companies an even bigger gateway to users — defines the policy problem. The CMA is effectively betting that competition among assistants, with mandated attribution, will discipline the market better than Google’s incumbency plus regulation. Publishers fear the cure could resemble the disease.

What happens next

The October 9 consultation close is the next milestone. The CMA will then weigh responses — expect Google to fight the annual prompt requirement and the assistant-eligibility criteria, and publisher groups to press for stronger attribution teeth. Final conduct requirements could follow within months, with compliance obligations landing on Google’s UK product calendar for 2027.

The bigger significance is precedential. If AI assistants become a mandatory choice-screen category in the UK, the definition of what regulators must treat as “search” changes everywhere. The CMA was the first regulator to designate Google under a DMCCA-style SMS regime; its choice to fold assistants into remedy design will be studied in Brussels, Washington, and beyond. The era of regulating search as ten blue links is over — the question now is whether regulators can keep pace with what replaces it.