The UK Competition and Markets Authority has published the responses to its consultation on a proposed user choice conduct requirement for Google’s general search services. That is the rulebook for the search engine choice screens Google will have to show on Android and Chrome in the UK. The alternative search engines have used their responses to go after one detail in particular. It is the metric that decides who gets the guaranteed slots.
Under the CMA’s draft, the five most “popular” eligible providers appear on every screen. Seven more are drawn at random. Popularity would be measured by Google Play app downloads. Almost everyone outside Mountain View seems to think that is the wrong number.
DuckDuckGo: the maths of the random slots
DuckDuckGo’s response runs the numbers on the “5+7” design. With 15 eligible providers, a provider outside the top five appears on 70% of screens. With 20 providers that falls to 47%, and with 25 it falls to 35%. DuckDuckGo says that in France its selection rate tripled once it moved into the top five from the random pool. It wants larger fixed blocks, such as 8+4 or 6+6, and a popularity metric based on actual search usage, for example a referral share service like Cloudflare Radar.
It also points out that, on the CMA’s own figures, the roughly 30 million Android devices already in use in the UK get no choice moment at all. That is because choice screens, redisplay and test-drive only apply to new devices. DuckDuckGo wants the non-skippable annual screen back, or at least a persistent notification of the kind Google used on existing Android devices in the EU. One useful data point: when Chrome’s switchback prompt stopped applying to its one-click extension, switching success was 30.9% higher than with manual set-as-default instructions, with no increase in churn.
Ecosia: “we lose 80”
Ecosia wants one app per corporate group. Its reasoning is that otherwise the biggest companies can take extra slots by running a separate search app and AI assistant app. It also does not want AI assistants mixed in with search engines. It welcomes the proposal to let providers deep-link users straight to the default search setting, and the removal of Google’s power to cap in-app prompts. Its reason is blunt: “Of every 100 people who set out to make Ecosia their default search engine, we lose 80 because the process is so cumbersome.”
Qwant and Yahoo: test the testers
Qwant combines a syndicated web index with its own European index. It wants the CMA to confirm explicitly that engines using syndicated results are eligible. It also wants the CMA, not Google, to approve the testing methodology used to decide whether a provider qualifies. On scope, it says new Google surfaces such as Circle to Search, Gemini and AI Mode should come under the rules once they reach a set share of UK queries, suggesting 5%. And it wants retention measured at 30 and 90 days, because a switch that reverts after a week is not competition. It notes that Google’s own cost estimates are around £5 million for extending the screen to Chrome on desktop and iOS, and up to £5 million for test-drive over five years.
Yahoo calls Play installs “irrational” as a popularity measure. It says Google’s global use of an installs-based metric has already removed Yahoo from the list of top alternatives in some markets where it is widely used. It suggests StatCounter or Cloudflare data instead. Karma Search wants AI assistants shown in a separate section from search engines.
Our take
Every respondent above makes the same argument: an app-install count measures marketing spend and Android distribution, not where people actually search. On the evidence submitted, that argument is hard to dismiss. DuckDuckGo’s France figure and Yahoo’s claim of being dropped elsewhere are the kind of real-world outcomes the CMA should weigh more heavily than modelled ones. They are still the companies’ own numbers, mostly unaudited, and some of Yahoo’s evidence is redacted. So I’d treat them as strong signals rather than settled fact.
The more interesting fight is the one about AI. The CMA’s draft lets AI assistants compete on the same screen, which is where the market is heading. But Ecosia and Karma are right that a “search engine” slot and an “AI assistant” slot are not the same product decision for a user. Qwant’s point about Circle to Search and AI Mode matters too. A choice screen that only governs the address bar risks regulating the 2023 version of Google search while query volume quietly moves elsewhere.
We have seen this before. The EU’s DMA choice screens produced short-term share gains for some providers but did not reshape the market, a pattern we track across the non-Google share data. Engines like Ecosia and Startpage are fighting over fractions of a percent. For brands in the UK, the practical point is unchanged. Bing-, Brave- and independent-index visibility is cheap insurance, and it will matter a bit more if this screen lands well.
The CMA expects to make its final decision by the end of the year. Responses from Google, Microsoft, OpenAI, Mozilla, Samsung and publishers’ groups are also on the consultation page, and we will come back to those.