DuckDuckGo told the UK Competition and Markets Authority that its revised plan for a search engine choice screen on Google's Android devices and Chrome browser would favour Google and the largest AI app makers, leave roughly 30 million Android phones already in use without any choice moment, and turn an annual choice screen into a notification users can dismiss. Aurélien Mähl, who works in public policy at DuckDuckGo, set out the objections today in a LinkedIn post that links the company's non-confidential response, a six-page document dated October 14, 2026.

In Short

The UK competition regulator wants Google to show people a list of search engines and AI assistants so they can pick which one they use by default on Android phones and in Chrome. DuckDuckGo, a smaller search company, says the way the regulator plans to pick who gets the best places on that list will mostly help Google and big AI apps, and that millions of phones people already own will never see the list. The regulator closed its consultation on October 9 and expects to decide by the end of 2026, so the rules that decide where people search in the UK, and where search advertising happens, are still being written.

What DuckDuckGo published

The response answers the CMA's second consultation on the user choice conduct requirement for Google's general search services, which the regulator opened on September 23, 2026, and closed at 5pm on October 9. The document carries the heading "DuckDuckGo response to the CMA's second consultation on the User Choice conduct requirement for Google's general search services" and the line "14 October 2026 - Non-confidential version". It is hosted on assets.publishing.service.gov.uk, the UK government's publishing domain, under the file name DuckDuckGO.pdf.

Its opening paragraph frames the whole submission. "The CMA presents its September 2026 proposals as a strengthening of the User Choice Conduct Requirement (CR). In our view, they are the opposite: the CMA's choice screen proposals favour the incumbent and will lock the existing ecosystem in place rather than stimulating growth from a wide range of providers," according to DuckDuckGo. The company adds that "the result is a choice architecture weaker than the January 2026 proposals and significantly weaker than what Google already operates in the EU."

Mähl's LinkedIn post condenses the argument into one line: "New choice screens in the UK could end up being worse than what they currently are in the UK and the EU." It then lists four design problems, three steps back and one area of progress. The post and the document agree on every figure they share.

The document is organised into six sections: the selection rules, eligibility of AI services, cadence and coverage, easy switching, reporting, and points from its response to the January 2026 consultation that the CMA did not take up. DuckDuckGo is a direct competitor to Google in search, and every figure about its own performance comes from its own data.

What the CMA proposed in September

The regulator's revised draft, a 71-page document, proposed a choice screen of up to 12 search providers on Android devices and Chrome, opened to AI assistants that meet three rewritten eligibility criteria. If 12 or fewer providers qualify, all of them appear in fully random order. If more qualify, Google selects 12: the five most popular plus seven drawn at random. For the first six months, popularity would be measured by a reasonable metric such as Google Play Store downloads; after that, by each provider's selection rate on the screen itself.

The draft also replaced a non-skippable annual screen from the January version with a skippable, visually limited notification tied to major Android or Chrome updates. It added a Device-Level Default Setting on Android, an API that tells eligible providers whether they are a user's default, and deep links into default settings. Google would report to the CMA every six months in the first year, then annually.

That package followed the CMA's decision to designate Google with strategic market status in general search and search advertising in October 2025, the first designation under the Digital Markets, Competition and Consumers Act 2024. The designation covers Google Search, including AI Overviews and AI Mode, as well as Google Ads and Search Ads 360 when used for search advertising. The CMA said at the time that Google accounts for more than 90 percent of general search queries in the UK.

The first round of proposed conduct requirements, published on January 28, 2026, put choice screens on Android and Chrome alongside measures for publishers. The publisher measure moved first: on June 3, 2026, the CMA imposed the Publisher Conduct Requirement, its first binding conduct requirement under the 2024 Act. The choice screen requirement is still a draft. According to the CMA's September 23 announcement, the regulator expects to make a final decision on it by the end of 2026.

How the selection rules would work

DuckDuckGo's central complaint is that three separate design choices compound. "Using Play Store downloads as popularity metric, allowing more than 1 app per provider and the 5+7 selection logic entrenches Google," according to the document, which cites paragraph 4.82 of the CMA's consultation and paragraph 6.a of the draft requirement.

Downloads as a popularity measure

The first problem, in DuckDuckGo's account, is the metric. "Play Store downloads, proposed for the first six months, do not measure search usage and favour apps with multiple use-cases," the document says, adding that this "is especially true for AI apps, whose recent massive download figures are driven by non-search uses." Mähl's post puts it more bluntly: "That rewards big AI apps with many uses beyond search."

The consequence DuckDuckGo spells out is direct. Well-known search brands, it says, "will not appear in the guaranteed five spots if the CMA uses Playstore downloads to determine popularity."

To show what a fixed slot is worth, the company draws on its own history in France. "In France, DuckDuckGo's selection rate increased x3 when it was included in the top 5 after previously being in the random 7," according to the document. It then reports "an additional +30-40% boost following our later inclusion in the better designed, fully randomized DMA choice screen," the version Google runs in the European Union under the Digital Markets Act. A footnote qualifies the French figure: "While some of the lower observed selection rate was due to the lower ranking (which won't be reproduced under the CMA proposal), appearing a fraction of the time had a logical impact on selections." The document does not give the underlying selection rates or the period they cover.

DuckDuckGo's alternative is to use the selection-rate data from the existing UK Android search choice screen for the first six months, place every other eligible app in the randomised section, and switch to the new screen's own selection rates once they settle. The company notes that the CMA rejected its earlier proposal to rank popularity by search market share. It supports the regulator's plan to use effective selection rates after six months, with each app's total impressions on the choice screen as the denominator.

More than one app per company

The second problem is the number of slots a single corporate group could hold. The CMA, according to DuckDuckGo, rejected a cap of one app per group and requires only that each service from a group be "sufficiently distinct". The response argues that "Google is one of the only two firms operating separate search and AI assistant apps, and it may claim they are "technically distinct"." It does not name the other firm. "Allowing both risks handing the largest groups extra slots, including several of the five guaranteed spots," the document continues.

Mähl's post names the beneficiary. "This mostly helps Google given Gemini's and its search app's clout. Most providers don't maintain two separate apps for search and AI," he wrote. His opening paragraph goes further: the recommendations "combine to hand the largest providers the guaranteed five slots - Google could even get two of those."

The 5+7 arithmetic

The third problem is the split itself. Five apps would appear on every screen. Every other eligible app would compete for seven random slots, so its exposure depends on how many providers qualify. DuckDuckGo gives the formula in a footnote: the probability of inclusion equals seven divided by the number of eligible providers minus five.

The results, according to the company:

  • With 15 eligible providers, an app outside the top five appears on 70% of screens.
  • With 20, on 47%.
  • With 25, on 35%.

Neither the response nor the post estimates how many providers would qualify, so the figures describe a range of outcomes rather than a forecast. The response also dismisses the overflow option that would hold the remaining providers one step away from the main list. "The "view all" link is no substitute for a slot," it says, and DuckDuckGo expects selections from that reserve list "to be negligible."

The fix it proposes is to raise the number of fixed places. The five-slot figure, the document says, "was originally used because only five apps appeared "above the fold" on a mobile device screen but is unnecessary with the fixed scroll obligation." It suggests 8+4 or 6+6 instead. Mähl's post makes the same point: "The cutoff should be increased to ensure fixed slots don't just go to big firms."

The CMA's September draft rewrote the eligibility test so that AI assistants could qualify. A service must address four categories of query, navigational, transactional, simple informational and complex informational; it must offer a functioning general search experience; and a significant proportion of its UK users must consider it a general search service. The earlier requirement that search be a core part of the product was dropped. Google had argued for at least nine categories, while the CMA settled on four.

DuckDuckGo welcomes the added clarity but says the criteria "lack anchors in actual search usage, especially when it comes to navigational queries." Mähl's post opens this section with a warning: "Eligibility for AI is easy to game without hard data to prove search usage."

The navigational test

The document treats navigational searches as the defining test. "Navigational queries are the clearest test of a general search service and must be strictly interpreted," it says. A search for "the guardian" in the UK should return the Guardian's website among the top results, and the document claims that "many of the popular AI chatbots with a web search function currently do not support this in a straightforward manner." Mähl's post proposes a simple check, asking a chatbot for "guardian" or "sky" to see whether it returns the links, and adds: "Users can be confused to get those as their browser search default."

As evidence of real search use, DuckDuckGo proposes checking whether a service appears in the UK top 12 of a referral-based search market share service such as Cloudflare Radar, a dashboard Cloudflare launched in 2020 that draws on traffic across Cloudflare's network. Referral data, the argument goes, measures whether users click the links a service shows them, which filters out services that are not used for search. The company anticipates the objection that this would shut out AI products: "It does not exclude AI search per se - OpenAI and Perplexity do appear in the UK top 10." As a secondary signal, it points to uptake of a service's search extension on Chrome and Firefox.

What the CMA's own surveys show

DuckDuckGo also turns to the regulator's own research. According to the response, the Accent consumer survey prepared for the CMA in June 2025 found that only 5% of users rely on AI most often to find a specific website and 3% to find a product to buy. Among AI users, the shares were 13% and 7%. "Mixing categories on one screen will confuse users," the document concludes.

It criticises the second study, by Thinks, on method. According to DuckDuckGo, the Thinks research stops at the moment of choice and does not examine how people use the service they picked or whether the results meet their expectations. The only exception is "a seven-day trial done by 20 participants," and even there, the company says, the report records no findings on result satisfaction for these services.

A separate screen for assistants

The response does not argue that AI assistants should be left without remedies. "Competition among AI assistants needs its own tools," it says, citing a finding in the CMA's SMS Final Decision that defaults are also a barrier to entry and expansion for AI assistants. DuckDuckGo proposes interoperability measures and a separate assistant choice screen, which it says is why it supported including Gemini within Google's SMS designation. When the CMA designated Google in October 2025, the Gemini assistant was outside the scope, and the regulator said it would consult before deciding whether to add it.

Cadence and coverage

The third section of the response, titled "Reverse the regressions on cadence and coverage", deals with how often users see the screen and which devices get it.

A notification instead of a screen

In January 2026, the CMA proposed a non-skippable annual redisplay. The September version replaced it with a "skippable and visually limited" notification, based on some users in the CMA's qualitative research preferring to delay or skip a screen that interrupted a time-sensitive task. "A skippable notification is not an annual choice screen," according to DuckDuckGo. Mähl's post calls the new version a "skippable (and therefore meaningless) notification."

The company offers three pieces of evidence for keeping a non-skippable screen:

  • When Apple showed its iOS browser choice screen in the EU a second time, the second showing generated an immediate install uplift comparable to the first, according to an annex to DuckDuckGo's response to the January consultation. PPC Land reported in August 2024 that Apple's updated browser choice screen would be shown to all EU users with Safari as their default, including those who had seen the earlier version.
  • In a survey cited by the company, 69% of users wanted to see the screen again after one exposure and 71% after two. "This is choice appetite, not fatigue," the document says.
  • The CMA itself estimates that a non-skippable screen costs users "a little over 5p of time per showing," according to DuckDuckGo, which cites paragraph 5.21 of the consultation. Showing the screen at a major software update, the company argues, means it does not interrupt a task.

As a minimum, DuckDuckGo asks for a persistent notification that cannot be dismissed without making a choice, which it says Google already implemented on existing Android devices in the EU.

30 million phones without a choice

The coverage point gives this story its headline. Under the September draft, choice screens, redisplay and the test-drive feature on Android would apply to new devices only. "Existing Android devices are dropped entirely," according to the response.

The company's calculation, based on footnotes 162 and 164 of the CMA's consultation, is that around 30 million Android devices already in use would get no choice moment, while redisplay on Android would reach only 2.5 million to 5 million devices in year three. The document gives that second figure as a range in square brackets rather than an exact number. "Only showing them on new devices regulates the new phone market, not the search market and misses roughly 30 million existing Android devices," DuckDuckGo writes in its summary.

The response also notes an inconsistency with the CMA's own earlier findings. The regulator had criticised Google's "UK Choice Package" because users were prompted at most once; under the new draft, according to DuckDuckGo, existing users get even less. On cost, it cites manufacturers who told the CMA that Google could implement redisplay through Google-managed components at limited cost to them, and argues that much of the cost is sunk because Google already showed choice screens to existing Android devices in the EU.

The September draft records the other side. The test-drive was limited to new Android devices and Chrome installations, with Google saying it has limited control over existing non-Pixel devices.

Users who already left Google

The September draft defines Relevant Browsers and Devices by whether Google was the out-of-the-box default, as Google requested, according to DuckDuckGo. Users who actively picked an alternative would therefore see the same screens as Google users. "A default won from Google is precisely the outcome sought," the document says, citing the SMS Final Decision's finding that consumers are likely to use the default they have chosen.

A six-month exemption in the draft, which lets Google skip the notification for users who changed their default in the previous six months, draws a separate objection. Because the exemption is provider-neutral, a switch back to Google could also suppress the next notification. DuckDuckGo wants it to apply only to deliberate switches away from Google.

Switching: the part DuckDuckGo welcomes

Not everything in the response is critical. The company lists four elements it "strongly" welcomes:

  • Deep linking from providers' websites in Chrome and from their Android apps straight to the relevant default setting.
  • The Device-Level Default Setting propagating to Chrome and the search widget on all relevant devices.
  • Removal of Google's ability to cap in-app prompting on Android.
  • Default-status detection through the Search Default Setting API on Chrome.

"Removing friction works," the document says, and supports the claim with the company's own figures. When Chrome's switchback prompt stopped applying to DuckDuckGo's one-click simple search extension, the extension outperformed manual set-as-default instructions by 30.9% in switching success rate, with no increase in churn, according to DuckDuckGo. The document does not describe the test design or the period. On Android, the company expects opt-in rates several times higher than its current widget prompt, "closer to the c. 40% seen for the browser default prompt."

Mähl's post calls this "real progress", with a condition: deep links "will make switching much easier, as long as Google does not add new friction." The document is more specific. It asks the CMA to rule out switchback prompts after a deep-linked change and notes that Google says it is introducing an updated, "fully neutral" confirmation dialogue on Chrome. "Whatever its design, no prompt is needed where an extension clearly states that it changes the default search engine: our data show no churn increase without one," according to the response.

There is a limit to the benefit, the company warns. Deep links lead only to settings that cover Chrome and the search widget, so access points locked to Google stay out of reach.

Reporting that providers never see

The fifth section concerns data. Under the September draft, Google's reports would go only to the CMA, every six months in the first year and annually after that, which DuckDuckGo says was what Google requested. With the first redisplay notifications not due before year three, the company argues that the regulator would see evidence on its most contested measures years after the fact. It wants reporting "preferably daily".

It also wants the data shared. "Eligible Providers cannot plan campaigns or improve their offer without their own data while Google has ongoing access to this data to support its own planning," the response says. At a minimum, according to DuckDuckGo, Google should share with each eligible provider its selection rates and display positions, daily.

Points the CMA did not take up

The final section revisits proposals from DuckDuckGo's response to the January 2026 consultation:

  • Never place Google in the top positions. According to DuckDuckGo, ranking Google lower is "the single most impactful design change, increasing selection of alternatives by up to +89%." The figure comes from annexes to that earlier response, which are not part of this document. Full randomisation, the company notes, still puts Google in the top five on around 42% of a 12-slot screen. PPC Land's calculation: five positions out of 12 equals 41.7%, which matches.
  • Replace Google's icon. On iOS, placing the chosen app on the dock or home screen in place of Google's "markedly improved retention and searches per app," according to the response.
  • Locked access points. DuckDuckGo welcomes the CMA's commitment to monitor other access points but wants an explicit review at the 12-month compliance stage on extending the requirement to Circle to Search and the web component of on-device search. The September draft excluded Circle to Search and other Android access points.
  • No silent fallback. Uninstalling the chosen app "should re-trigger the choice screen rather than silently revert to Google Search."

What the document does not establish

Several of the response's strongest numbers rest on DuckDuckGo's own data or on annexes that are not public in this version: the threefold French selection-rate increase, the 30-40% uplift on the DMA screen, the 30.9% switching improvement and the 89% effect of ranking Google lower. The document does not give sample sizes, dates or methods for these figures, and the French comparison mixes the effect of fixed placement with the effect of ranking, as the company's own footnote concedes.

The 70%, 47% and 35% exposure figures are arithmetic on a hypothetical number of eligible providers. The response also does not identify the second company that, in its account, runs separate search and AI assistant apps.

The response is also an advocacy document from an interested party. A company outside the top five by downloads stands to gain from every change it proposes: more fixed slots, popularity based on search selections, a ban on Google appearing near the top. The CMA's draft records Google's opposing positions, including its argument that the costs of the requirement outweigh its benefits. The regulator estimated five-year costs of about 5 million to 20 million pounds, against UK search advertising revenue of 10 billion to 20 billion pounds in 2024, and concluded that the benefits likely outweigh the costs.

Why this matters for marketers

Defaults decide where queries go, and queries decide where search advertising budgets are spent. With Google above 90% of UK general search, the choice screen is one of the few levers the CMA has to change how that traffic is distributed. Its design, down to how slots are allocated, will shape whether any shift goes to independent search engines, to large AI assistants, or nowhere at all.

European evidence suggests the effect of choice screens on market share is modest. PPC Land reported in July 2026 that DMA browser choice screens added roughly 6 million Firefox users in the EU, according to an NBER working paper, while Google kept about 90% of the search market. The same paper found DuckDuckGo outperforming Ecosia and Qwant in European usage. A month later, Brave Search gained a place on Chrome's global default search engine list, a listing PPC Land noted is not the same as becoming the default.

The UK proposal differs from the EU model in one significant respect: it lets AI assistants onto the screen. For advertisers, that raises a practical question about where commercial queries would go if a UK Android user picked an assistant as the default search service. DuckDuckGo's argument about navigational queries goes to that point. A brand search such as "the guardian" is also where branded search campaigns run, and an assistant that does not return the site's link changes what happens to that click.

The reporting dispute matters for a narrower group: the search providers themselves. DuckDuckGo's complaint that eligible providers "cannot plan campaigns" without selection-rate data shows how those companies treat the choice screen, as an acquisition channel with conversion rates that can be measured and marketed against. On the EU side, the European Commission in July required Google to share anonymised search data with eligible rivals and to open Android features to competing AI assistants. That is a different kind of data, but an obligation that runs to competitors rather than only to the regulator.

Google has argued since the designation that the UK regime could slow innovation. When the designation was announced, Oliver Bethell, Google's Senior Director of Competition, warned that many of the proposed interventions would inhibit UK innovation and growth.

What happens next

The consultation closed on October 9, 2026. The CMA has said it expects to reach a final decision on the user choice conduct requirement by the end of 2026. Under the September draft, if the requirement is imposed, Google's approval process for eligible providers would begin one month later, with the other obligations following within six months. Whether the regulator changes the slot split, the popularity metric, the treatment of existing Android devices or the redisplay cadence will be visible only in the final text.

Timeline

Summary

Who: DuckDuckGo, through its non-confidential consultation response and a LinkedIn post by Aurélien Mähl, who works in public policy at the company; the UK Competition and Markets Authority; and Google, whose Android devices and Chrome browser the requirement would cover.

What: DuckDuckGo argues that the CMA's revised choice screen design would favour Google and large AI apps. It objects to Play Store downloads as a popularity measure, more than one app per corporate group and a 5+7 slot split under which an app outside the top five would appear on 35% of screens if 25 providers qualify. It also objects to replacing the annual screen with a skippable notification and leaving around 30 million existing Android phones out. It welcomes deep links to default settings and asks for daily reporting shared with providers.

When: The response is dated October 14, 2026, after a consultation that ran from September 23 to October 9, 2026. Mähl published his post on October 26, 2026. The CMA expects to decide by the end of 2026.

Where: The United Kingdom, on Google Android devices and the Chrome browser, under the Digital Markets, Competition and Consumers Act 2024.

Why: The choice screen is the CMA's main tool for loosening Google's hold on more than 90% of UK general search queries. How slots are allocated, which AI services qualify and which devices are covered will determine whether the screen moves any search traffic, and with it search advertising, away from Google.