Steering Far and Beyond: The European Commission Fines Google for a Breach of Article 5(4) DMA

In July, the European Commission issued a press release communicating that it had fined Google for preventing app developers from freely communicating and promoting offers and concluding contracts with users in distribution channels of their choice, including third-party app stores. Two months later, in September, the DMA’s enforcer published the non-compliance decision relating to how Google had breached the steering mandate under Article 5(4) DMA.

The case jumped on the bandwagon of the European Commission’s previous non-compliance decision against Apple for similar reasons (see comment here). The present case relating to Google builds on some of the rationale brought by that first case surrounding Article 5(4) DMA and expands on several concepts that were left unclear by the regulator, notably how gatekeepers can introduce limitations to steering and how to benchmark the initial acquisition fee that the steering mandate does allow for.

The Non-Compliance Decision Does Not Work Wonders

Google’s approach to complying with Article 5(4) DMA was quite similar to that initially presented by Apple back in 2024. By default, app developers could not access the technical solutions that should have applied by virtue of the DMA. That is to say, the General Business Terms applied to all app developers. The business terms do not allow steering and charge 30 or 15% per transaction completed by an app user, depending on the app developer’s total yearly earnings (paras 41-47 of the non-compliance decision). Needless to say, the European Commission considered that this situation already infringed Article 5(4) DMA, since they “completely prevent app developers from communicating and promoting, within the app, offers that can be purchased outside the app, and to conclude contracts with acquired end users following such communication or promotion of offers, either within or outside the app” (para 56).

In 2024, Google introduced the option for Google Play developers to promote offers with their Google Play apps and show hyperlinks that take users out of the apps to their external sites to conclude contracts for those offers. This was articulated through the External Offers program, which app developers had to sign up for (page 151 of Alphabet’s 2024 compliance report).

The External Offers program applied new terms and conditions, including the imposition of a new fee model. Therefore, Google started to charge two separate fees: i) an initial acquisition fee, limited to a period of 2 years amounting to 5% for automatically renewing subscriptions and 10% for other offers that are consumable within the app; and ii) a fee for ongoing services amounting to 7% for automatically renewing subscriptions and 17% for other offers that are consumable within the app. In 2024, both fees would be charged upon the app developer’s agreeing to the conditions of the External Offers program for 2 years, corresponding to the time in which the initial acquisition fee would be due. Following that time, developers would have the possibility to opt out of receiving ongoing services from Google Play, subject to users agreeing and enabling that opt-out (page 152 of Alphabet’s 2024 compliance report).

According to the European Commission, the External Offers Program did not restrict the number of link-outs that developers can show in any app distributed and they have broad freedom to design the wording of their in-app promotional messages (para 51 of the non-compliance decision). Notwithstanding, Google did introduce restrictions on the app developer’s steering of users. First, app developers could only direct end users to its own digital content or items associated with the app, but not to destinations that are unrelated to the app or relate to third-party content. Second, every time an end user follows an external link, the app developer must inform the end user in a prominent manner that the app developer is the merchant and therefore bears sole responsibility for security and safety. Third, the app developer cannot pre-populate fields on the destination page with the end user’s personal data (such as their payment details once they have already completed a transaction) from the app (paras 51 and 53 of the non-compliance decision).

The European Commission triggered one of its first non-compliance procedures against this conduct in March 2024, only days after Google had released its compliance report. After the DMA’s enforcer sent its preliminary findings remarking on the fact that Google Play did not comply with Article 5(4) DMA, Google announced updates to its External Offers program in August 2025. The update included changes offering developers the capacity to link to owned app downloads, as well as to their own digital content and regarding the applicable fees. The gatekeeper maintained the same initial acquisition fee, whilst it introduced a tiered model for the ongoing services fee to increase developers’ optionality, consisting of two tiers (required tier 1 and optional tier 2). Links to purchases when end users clicked on in-app content amounted to the imposition of the same fixed 10% fee, whereas links to downloads triggered the imposition of higher fixed rates per transaction in Tier 2 as opposed to Tier 1. By default, developers received both Tier 1 and Tier 2 ongoing services, and once enrolled, they could choose to opt-out of Tier 2 (page 159 of Alphabet’s 2026 compliance report).

The non-compliance decision covers the period from March 2024 to December 2025 (para 531 of the non-compliance decision). That is to say, since Article 5(4) DMA was applicable to Google and until the new measures proposed in August 2025 became live. The European Commission takes a snapshot of conduct that certainly is in breach with the DMA and cuts out of the picture the progression of that behaviour into the future when it becomes more convoluted, making it more difficult to be captured through only one non-compliance procedure. This is precisely what the enforcer also did when it captured Apple’s steering practices back in May 2025 (para 46 of Apple’s non-compliance decision).

By doing so, the enforcer sidestep one of the main objectives when the DMA was adopted, which was to tackle Big Tech conduct in an effective and quick way. One cannot possibly challenge the fact that the non-compliance decision was taken quickly (although it was issued one year later than the previous non-compliance decision), but effectiveness cannot be assumed just as promptly. To the extent that the non-compliance decision deliberately avoids pronouncing itself on conduct that takes place in the moment that it is released, it is by itself backward-looking. This can be taken lightly, by arguing that the European Commission, when it triggers its punitive powers, surely must be allowed some leeway to impose sanctions on past conduct.

But then a related question arises, which is how the non-compliance procedure contributes to deliver fair outcomes to consumers in digital markets. This is precisely what the post will navigate, because there are some points in the non-compliance decision that do contribute to the protection of consumers and the promotion of fairer outcomes when it comes to steering in Google’s app store, Google Play. As a matter of fact, Google has altered its conduct following some of the principles outlined by the European Commission in the non-compliance decision. For instance, it has lowered the initial acquisition fee to 0%, whilst it has maintained the 10% ongoing services fee. As we will now see from the European Commission’s analysis (and as I have also argued in relation to other cases, see here), that does not necessarily mean that the infringement of Article 5(4) DMA has actually come to an end.

Stretching Out the Mandate on Steering under Article 5(4) DMA

Back in May 2025, when the European Commission fined Apple for breaching the steering mandate, the enforcer set out the first few principles surrounding its interpretation. In that decision, it clarified that steering in the sense of Article 5(4) DMA comprises not only ensuring that app developers can engage in transactions with their end users, but also in any form of communication and conclusion of contracts. Related to the scope, the EC also highlighted that the conditions surrounding steering should be developed and implemented in a neutral manner, based on tangible, verifiable data collected by Apple, and that potential limitations to their exercise could not go beyond what is objectively necessary and proportionate (para 312 of Apple’s non-compliance decision).

Additionally, the EC also fleshed out the modalities of payment that the gatekeeper can apply when app developers steer their users because of the efforts that they dedicated in their app store so that they could reach them. Following the letter of Article 5(4) DMA, steering must be performed free of charge, even though gatekeepers can be remunerated for facilitating the initial acquisition of end users by the app developers under three conditions. First, a fee can only be charged when it is related to the initial acquisition only, both in terms of time and scope, which means that it cannot be extended to already acquired end users. Second, it must be commensurate with the value of the initial acquisition and must take into account any other, direct or indirect, remuneration received from business users for facilitating the initial acquisition. Third, it must not remunerate the gatekeeper for gatekeeper value. The EC even went on as far as going that “Apple is prohibited from imposing any other type of fee (in addition to a potential fee for initial acquisition) that covers services linked to the acquisition of end users” (para 312(c) of Apple’s non-compliance decision).

Those principles have now been expanded with the EC’s interpretation of Google’s technical implementation of Article 5(4) DMA. As it did with the previous non-compliance decision, the EC found several instances where the gatekeeper infringed the provision in different ways, namely by: i) limiting third-party destinations in steering; ii) restricting link-outs and preventing app developers from pre-populating URLs; iii) introducing friction through the prompt that was shown to end users upon steering; and iv) excluding web view from the scope of application of the External Offers program. The EC also took issue with Google’s fee imposition and considered it to fall outside of the legal possibilities opened by Article 5(4) DMA. I will cover those in turn.

The Scope of Steering and Non-Proportionate Limitations to Communicating with End Users

Technically, Google did a better job in complying with the general terms of steering because it did allow all modalities of steering comprised under Article 5(4) DMA that Apple had, in turn, restricted. Substantially, the EC still took issue with the fact that app developers are restricted from showing, in their apps, link-outs that lead to destinations hosting something other than digital content or items associated with their respective apps, such as third-party digital content or items (para 85 of the non-compliance decision).

According to the EC, app developers should be able to freely steer end users to any type of offer, including to third-party offers, in apps or on app stores not operated by the app developer itself. For instance, that would include “situations where a developer may wish to steer an end user to an alternative app store provided by a third-party developer” (para 91). Restrictions operating against third-party content being available upon steering would, therefore, be contrary to the objective of Article 5(4) DMA because they would “de facto allow the gatekeeper to dictate the terms of how and where app developers interact with acquired end users” which would, in turn, “increase those app developers’ dependence on the gatekeeper CPS in question and further entrench its gatekeeper status” (para 93).

Furthermore, the EC also takes issue with Google’s repeated imposition of the prompt following every link-out and terms it “an unjustified restriction on developers’ ability to freely communicate, promote offers and conclude contracts” (para 114). As opposed to the limitation on third-party content, this aspect touches upon Google’s burden of proof in demonstrating that the prompt was neutral and objective (para 115). In other words, it is not that the prompt per se is contrary to Article 5(4) DMA, but rather that the gatekeeper was compelled to show how it abided by the principles of proportionality and objective necessity. For instance, the EC argues that the wording and the frequency of the prompt entailed that Google could hardly argue in favour of it being objectively necessary for informing end users of potential risks arising from steering (paras 117 and 118).

The enforcer also applies this same rationale to its consideration of Alphabet’s prevention of app developers from passing on additional parameters in the URL in the hyperlink they provide in their app for linking out. In practical terms, the restriction required end users to re-enter the required information (e.g., their payment information) each time that they were steered out of the developer’s app, unduly degrading their experience and the usefulness of steering altogether (para 120). The EC discredited Google’s arguments in those the restriction was justified because the users could already use the auto-fill payment details functionality available on their browser. Given that the auto-fill feature requires the end user to have entered and saved their information previously, the EC considers it constitutes an additional step in steering, which Article 5(4) DMA does not allow for (para 121).

The most interesting aspect within the analysis of these restrictions is that the EC seems to deviate from the standard of proof applicable to the gatekeeper, as opposed to its Apple non-compliance decision. The EC notes that Google is not prevented from taking measures that are objectively necessary and proportionate to protect security, but that it “did not provide elements showing that the restrictions are either objectively necessary or proportionate to protect the security of end users” (para 122, emphasis added). If one reads the paragraph in isolation, it hints at a fascinating nuance: the gatekeeper does not have to meet the very high threshold of demonstrating that its restrictions are both objectively necessary and proportionate, but rather can demonstrate one or the other, without one prevailing over the other. Objective necessity could apply to measure whether the goal (i.e., preserving security) could be attained through less restrictive means, in the absence of proportionality assessing whether the magnitude of the restriction outweighs the provision’s goal. In turn, proportionality could assess this aspect without accounting for less restrictive alternatives.

Engaging within such an interpretation would be tantamount to establishing that gatekeepers can simply overcome this high threshold without appropriate safeguards. As a matter of fact, the EC seems to correct its own standard later on in the decision by analysing those restrictions under the lens of both principles. For instance, that Google failed to provide “sufficiently specific evidence to demonstrate the existence and magnitude of the purported risks (of restricting third-party destinations in steering) nor why a complete restriction (…) is a necessary measure and the most proportionate one to mitigate these risks” (para 127). Or by asserting that “the frequency and messaging of the prompt (triggered by the link-out) must be necessary and proportionate” (para 144). Therefore, we can only assume that the EC’s initial assertion was just a slip of the pen with no intention to bring further repercussions to the burden of proof applied to the restrictions introduced by the gatekeepers.

Precisely on this point, the EC distances itself from the legal requirements that it did add to the provision in Apple’s non-compliance decision. On that decision, the EC highlighted that to the extent that the gatekeeper claimed that a disclosure sheet (prompt, in this case) is necessary and proportionate, “it is for (the gatekeeper) to put forward a thoroughly conducted analysis sustaining its argument. Apple has neither conducted any A/B testing, nor any other testing or consultation which could give support to its claim that the design and wording of the disclosure sheet is appropriate to inform users in a neutral and objective way (…)” (para 125 of Apple’s non-compliance decision).

As a response to this requirement, Google conducted its own internal survey on the reactions to alternative warning messages and the effectiveness of showing its prompt to alert on security and how it was necessary so that end users remained aware of potential risks. However, the European Commission criticised the gatekeeper for not engaging in live testing of end user in-app behaviour, which it had already communicated to Google prior to its conducting of the survey (paras 130 and 132 of the non-compliance decision). This indicates that the EC’s assertions in its previous non-compliance decision surrounding the need to conduct A/B testing is not a recommendation. Instead, it is a requirement that the gatekeeper must fulfil in case it wishes to introduce a prompt tailored to the Article 5(4) mandate.

The Free of Charge Clause and Permissible Fees

The EC’s non-compliance decision against Google does deviate from its previous non-compliance decision under Article 5(4) directed at Apple in the way in which it describes the aspects that must be taken into account for considering remuneration for facilitating an initial acquisition as abiding by Recital 40 and the provision (for further explanation on their interplay, see here). Building on the framework set out by the EC in May 2025, the enforcer goes a step further. Aside from asserting that the initial acquisition fee must be commensurate with the value of the initial acquisition, it then adds that the gatekeeper “should account for the fact that the value generated by an end user for the developer will always depend on a variety of factors, in particular on the marketing efforts of the developer and the attractiveness of the app” (para 290). This assertion comes as a reaction to Google’s own allegations to the preliminary findings, arguing that the factor that was to be taken into account was the user lifetime value derived from the end user’s initial acquisition. In other words, the EC discredits Google’s reliance on that metric and proposes two additional elements that it should have taken into account when devising its initial acquisition fee. To the extent that it did none of the above, then the initial acquisition fee cannot be held as fulfilling the legal requirements set out in Recital 40 and Article 5(4) DMA.

The EC reiterated some of the arguments that it had already presented to Apple when it assessed its compliance with Article 5(4) DMA, such as the lack of proximity of its imposition with the time of the download (para 333) or its duration (para 335). However, the reasoning of the EC, in some other instances, demonstrates to be quite novel because it frames new types of arguments that it had not exploited when it was scrutinising Apple. For instance, the EC also took issue with the fact that the initial acquisition fee’s application in two scenarios duplicates the remuneration that it had already obtained for the end user’s acquisition, namely; i) for those users who had downloaded an app prior to the compliance deadline, where Google already charged its standard 30/15% fee; and ii) in paid apps, where it had already charged a fee on the purchase price of the app (paras 297 and 298). Those two instances reinforced the message that Google’s initial acquisition fee did not accurately capture end user initial acquisition.

After that, the enforcer went on to analyse whether the fee could be considered as excessive by being placed at 5%/10% of all transactions performed by the end user having followed steering from the app developer’s app. To do that, it first clearly delineated the terms by which the initial acquisition fee should be judged. Contrary to Google’s allegations, it emphasised that only Article 5(4) regulates the fee that gatekeepers can charge following steering, and that the FRAND notion deriving from Article 6(12) was not devised by the EU legislator to govern the fees that a gatekeeper can impose in the context of steering (para 328).

Then, it went on to defend that a benchmark analysis could enable it “to determine whether the fee charged for a specific service exceeds the market value of that service” (para 363). Let’s not forget that this is the EC, which has repeatedly eluded any discussion about yardsticks and benchmarks to meter the effective application of any provision. Within this exercise, the EC surprisingly represented Apple’s initial acquisition fee as “the most suitable and comparable benchmark for the initial acquisition fee charged by Alphabet in terms of business users and end users and the nature of the service it provides” (para 366). That is, the initial acquisition fee that it has not sanctioned or supported following the findings it produced in its May 2025 non-compliance decision. The situation even gets more surprising, given that in footnote 292, the EC highlights the fact that it might subject Apple’s initial acquisition fee to further scrutiny under Article 5(4) DMA. In other words, the EC believes that the most suitable and comparable benchmark for Google’s initial acquisition fee is a fee that another gatekeeper has imposed without its acquiescence, even when it recognises that it might sanction that other gatekeeper for the same reason.

In principle, I do agree that Google’s services are most comparable to those of Apple to the extent that they both operate an app store, they are both gatekeepers and they have both decided to apply an initial acquisition fee. But the comparability of both scenarios ends right there, insofar as Apple’s approach to compliance after the non-compliance procedure is anything but perfect. Drawing inspiration (or comparisons) with that scenario, in my own view, can only deliver fallible results that do not really account for what an authentically robust benchmark could deliver. Here’s why: Apple decided to rename its initial acquisition fee as the store services commission, and maintained a 15% fee on all out-of-app transactions completed upon steering in the following 7 days to the download of the app. The change was effective in early October 2026. Now one faces the unavoidable problem that competition law is confronted with, too.

As far as the European Commission does not disclose the particular benchmark that it applies in the non-compliance decision, it directly lends itself to misinterpretation. One of two options will apply. Either the EC’s reference to a benchmark with reference to Apple’s August 2024 measures in para 362 must be taken at face value and, therefore, the benchmark against which Google’s fees must be compared is that where the initial acquisition fee was set at 5% and applied 12 months after an initial install, followed by a 10% store services fee applied for the same period. Or the EC places a benchmark looking at the current configuration of Apple’s steering fees. The most updated approach eliminates the initial acquisition fee by name and renames the only fee applicable to steering as the store services commission. The fee charges 15% from out-of-app offers with no time limit in terms of its distance from the initial download that justifies the charging of the fee in the first place under Article 5(4) DMA.

Subjecting the benchmarking to the first option would deliver a synecdochical outcome. In its non-compliance decision against Google, the EC declared that its ongoing services fee (comparable to Apple’s own store services fee) “is not allowed under Article 5(4) DMA”, insofar as “it is not limited in time and Alphabet is compensated for the value of the initial acquisition already via the initial acquisition fee” (para 434). The enforcer terms the ongoing services fee as “an ongoing user acquisition service” that gatekeepers are prevented from charging under Article 5(4) DMA, since it only leaves some space to charge the initial acquisition fee as an exemption to the free of charge rule (paras 435 and 445).

From this statement, it naturally follows that Apple’s baseline two-tiered fee structure proposed in August 2024 cannot be anything to go by in the context of the regulator’s appraisal of Google’s own compliance strategy. The only reasonable explanation that would follow from the benchmarking exercise would be to adopt quite a simplistic approach that presents a dangerous precedent. To the extent that Apple decided to charge 5% for the initial acquisition fee, whilst Google charged 10% for the same type of transactions, then the European Commission considers the latter’s as excessive (drawing on para 361). This would precisely entail that the regulator is engaging in price setting in its more direct manifestation, by finding it inadmissible that Google prices its initial acquisition fee as it sees fit.

But actually, if one looks at the broader picture, that conclusion cannot be right either. The EC now considers charging 10% of the steered transaction as excessive. Let’s take that as a given from the non-compliance decision directed at Google. Then, there is something innately wrong with how Apple is now framing its initial acquisition fee, as applicable from October 2026, insofar as it has now been merged into a rounded-up 15% fee applicable without any time limit in sight.

In my own view, these inconsistencies demonstrate the danger in trying to set a benchmark that is not well established nor recognised as the industry’s baseline for operating. I do understand that the initial acquisition fee under Article 5(4) DMA is a new concept that cannot be promptly assimilated to pre-existing fee structures, but the ex ante regulation does not operate in a vacuum, either. The EC is not of the same opinion. It systematically rejected every alternative benchmark proposed by Google relating to the setting of fees in the context of OEM agreements (para 415), of other non-gatekeepers (418) or of third-party marketing expenditure (paras 422-427). And the discussion was, therefore, settled in favour of benchmarking against the Apple scenario, be that as it may.

The Additional Principles Applicable to Steering

Resulting from its analysis of the breaches that Google incurred with respect to Article 5(4) DMA, the EC set out some principles that the gatekeeper must abide by when framing the next iteration of its compliance solution relating to the provision. This is not a given in every single non-compliance decision issued by the EC in the DMA’s context. For instance, the EC did not engage in this exercise on its non-compliance decision relating to Google’s self-preferencing conduct (see here for comment).

The DMA’s enforcer followed the steps of its non-compliance decision against Apple and added some nuance to the content of Article 5(4) DMA. It did so by fleshing out that gatekeepers must not only enable steering in as many destinations as possible, but also in as many types of destinations as possible. Even though app developers will be more prone to linking end users out to their proprietary content, ranging from their own items, products, services or promotional offers, gatekeepers should not effectively limit the capacity of those app developers to link out to other third-party digital operators on their websites, app stores or apps. Additionally, the EC highlighted that steering must be allowed in a technology-neutral manner. In Apple’s case, that entailed allowing web view in-app, but the EC now finds that this modality of steering must also be included within the conditions that make steering possible, i.e., through the charging of the initial acquisition fee, and not the regular fee applied to all transactions that do not fall within the definition of a steered transaction.

To wrap up those principles, the non-compliance decision also emphasises that the gatekeeper must ensure that no fee is imposed for steering, that does not fall within the scope of facilitating the initial acquisition of end users. The EC seems to be reiterating that gatekeepers cannot have additional fees triggered by steering that do not correspond with the initial acquisition fee. Therefore, Google’s current enforcement strategy by which it does not charge an initial acquisition fee but imposes an ongoing service fee for every link-out, does not fall neatly into the EC’s idea of effective enforcement. Just as with other non-compliance procedures, now we will have to wait and see how the enforcer reacts to this change. If one has to go with past enforcement action, no repercussions will follow from the gatekeeper’s deviation of the EC’s initial plan of compliance.