The EU General Court’s Ruling on Meta’s Gatekeeper Designation (T-1078/23)

The General Court has handed down its judgment in Meta v Commission (T-1078/23), ruling on Meta’s challenge to its DMA gatekeeper designation. Two services were at stake: Facebook Messenger and Facebook Marketplace. Messenger stays designated and the decision on Marketplace is annulled though this service had already been de-designated by the Commission anyway.

Annulments make headlines (at least in the General Court’s press release). But on closer inspection, the Messenger part of the judgment is likely the more consequential one. It addresses a crucial question for digital markets: when does a service embedded within a larger ecosystem constitute its own core platform service (CPS), and when is it just a feature? Or in other words: how to measure the digital world? With AI being woven into search engines, operating systems, and social networks at speed, knowing how to draw those lines matters enormously. The Marketplace chapter raises a different kind of concern. It is less about the substance of the designation than about whether last-minute service changes can be weaponised to derail Commission proceedings.

By Sebastian Steinert and Kena Zheng

I. Messenger: Measuring the Digital World

Daniel Kehlmann’s novel Measuring the World follows two great German scientists, Gauss and Humboldt, each trying to map the boundaries of a world that resists neat categorisation. The General Court faced a structurally similar challenge with Messenger. It was undisputed that Messenger fulfils the definition of a messaging service (NIICS, in the DMA’s terms). The question was whether it can be treated as a standalone CPS, or if it is rather a sub-feature of the Facebook social networking CPS.

This delineation carries significant consequences, directly affecting how the DMA will be applied, and, more specifically, the scope of obligations imposed on gatekeepers. The DMA’s prohibition on tying under Article 5(8), the prohibition on combining personal data without consent under Article 5(2) and the self-preferencing rule under Article 6(5), all apply only to distinguishable services. Whether Messenger is a separate CPS or a feature of Facebook therefore determines whether a whole range of DMA obligations are even triggered solely for Messenger. The same question is now being discussed with regards to AI integrated into operating systems, search engines, and other digital services. The Court itself acknowledged that the question is tricky. As it noted, “the EU legislature was fully aware of the complexity of the composition of certain platform ecosystems that include, inter alia, complementary services which could overlap or be interconnected” (para 98, emphasis added).

1. When a Feature Becomes a Core Platform Service

Meta’s position, maintained from its notification of 3 July 2023 through the Court proceedings, was that even if Messenger technically satisfies the definition of a number-independent interpersonal communications service (NIICS) under Article 2(9) DMA, it should be treated as part of the Facebook online social networking CPS rather than designated separately. Meta’s core argument: Messenger is the chat functionality of Facebook, nothing more.

This refers to an interesting legal question within the DMA’s architecture. When a service can fall into two different CPS categories (here, both a NIICS and potentially part of an online social network) can the Commission simply pick one, or must it conduct a comparative assessment to find the best fit? Meta argued for the latter: first, an a priori assessment of which CPS categories are in play, then an in concreto assessment to determine the closest match (para 83).

To support this legal test, Meta ironically referred to the “autonomous nature” of the DMA’s provisions, namely the definition of online social networking services (para 86). In the Article 5(2) non-compliance proceedings, by contrast, Meta argued vigorously against an autonomous interpretation of DMA provisions when that approach was being used against it (see our blog post on that case). The DMA’s autonomy, it seems, is an argument deployed selectively.

The Court did not explicitly endorse Meta’s in concreto framework, but it found that the Commission had complied with it anyway. The Commission’s analysis first established that Messenger meets the NIICS definition, then considered whether it should nonetheless be treated as a mere ancillary feature of the Facebook social network

Meta brought forward different arguments why Messenger is just a Facebook feature. The Court rejected them all.

1.1. Messenger as a Standalone App intertwined with Facebook

Meta’s argument that Messenger is accessed through the Facebook app collapses before the fact that a standalone Messenger application exists. And this app constitutes the entry point for the [confidential] majority of Messenger’s users (para 42). Messenger’s standalone nature is also showcased by the fact that it can be used with a deactivated Facebook account (para 43). The service delineation must be made from the objective perspective of end users, but how a provider configures and markets its services is relevant evidence of how users will experience them (para 60). Here, the Court sided with the Commission finding that Meta designed, presented, and positioned Messenger as an autonomous service in its own communications to investors and the public (para 59).

Meta also tried to use the large overlap in user groups as evidence for a single service. The firm pointed to data showing that between 80% – 100% of Messenger’s monthly active users were also monthly active Facebook users, and that 100% of Messenger communications took place between Facebook account holders (para 50). The Court’s response: those figures simply reflect Meta’s own design choice to require a Facebook ID to use Messenger. The overlap is a symptom of tying, not evidence of a single service.

Referring to the DMA’s Annex, Meta argued that Facebook and Messenger serve the same purpose, namely “connection and communication” and therefore they should be considered one service. While it is true that the “purpose” can be a valid criterion of delineation (Section D(2)(c)(ii) of the Annex), the Court highlighted that this criterion is an alternative to the delineation along CPS categories (Section D(2)(c)(i) of the Annex). Once Messenger qualifies as a NIICS and Facebook as an online social network, the category distinction is sufficient and no further “purpose” comparison is required (paras 80-81).

Finally, none of Meta’s comparisons with other designated services carried sufficient weight. Meta tried to take the Commission’s analysis in other cases as precedent for Messenger, but the Court didn’t see enough similarities. Apple’s iMessage was never offered alongside an online social networking CPS, so the delineation question between a NIICS and a social network simply did not arise in that decision (para 67). TikTok and LinkedIn’s chat functionalities lacked the key characteristics that made Messenger standalone: a dedicated application as the primary access point, and usability with a deactivated main account (para 68). Booking.com and RentalCars.com were accepted as a single CPS because all services were predominantly accessed through a single interface even if separate apps exist (para 70). The principle here is settled since the Court’s ByteDance ruling: CPS designation is an individual assessment of the specific circumstances of each case, and pointing to other decisions without demonstrating genuine comparability gets you nowhere. After having confirmed that Messenger indeed is an independent CPS, the Court addressed Meta’s arguments trying to rebut the presumptions of being an important gateway.

1.2. Meta joins ByteDance Below the Hurdle of Rebuttals

Messenger significantly exceeded the quantitative user number thresholds in Article 3(2)(b) DMA, triggering the presumption of being an important gateway. Rebutting this presumption requires exceptional arguments that are “sufficiently substantiated” and “manifestly” call the presumption into question (Article 3(5) DMA). The court confirmed the high standard of proof, already established in its ByteDance ruling (para 121).

• Pointing to Messenger’s share of B2C traffic without considering data on usage intensity or time spent is not sufficiently substantiated (para 127).
• Multi-homing data covering only Germany in 2021 says nothing meaningful about the EU as a whole (para 127).
• Comparing Messenger to email services fails, because emails, unlike Messenger before designation, enables interoperability between users of different providers (para 128).
• The claim that business users cannot initiate contact on Messenger was inadmissible before the Court for having been raised for the first time in litigation and in any case, the DMA does not require business users to initiate contact for a service to function as an important gateway (para 133-134).

Comparisons with how iMessage or Gmail were treated in other Commission decisions carry no weight “with no consideration being given to the circumstances in which those services operated as a whole” (para 143).

On calculating end user numbers, Meta argued that only Messenger users who were not simultaneously Facebook users should be counted. Meta’s legal basis for this argument is Article 3(9) which mentions that services should be designated which “individually” are an important gateway (para 146). The Court disagreed with overinterpreting the word “individually”. Based on the definition of NIICS end users in the DMA’s Annex and Recital 20, the Court found that only counting the users which do not also use Facebook would not adequately represent the role and reach of Messenger (para 155).

2. Lessons for Digital Geography

A few general learnings stand out for future designation discussions.

The Value of the Annex: The DMA’s Annex has the same legal value as the substantive DMA provisions (para 74). While this is not a surprise in EU law, the specifics are interesting. Where the Annex specifically instructs undertakings on how to delineate services (Section D(2)(c)), the Court ruled that this methodology is also binding for the Commission, because otherwise the Annex would not provide legal certainty (para 76). The Annex provides for two alternative delineation methods: along CPS-category lines and based on purpose. When two services fall into different CPS categories, no purpose analysis must be conducted.

End User Perspective: The primary lens of delineation is the objective point of view of end users (para 60). How do users access and use the service, does it have standalone relevance, and does it come with its own specific tools and monetisation logic? Investor communications and product positioning can supplement that analysis.

Comparisons with other designation decisions are of very limited use: On multiple occasions in the ruling, the Court has rejected Meta’s efforts to take other services and designation decisions as precedents (e.g. paras 62-71, 141-143, 172-177). Given the highly individual nature of each assessment, it will likely continue to be difficult for Gatekeepers to raise “equal treatment” arguments.

One final argument deserves a brief mention, and it concerns innovation. Gatekeepers do not become tired of framing the DMA as anti-innovation (Apple is front runner with this claim). With Meta’s “technological deterrence” point, the firm suggested that designating standalone features within a platform ecosystem as separate CPSs would discourage companies from developing new services (para 161). The Court rejected this as legally irrelevant. Article 3 DMA does not list innovation among the criteria for gatekeeper designation. Recital 107 DMA does identify innovation as a goal, but it is a goal of the contestable and fair digital sector the DMA aims to create. The DMA fosters innovation by enabling challengers to compete, it does not protect incumbents from regulatory scrutiny when the CPS they develop grow large enough to meet the designation criteria.

Now let’s move to the second part of the judgement: the annulment of the Marketplace designation.

II. The Phantoms of the Marketplace

The central question for Marketplace was whether it had enough business users to qualify as an online intermediation service (OIS). The OIS definition points to the P2B Regulation and requires that undertakings be to offer goods or services to consumers. The Commission used Meta’s internal “power sellers” concept as its proxy for the existence of business users: those posting at least 28 listings in any month of the year and at least 80% of the listings fall into the same category. Those “power sellers” are the phantoms of Marketplace. Meta tried to keep them a secret and later had them disappear by changing its Terms and Conditions altogether. Today we know that they in fact did not disappear but remained present on marketplace after the designation decision (as true phantoms do).

1. Two Preliminary Battles: Standing and Confidentiality

Before the Marketplace designation could be scrutinised on the merits, the Court had to resolve two preliminary questions on standing and on confidentiality.

On standing: the Commission had already repealed the Marketplace designation in April 2025, finding that Meta’s service changes in December 2023 and January 2024 had driven business user numbers below the 10,000-users threshold. Did Meta still have a legal interest in pursuing the annulment? Yes. A Commission repeal operates ex nunc (from the moment of repeal forward) while a General Court annulment operates ex tunc (as if the designation never existed) (para 30). The original designation required Meta to comply with DMA obligations for Marketplace by 7 March 2024. Those obligations bound Meta for the period from September 2023 to April 2025, and their legal effects cannot be erased by a later administrative repeal. The potential basis for damages proceedings alone was sufficient to maintain Meta’s interest (para 29).

On confidentiality: Meta had sought to keep the term “power sellers” out of the public judgment, describing it as commercially sensitive information relating to how it internally categorises certain users. It had already succeeded in labelling it as [confidential] in the Commission’s designation decision. Now, the Court lifted the mask. Meta itself had described the concept as a “legacy” that was “no longer used internally well before the adoption of the contested decision” (para 23). It had also been discussed openly at the hearing. The phantom had already been seen and there was nothing left to conceal.

2. Timing Is of the Essence

What is the right time frame to assess the existence of a gatekeeper position? The DMA provides for a three-year retrospective period to assess whether a service meets the quantitative thresholds in Article 3(2). The Commission had used this same three-year window to assess whether Marketplace qualifies as an OIS at all, i.e. whether it fulfils the CPS definition. The Court found that was an error. The three-year period is strictly for the quantitative thresholds establishing the “entrenched and durable” position of the gatekeeper. For the prior question of whether a service meets the definition of a particular CPS category, the general principle of EU administrative law applies: “the legality of an EU act must be assessed on the basis of the facts and the law as they stood at the time when the measure was adopted” (para 226, emphasis added). The Commission had to assess Marketplace as it was on 5 September 2023, not as it had been over the preceding three years. And this is where things get complicated, because Meta made sure the Marketplace on designation day looked rather different from the one the Commission had been analysing before.

3. Last Minute Changes Hit Time-pressured Commission

The Commission sent its preliminary view to Meta on 26 July 2023, laying out its provisional reasoning including its reliance on the “power sellers” proxy to demonstrate the presence of business users on Marketplace. Five days later, on 31 July 2023, Meta changed the listing rules on Marketplace: a maximum of 20 listings per category per month (para 215). Meta informed the Commission that the “power sellers” proxy was now moot: under the new rules, no compliant users would qualify. To be fair, Meta had flagged this change in its notification of 3 July 2023. But the timing, coming after the Commission’s preliminary reasoning, posed difficulties for the Commission which had to work under a strict 45-working-day designation deadline (Article 3(4) DMA). The clock was already ticking. There was limited procedural space to rebuild an entirely fresh evidentiary basis.

Therefore, the Commission’s response to Meta’s July changes was slim. It pertained that users “could still be considered business users depending on the factual circumstances” even under the new listing limits (para 244). The Court found this insufficient. The Commission failed to specify which factual circumstances it considered relevant. And it failed to explain why those circumstances applied specifically to Marketplace after 31 July 2023. The Commission tried before the Court to explain that users could still publish 240 listings per year in the same category, or that a single listing could cover multiple items, so that business users were still present beyond the eliminated category of “power sellers”. The Court did not accept this: the Commission relied on the (no longer existent) proxy of “power sellers” in its decision, and any additional arguments came too late (para 265).

4. The Phantom was Real

Here is the irony at the heart of the Marketplace annulment. We know from the de-designation decision of April 2025 that power sellers did not disappear from Marketplace after the July 2023 changes. Meta’s own submissions in the de-designation proceedings revealed that approximately 40,000-50,000 power sellers were still present after the designation, still comfortably exceeding the 10,000-users threshold. Only Meta’s further, more comprehensive changes in December 2023 and January 2024 finally pushed the numbers below the threshold. They removed some 150,000-200,000 business users through new C2C enforcement guidelines. So, the phantom was real. The time-pressured Commission just failed to document it sufficiently in September 2023. Had the Commission specified in the contested decision that the listing restrictions did not eliminate commercial activity on Marketplace, merely the specific internal proxy used to measure it, and that a substantial number of business users remained identifiable under alternative measurement, it would very likely have survived review. The annulment is therefore less a substantive verdict on whether Marketplace was a gatekeeper than a failure of adequate reasoning under Article 296 TFEU.

5. A Door Opened for Evasive Tactics?

This matters beyond the specific facts of the Marketplace case. What the ruling effectively signals is that last-minute changes to a service (even made after the Commission has published its preliminary view, once the evidentiary strategy is known) can unravel a designation decision if the Commission does not address those changes with sufficient precision in the final decision. For a regulator working under the 45-day designation deadline, that is a meaningful vulnerability.

We have already seen, in the compliance context, how gatekeepers respond to the DMA with continuous service changes designed to make the regulatory object harder to pin down. In the Article 5(2) proceedings, Meta’s internal documents showed it was aware from the outset that its first “pay or consent” compliance model was insufficient, yet it rolled it out anyway. It then introduced an improvement with a less-personalised option that was still likely non-compliant (featuring mandatory ad breaks and complicated choice screens), and has now moved to another iteration with less-personalised ads and no ad breaks. Each change shifted what the Commission had to assess.

But the compliance context is different from designation. The compliance deadline is set to six months after designation (for most services that is 7 March 2024), so a non-compliance decision can always refer to the period from that date onward. Any subsequent service change only affects the duration of the assessed infringement, not the validity of the earlier finding. Designation has no such anchor. It must reflect the service as it stands on the day of the decision. Our SCiDA Researchers Rupprecht Podszun and Jasper van den Boom have already flagged the risks that delayed compliance poses for effective DMA enforcement. The Marketplace section of this judgment adds another dimension to that risk: the moment of designation itself may become a target for evasive tactics. The Commission’s answer will need to be procedural resilience: anticipating mid-procedure changes and building an alternative reasoning if necessary to react to late factual changes.

III. A Ruling with Lasting Implications

On the surface, the 3 June 2026 judgment changes very little. Messenger stays designated as it always was. Marketplace was already de-designated before the judgment. The map of DMA-covered services looks the same as it did the day before.

Underneath that surface, the ruling carries two significant lessons. First, it provides the most detailed judicial guidance yet on how to delineate integrated services within platform ecosystems under the DMA. Services that fall into different CPS categories can be designated separately, even when tightly integrated. The user perspective is primary, but public positioning, specific tooling, and standalone functionality all feed into the analysis. With AI now being folded into digital services at speed, the method the Court validated here will likely be relevant again soon.

Second, the Marketplace chapter shines a light on a tactic that deserves closer regulatory attention: strategically timed changes that shift the regulatory object during a live designation procedure. Following this judgment, the Commission must assess the latest factual state of a service thoroughly, even if changes arrive the day before the decision is signed. Where such changes appear designed to confuse or obstruct the Commission’s assessment, the DMA already provides a tool: Article 13 DMA’s anti-circumvention provision. This provision could be triggered to prevent the phantoms of designation tactics move to compliance procedures.

The judgement will be remembered for its instructions on how to measure the digital world (Messenger) and how to react to the phantoms within it (Marketplace).