More than a decade after the Commission first opened its investigation into Google Android, the Court of Justice (CoJ) has finally brought the curtain down. On 2 July 2026, the CJEU handed down its judgment (C-738/22 P, Google and Alphabet v Commission), upholding the General Court’s ruling (Case T-604/18, Google and Alphabet v Commission) which imposed a €4.1 billion fine on Google and largely confirmed the Commission’s decision. The judgment not only marks the end of a decade-plus saga but also confirms and builds on recent case law on abuse and the Commission’s Draft Guidelines on the application of Article 102 TFEU (the Draft Guidelines). In doing so, it clarifies important methodological questions in abuse of dominance analysis with implications for digital markets enforcement well beyond Google.
By Ece Ban
Background: how did Google end up here?
Before turning to the CoJ’s judgment, it’s worth remembering the conduct the Commission’s 2018 decision targeted. In 2018, the Commission found that Google had abused its dominant position by imposing contractual restrictions on original equipment manufacturers (OEMs) and mobile network operators (MNOs), strengthening its dominance in the market for general search services. Commission’s 2018 decision targeted three contractual restrictions:
– Mobile Application Distribution Agreements (MADAs): OEMs seeking to license the Play Store were required to pre-install Google Search and Chrome browser.
– Anti-Fragmentation Agreements (AFAs): as a condition for pre-installation of the Play Store and Google Search, OEMs had to commit not to sell devices running versions of Android (so-called ‘Android forks’) that were not approved by Google.
– Revenue Share Agreements (RSAs): OEMs and MNOs received a cut of Google’s advertising revenue in exchange for agreeing not to pre-install a competing general search service within an agreed portfolio of devices.
The Commission found that these contractual restrictions amounted to four separate infringements of Article 102 TFEU, all sharing a common objective: cementing Google’s dominant position in general search services. This shared objective is what led the Commission to treat them as a single continuous infringement.
On appeal, the General Court largely agreed, though it annulled the part concerning RSAs and trimmed the fine accordingly. Google appealed the GC’s judgment to the CoJ. The CoJ, in line with AG Kokott’s opinion, dismissed Google’s appeal. Some of Google’s arguments raised the core methodological questions in establishing abuse: the role of counterfactual analysis, the AEC test, and finally, the single continuous infringement finding. The CoJ’s answers on each are worth unpacking.
The CoJ Holding: Three Takeaways for Abuse Enforcement
(i) No Crystal Ball Needed: Causation Analysis in Abuse Enforcement
As a first ground of appeal, Google claimed that the General Court erred in its review of the causal link between the MADA pre-installation conditions and their alleged exclusionary effects. Google raised this ground in four parts, each addressing a different aspect of how the causality analysis should be scoped.
As a first part of the first ground, Google argued that, in assessing the exclusionary effects of MADA pre-installation, the General Court wrongly took into account the impact of RSAs (although they were not found to be abusive) and thus failed to show that the alleged exclusionary conduct was attributable to MADA pre-installation conditions. Google further argued that the RSAs would have had the same exclusionary effects in the absence of the MADA pre-installation agreements. The CoJ dismissed the argument. It held that assessing whether conduct is capable of producing exclusionary effects doesn’t follow a single fixed template: the Commission can draw on different methods, so long as it accounts for all the relevant factual circumstances. On that basis, whether the RSAs were themselves abusive was not relevant as they were still part of the contextual background against which the MADA conditions had to be assessed (para 197).
Second, Google challenged the causal link from a different angle: it argued that the Commission had failed to establish that the low download numbers were attributable to status quo bias. The argument was that, in analysing the effects of the MADAs, the General Court took into account the low download numbers of competitor apps. According to Google, however, these low numbers were attributable to users simply preferring Google’s products, rather than to any status quo bias created by MADA pre-installation. At its core, this part asks how far the Commission must go in ruling out alternative explanations before a causal link between conduct and effect can be established. The Court noted that it may not always be possible to objectively determine whether user behaviour is attributable to superior performance, echoing its reasoning in Servizio Elettrico Nazionale (para 99). Following that logic, it held that once bias distorting consumer choice has been established to the requisite standard, such a finding, in principle, precludes attributing choice to the undertaking’s superior performance (para 203).
Third, Google raised a narrower evidentiary complaint: that the General Court had wrongly blurred the line between pre-installation and default settings, treating evidence about one as if it applied equally to the other, without requiring the Commission to show that the status quo bias was attributable specifically to pre-installation. The CoJ found this part partly inadmissible, and, to the extent it was admissible, unfounded.
The fourth part is perhaps most consequential for future Article 102 TFEU enforcement, since it goes directly to the methodology for establishing causation: does establishing a causal link between the conduct and effects require a counterfactual analysis? Google argued that the General Court was wrong not to require the Commission to conduct a counterfactual analysis, an argument it had also raised in Google Shopping. Google’s argument was that the MADA pre-installation conditions were the non-monetary consideration OEMs paid for a free-licensing model, and that without that non-monetary consideration, an open-licensing model like the Android OS might not have existed at all. According to Google, the effects of MADA conditions had to be assessed with reference to a proper counterfactual that accounted for this. The CoJ disagreed. Following its holding in Google Shopping (para 231), it held that establishing a causal link does not depend on any single methodology and that the Commission is not under an obligation to carry out a counterfactual analysis (paras 225, 228), provided it duly takes into account all the relevant circumstances.
(ii) AEC Test Marginalised in Digital Markets: ‘Not Possible, Nor Does It Make Sense’
The second ground of appeal takes on the AEC test – and this is where the Court doesn’t just confirm existing case law but moves beyond it regarding its application in digital markets. It’s the part of the judgment worth sitting with for what it might mean going forward. Google argued that the Commission had failed to show the MADA pre-installation conditions were capable of foreclosing as-efficient competitors, and, in particular, that the General Court was wrong not to require the Commission to run the AEC test.
Before responding to Google’s arguments, the CoJ first set out some general considerations. First, the CoJ endorsed the now-familiar two-step test for finding conduct abusive: (i) a departure from competition on the merits, and (ii) the capability to produce exclusionary effects (highlighting that these two conditions are cumulative (para 264)).
On the first limb, the Court noted that competition on the merits can be assessed by reference to whether an as-efficient competitor could replicate the conduct (para 268, citing Servizio Elettrico Nazionale). In this case, however, that kind of analysis wasn’t required for tying, as it constitutes a departure from competition on the merits without there being a need for such analysis. On the second limb, the Court held that this requires showing a restriction of competition through the exclusion of as-efficient competitors, while noting that conduct may also amount to abuse where it impedes competitors at an earlier stage or prevents them from entering the market – referring to the European Superleague case.
The first part of the Court’s analysis regarding the AEC test does not come as a surprise and fits with the incremental building of case law (see Google Shopping, para 263; Servizio Elettrico Nazionale, para 81). It confirms that an AEC analysis is only one way of showing departure from competition on the merits, while holding that there is no obligation on the Commission to examine the efficiency of actual or hypothetical competitors of the dominant undertaking (para 272). The Court emphasises the non-workability of such an analysis in certain markets, specifically ecosystems characterised by significant entry barriers and network effects. Furthermore, in paragraph 278, the Court made a broader statement regarding the applicability of the test in the digital economy, holding that in such markets, where innovation, access to data, multi-sidedness, user behaviour, and network effects play a decisive role, and where high barriers to entry shape competitive dynamics, it is ‘not possible, nor does it make sense’ to base the analysis of exclusionary effects on whether an as-efficient competitor could replicate the conduct. The Court further emphasised the distinction between price and non-price conduct and held that ‘this applies all the more where the conduct itself does not lend itself to a quantitative, price-based analysis’. The final part of the Court’s analysis can be read as a direct extension of the reasoning the Court used in Google Shopping, where it held that the AEC test did not apply because no objective, reliable results could be obtained given the market’s specific structural features. Google Android generalises that logic beyond the narrow facts of self-preferencing in search results to the broader category of market characteristics in digital markets. Building on that earlier case law, the CoJ appears to be carving out a distinct evidentiary burden, bifurcating along a conduct-type and market-characteristics axis.
(iii) Single continuous infringement
Another question addressed by the CoJ is what happens to a finding of a single continuous infringement when one element of the infringement finding is annulled. Google argued that if the RSAs were not abusive, the finding of a single continuous infringement should fall too. The Court disagreed. It held that, notwithstanding the annulment of the finding that the RSAs were abusive, the ‘overall strategy’ underpinning the infringement remained intact. Just as with the effects analysis, the CoJ held that the non-abusive nature of the RSAs did not preclude them from being taken into account as part of the contextual assessment for the purposes of establishing a single continuous infringement (para 389). Therefore, a finding of a single continuous infringement does not require every element to survive appeal; it requires only that the remaining conduct still tell the same story.
Takeaway: where does it leave us?
The judgment of the Court of Justice in Google Android closes out a saga that has spanned more than a decade, but it does more than bring a long-running case to an end. A few broader observations stand out for the development of abuse-of-dominance enforcement, specifically in digital markets.
First, the judgment builds on existing case law, as mentioned above, and also lends judicial weight to the Draft Guidelines. The Court’s use of the two-limb test for exclusionary abuse, as well as its approach to AEC (Draft Guidelines, para 27) and counterfactual analysis (Draft Guidelines para, 66-67) tracks closely with the Guidelines.
Second, the judgment marks the AEC test’s growing marginalisation in digital markets, and a bifurcation along conduct type and market characteristics.
A broader takeaway runs through different parts of the judgment: the CoJ’s reasoning on the counterfactual, status quo bias, and the AEC analysis all come back, at bottom, to the same question: the workability of methodological tools and whether they can be meaningfully applied to the markets they are meant to govern. The CoJ’s ruling reflects a broader acceptance that Article 102 TFEU enforcement must account for the unique characteristics of digital markets, rather than mechanically applying analytical tools developed for traditional markets.
