AG Kokott issues Opinion in Google Shopping with the Commission looking set to win round 3 with a knockout

Blogpost 8/2024

AG Kokott recently gave her Opinion on the ongoing saga between the European Commission and Google. She advised the Court to dismiss all of Google’s grounds of appeal against the judgment of the General Court which upheld the Commission’s decision to fine Google €2.4 billion. In her Opinion, AG Kokott raises and answers questions that ‘are of great legal and practical importance’ (para 2). In particular, the AG considers ‘self-preferencing’, ie. where a dominant firm favours its services over those of rivals, to be an independent form of abuse. This means this type of prohibited conduct under Article 102 TFEU has its own separate criteria for application. Importantly, these do not include the strict indispensability test akin to that found in Bronner which requires that rivals have no alternatives to the input of the dominant firm. Not applying Bronner greatly facilitates the finding of an abuse under article 102 TFEU.

Furthermore, with regard to the notion of competition on the merits, it is considered that in certain circumstances, where a dominant online platform decides to alter its business model from offering a fundamentally open infrastructure to offering a more restricted infrastructure, this may not constitute normal competition. The Opinion also finds that the ‘as efficient competitor’ (‘AEC’) principlewhich provides that article 102 TFEU only protects competitors as efficient as the dominant firm, and the AEC test, which is one way of implementing the principle, are not relevant in cases involving super dominant firms like Google. This is because in markets with super dominant firms (e.g. over 90% market share) and high barriers to entry, less efficient competitors may play an important role. The AEC test is also not relevant when the conduct concerns non-price abuses thus potentially eliminating its role in a self-preferencing assessment all together. If this Opinion is endorsed by the Court, it will give competition authorities across the EU the necessary legal criteria and guidance on how to address self-preferencing as an independent abuse under article 102 TFEU. However, there are still aspects of the legal test that remain unclear and must be addressed going forward.


The Google Shopping Saga

 Before analysing the AG’s findings, it is worth briefly explaining how Google’s case ended up before the Court. In 2017, the Commission fined Google €2.4 billion for favouring its own comparative shopping service(‘CSS’) over competing CSSs. As the AG describes, the Commission found that Google abused its dominant position by making ‘results from its comparison shopping service more visually opulent than those from competing comparison shopping services’ which had ‘the effect of increasing data traffic from Google’s general search results website to its product search results website and reducing data traffic to the websites of competing comparison shopping services’ (para 21). In 2021, on appeal, the General Court upheld the €2.4 billion fine and the Commission’s findings that Google abused its dominant position although it dismissed the finding that the abuse was on the market for general search services but rather on the market for specialized product searchers such as CSSs (para 32). In 2022, Google then appealed the judgment of the General Court on various grounds which will generally be addressed below. The final judgment of the Court is still pending.


The legal test for self-preferencing and competition on the merits

AG Kokott’s findings

One of the main contributions of this Opinion is that it sets out the legal test for self-preferencing. The test is largely based on the existing case law of article 102 TFEU but offers novelty in the sense that it adapts the legal tests from past case law to a new type of abuse. In order to provide a test, AG Kokott refers to article 102(c) which expressly governs the unequal treatment of trading partners or competitors (para 75). The AG then states that although the list of provisions in article 102 TFEU is not exhaustive, unequal treatment that is similar to and as harmful as article 102(c) TFEU (eg. self-preferencing) may also be classified as an abuse (para 76). After having identified the equal treatment criteria, the AG also refers to the general criteria in the case law such as the role of the special responsibility of a dominant firm and the fact that competition has already been weakened through recourse to means outside the scope of competition on the merits leading to a distortion of competition (para 77). AG Kokott also seems to suggest that in order to determine whether a difference in treatment through self-preferencing exists, it must, similar to article 102(c), put competing CSS’s at a competitive disadvantage in a way which deviates from competition on the merits (para 78). Last, as will be discussed further on, the Bronner criteria, and in particular indispensability, are not part of the legal test for self-preferencing (para 81).

AG Kokott then applies the notion of competition on the merits or normal competition to Google’s conduct. Competition on the merits is an abstract notion which provides that a firm may acquire a dominant position on its own merits and thus not every exclusionary effect on rivals is detrimental to competition so long as the dominant firm’s conduct constitutes normal competition. In order to find an abuse under article 102 TFEU, it is therefore necessary to not only establish anticompetitive effects but also that the conduct deviates from competition on the merits. The Court has in SEN provided a negative definition stating that implementing a practice which makes no economic sense, other than to exclude competitors in order to then raise prices obtained from a monopoly position, does not constitute competition on the merits.  Similar to the General Court, the AG refers to the business model of Google to demonstrate how its conduct was outside the scope of normal competition. Google’s business model was based on offering a “fundamentally open infrastructure designed to attract maximum number of internet users and to generate maximum volume of data traffic” and therefore Google has always granted competing CSSs access to its general search result page to maximize these benefits (para 92). Therefore by changing its business model to place its own CSS at the top of its search page, Google had diverted traffic to its own CSS, not based on being better quality or more useful for consumers, but rather from the exploitation of its dominant position (para 97). Furthermore, Google’s previous comparative shopping service called Froogle was unsuccessful and only once Google changed its business model did that situation change (para 97).


One of the issues that may be raised with the AG’s Opinion is that although it sets the foundations of the legal test for self-preferencing, further clarifications are still necessary for the test to be consistently applied. By referring to article 102(c) TFEU, it would seem that the legal test involves first identifying whether dissimilar conditions are applied to equivalent transactions as found under article 102(c). This requires determining whether the dominant firm favoured its own downstream firm (i.e. applied dissimilar conditions) over rivals using the same input (i.e. equivalent transactions). Yet how this is to be applied in the context of self-preferencing requires more clarity. The case law on article 102(c), for example, offers the dominant firm the possibility to provide objective justifications for applying dissimilar conditions such as the fact that providing a service to downstream competitors entails a higher cost for the dominant firm.

Second, the AG seems to require a competitive disadvantage on downstream trading partners which deviates from competition on the merits. Furthermore, she also seems to suggest that there must be a competitive disadvantage that is at least capable of adversely affecting competition(para 98). Competitive disadvantage, like the application of dissimilar conditions to equivalent transactions, is another fundamental requirement for the application of article 102(c). The issue is that the AG has not clarified whether the specific case law of article 102(c) is also relevant for self-preferencing or not. The Court’s judgment in Case C‑377/20 MEO requires a high threshold for determining a competitive disadvantage whereas in  Case C-95/04 P British Airways and Case T‑301/04 Clearstream, the threshold was a lot lower. This is because in these latter cases, competitive disadvantage (ie. effects) was either presumed due to the market structure or applied without determining whether the discrimination actually affected the ability of trading partners to compete. It would seem that the AG considers the threshold of effects to be quite low for unequal treatment by self-preferencing in the specific case (para 92). This may be due to Google’s super dominance in the market for general search and its higher special responsibility not to distort competition (para 152). The General Court in Clearstream, for example, used the monopoly position of the dominant firm to justify applying a lower threshold of competitive disadvantage. These considerations are of course fundamental for the application of self-preferencing since applying a low or high threshold of effects can significantly affect the outcome of an investigation by competition authorities in the future.

With regard to competition on the merits, AG Kokott provides valuable insight as to how the underlying logic of this abstract concept may fit into the context of a case of self-preferencing. What AG Kokott essentially says is that had Google competed under normal conditions, it would have applied its business model to its own CSS which would most likely have been prone to the same demotion as rival CSS’s due to it not being better quality or more relevant for consumers. Normal competition would mean that Google’s CSS would appear at the top if its CSS was the best quality and most relevant for consumers.

Scholars have been critical of this reasoning due to the ambiguity and subjectivity of determining when a change in a dominant firm’s business model constitutes competition on the merits or not. This assessment of competition on the merits does arguably lead to an element of subjectivity. However, AG Kokott does not base her reasoning merely on the fact that Google changed its business model. Rather, the AG supports this argument with the specific circumstances that Google’s previous CSS called Froogle was unsuccessful up until the change in the business model supporting the argument that the only purpose was to eliminate rivals. This would suggest that if Google’s CSS was successful before it changed its business model, then it might have been considered competition on the merits. Therefore it might not always be the case that a dominant firm that self-preferences by changing its business model does not constitute competition on the merits. The elements of ambiguity and subjectivity are of course still present to some extent. It may be worth noting here that more economic approaches involving the AEC principle or test also leave a degree of these elements.


Bronner not applicable to discriminatory self-preferencing

AG Kokott’s findings

 One of the main arguments in Google’s appeal was that its Shopping Units displayed on its general results page is a separate infrastructure and therefore Google refused access to this separate infrastructure thus requiring the application of the Bronner criteria. The AG (unsurprisingly) dismissed applying the Bronner criteria to self-preferencing by relying on well-established case-law such as  Case C‑152/19 P Deutsche Telekom. In her analysis, AG Kokott provides a complete explanation of the underlying logic (ie. freedom of contract and incentives to invest) underpinning the Bronner criteria for refusal to supply (para 83 to 87). Furthermore, she explains that these criteria exist to ensure that the main purpose of article 102 TFEU, which is to safeguard competition as a whole and not the interests of individual competitors, is protected (para 87). The AG then holds that in light of the rationale of the Bronner criteria, this cannot be applied to unequal treatment through self-preferencing (para 88). In this regard, AG Kokott means that since Google had already granted access to its services, there would be no interference with Google’s freedom of contract nor would it affect its incentive to invest in essential infrastructure (para 91). It would also be consistent with the main purpose of article 102 TFEU to not protect individual competitors because access had already been granted and therefore firms would not be free riding on the dominant firm’s investments. (para 87).

After stating that Bronner is not applicable, AG Kokott considers self-preferencing to constitute an independent form of abuse like margin squeeze (para 89 and 90). This is in line with what was held by the General Court and is based on case law such as Case C‑165/19 P Slovak Telekom. The AG states that like margin squeeze, Google’s conduct involved ‘unequal treatment between that undertaking and its competitors in relation to conditions of access to an input that is essential to business on the downstream market’ (para 95). Last, referring to the case law on margin squeeze, AG Kokott considers that the fact that a dominant firm’s input, such as Google search, is indispensable can further support the fact that self-preferencing may lead to anticompetitive effects (para 89 and 104).


The Bronner criteria were developed by the Court for cases of refusal to supply and, in particular, establishes the requirement that the dominant firm’s asset or input must be indispensable for competitors. The reason for applying indispensability, which is extremely difficult to fulfil, is that the Court is well aware that obliging a dominant firm to deal with rivals infringes upon their right to freedom of contract and also affects their incentives to invest in essential infrastructure since competitors could free ride on their investments. However, as the AG held, the rationale of Bronner was not affected since Google already granted access to its service.

When referring to an independent form of abuse, the AG means that self-preferencing is a form of discrimination, like refusal to deal or margin squeeze, but independent in the sense that it involves a different type of behaviour that therefore requires the application of separate criteria. The reference to margin squeeze, which has clearly been established by the Court as an independent form of abuse distinct from refusal to supply, is also significant as this means that the strict Bronner criteria are not applicable. Therefore, since self-preferencing is an independent form of abuse, it has its own specific criteria to be applied that does not require demonstrating indispensability.

It has been argued by commentators that indispensability should be applied to all forms of exclusionary discrimination which would include self-preferencing. The Court has been wary of adopting this approach for a variety of reasons but it seems that the major concern is that it would make finding an abuse under article 102 TFEU very difficult. AG Kokott, like the Court, is (unsurprisingly) unwilling to extend indispensability to new forms of exclusionary discrimination. It is worth noting that this argument can also be supported by economic literature where it has been demonstrated that monopolist firms, similar to Google, which are vertically integrated into downstream markets, often have the incentive to exclude rivals. Therefore, applying the very strict and demanding indispensability test, except for in the case of refusal to supply, may be considered unreasonable since it would result in anticompetitive conduct going unpunished. Regardless of the incentives of the dominant firm, the implications of not applying Bronner and in particular indispensability to self-preferencing will significantly facilitate competition authorities’ investigations going forward.

Last, what is also particularly relevant from this Opinion is the fact that a firm like Google that is ultra-dominant and provides a service that is almost akin to an essential facility and thus indispensable, may also facilitate finding an abuse due to the likelihood of anticompetitive effects being a lot higher. Like in the margin squeeze case law referred to above, providing unfavourable access conditions to an indispensable input to downstream competitors can significantly harm those competitors and thus competition. Therefore, the characteristics of the specific market are what will determine whether the conduct of the dominant firm is an abuse or not. Had Google’s market share been lower and its service less like an essential facility, then its conduct may have been considered competition on the merits since competitors would have had alternatives.


Role of the counterfactual and the AEC test

AG Kokott’s findings

With regards to the role of the counterfactual, AG Kokott rejects Google’s appeal finding that the Commission was not under the obligation to carry out a counterfactual analysis of the effects on rival CSS’s. The counterfactual, as explained by the AG, ‘reflects an actual situation ‘that is initially similar but whose development is not affected by all of the practices at issue’’ (para 168). The aim in this case was to identify whether there is a causal link between Google’s conduct and the exclusion of rival CSS’s (para 160). Of particular note in this analysis was Google’s argument that the counterfactual analysis should have assessed its conduct separately, i.e the promotion of its own CSS and separately the demotion of rival CSS’s (para 178). AG Kokott however specifically establishes that the conduct of self-preferencing in this case involved both the promotion of Google’s own CSS and the deduction of rival CSS’s (para 179). The AG considers that these elements are inextricably combined and operate together to support Google’s self-preferencing (para 179).

Moving on to the AEC test, Google challenged the fact that the Commission did not apply the test and examine the efficiency of competing CSS’s. In line with past case-law, AG Kokott refers to the AEC principle which establishes that article 102 does not protect less efficient competitors (para 193). She is (again unsurprisingly) very clear that this does not mean that less efficient or smaller competitors that do not possess the same economies of scale as dominant firms do not merit protection (para 193). In line with the Court’s judgment in Case C‑23/14 Post Danmark II, where she was also the AG, if the structure of the market presents for example high barriers to entry which make it unlikely that competitors can become as efficient on the dominant firm, the pressure of less efficient competitor ‘is capable of ensuring that the market structure and the choices available to consumers do not deteriorate further’ (para 195). Therefore she says that the ‘as efficient competitor test’ is not applicable in this context. Furthermore, in line with the case law, she holds that the AEC test is not generally applicable, let alone an essential prerequisite, for demonstrating that a conduct is within the scope of competition on the merits (para 196).


What is interesting in the analysis of the counterfactual is that the AG clearly establishes that the conduct of self-preferencing, in this specific case, involved both the promotion of Google’s own CSS and the deduction of rival CSS’s. Furthermore, these elements were inextricably combined to support Google’s abuse through self-preferencing. This combination of elements is not a condition that AG Kokott brings up when establishing the criteria for an abuse. It would seem that, according to the AG, at least for determining anticompetitive effects, this combination is relevant. It is important to note that it would seem that this combination of elements was specific to Google’s self-preferencing and is not necessarily to be applied in all cases of self-preferencing. Going forward, this leaves a degree of uncertainty since it is still not clear whether self-preferencing always requires only the element of demotion or promotion or both.

With regards to the structure of the market and the role of the AEC principle, AG Kokott is consistent with the case law and refers to Post Danmark II where the emergence of an as efficient competitor is practically impossible. In addition, in the footnotes, the AG also refers to paragraph 101 from Case C‑377/20 SEN to support this point. The reference to this paragraph is not entirely correct since the Court in SEN did not says that the emergence of an as efficient competitor was practically impossible, as held in Post Danmark II, but rather that it was impossible for a hypothetical as efficient competitor to replicate the exact conduct of the dominant firm but not similar conduct. Another critique of the application of both these judgments (SEN and Post Danmark II) is that they both concerned legal monopolies and thus the dominant positions were not obtained on their own merit. Google on the other hand earned its dominant position on its own merit, and thus was very efficient, which could support the fact that the AEC principle should at least be considered. This is because, due to Google being very efficient, one may also consider it reasonable or even fair to address whether the reason other CSS’s were eliminated from the market was due to their inefficiency. In the context of this specific case, demonstrating how competitors were less efficient may be hard but in other scenarios, it may be asked whether competitors where able to for example replicate the conduct of the ultra-dominant (or even monopoly) firm.

On a final note, AG Kokott also considers that the AEC test should not be applied to non-price abuses such as self-preferencing. Academic literature has, nonetheless, shown that the AEC test can be adapted to non-price abuses. Furthermore, the Court in Case C‑680/20 Unilever suggests it could, among other tests, be relevant for non-price abuses. AG Kokott is critical of the Court in Unilever. She advises the Court that this judgment either be clarified if not corrected (para 197). Although not providing an elaborate reasoning for her view, it would seem that AG Kokott is quite sceptical towards the AEC test in general and therefore extending it to non-price abuses would lead to it having a greater role in the application of article 102 TFEU.


Key takeaways and implications for digital markets and beyond

The legal and practical implications of this Opinion, like AG Kokott states, are significant. First and foremost, all of Google’s appeals are dismissed by the AG with a hefty fine of €2.4 billion being upheld. With regards to the future of article 102 TFEU, self-preferencing is considered an independent form of abuse like margin squeeze. Furthermore, the legal test has to some extent been established but it is still unclear if and how dissimilar conditions to equivalent transactions should be applied and also what the threshold of effects ought to be. In addition, the notion of competition on the merits and its application to business models of dominant online platforms is now becoming well established. Moreover, the Bronner criteria are unlikely to be extended to future cases of discrimination, except for refusals to deal. Last, in markets with high barriers to entry where inputs held by dominant firms are equivalent to essential facilities, the role of the AEC principle and test appears less relevant. All in all, solid legal foundations have been set, but without further clarity regarding certain aspects of the legal test, the application of self-preferencing under article 102 TFEU will become messy and fragmented across the EU. Providing more specific and clear criteria is key to ensuring national competition authorities and courts apply the self-preferencing legal test consistently.

Moving forward, dominant firms will have to be wary of conduct involving self-preferencing. However, the specific market structure in question, with high barriers to entry and an input that was somewhat equivalent to an essential facility, seems to be what led to the finding that Google had abused its dominant position. Furthermore, cases involving discriminatory leveraging by dominant firms into downstream markets often involve firms that are monopolies. Self-preferencing cases are therefore unlikely to be brought under article 102 TFEU against firms that do not operate on a market with these characteristics. Economic dependence may be the exception. Dominant online platforms will still have to be cautious, in particular since the framework for applying self-preferencing is still to some degree subjective and ambiguous. Changing a business model from a fundamentally open infrastructure to a more restricted infrastructure may not be considered normal competition. Additionally, a dominant platform favouring its own service over rivals’ due to higher costs of serving them may not be regarded as a legitimate economic justification.