Discriminatory Leveraging Plus: The Standard for Independent Self-Preferencing Abuses after Google Shopping (C-48/22 P)

Table of Contents: 1. Introduction and background – 2. Dispensability of the indispensability criterion for independent discriminatory leveraging abuses – 3. Standard of the discriminatory leveraging abuse: what is the plus? – 4. The what-if dilemma: causation and the burden of proof – 5. The never-ending tale of the as-efficient-competitor-test – 6. Conclusion.

Abstract: Almost fourteen years after the European Commission (EC) opened proceedings against Google for abusive behaviour, the European Court of Justice (ECJ) has ruled, confirming in its findings the EC decision of 2017 and siding with the 2021 decision of the General Court (GC) as well as the Advocate General opinion. In its Google Shopping judgment, the ECJ paved the way for an independent self-preferencing abuse and outlined the necessary prerequisites for such a type of abuse, which are generally transferrable to other situations. Discrimination – self-favouring and demotion of competitors – plus potentially anticompetitive or exclusionary effects seen in light of all the specific circumstances and accompanying practices constitutes an own form of abuse. Bronner is not applicable. The Court further emphasised the burden of proof for infringements on the Commission but applied a workable standard of proof that allowed the Commission to discharge its burden of proof without applying a full counterfactual analysis. It also clarified the applicability of the AEC-test.

Keywords: self-preferencing – self-favouring – discrimination – effects – AEC test – as-efficient-competitor test.

1.   Introduction and background

Almost fourteen years after the European Commission (EC) opened proceedings against Google for abusive behaviour, the European Court of Justice (ECJ) has ruled,[1] confirming in its findings the EC decision of 2017[2] and siding with the 2021 decision of the General Court (GC)[3] as well as the Advocate General opinion[4]. In its Google Shopping judgment, the ECJ paved the way for an independent self-preferencing abuse and outlined the necessary prerequisites for such a type of abuse, which are generally transferrable to other situations. The Court further emphasised the burden of proof for infringements on the Commission but applied a workable standard of proof that allowed the Commission to discharge its burden of proof without applying a full counterfactual analysis. It also clarified the applicability of the AEC-test.

1.1. The disputed conduct

First things first: The problematic conduct in Google Shopping essentially results in Google’s preferential treatment of its own comparison-shopping service thereby leveraging its market power from the dominant market position in horizontal search to vertically integrated markets and, at the same time, demotion of its competitors.

In detail, Google gave results from its own Google shopping service a more prominent position than other comparison-shopping services on its general search page. The results of the Google shopping service would be displayed separately from and above the general search results (i.e., in highlighted boxes), including images of the purchasable items. The other comparison-shopping services were only shown in the general search results (i.e., as a typical blue link). Here, they were not only subject to the relevance-based ranking of the search results but demoted in their positioning within the search results by a change in the Google search algorithm (e.g., the so-called Panda algorithm). The Google shopping section, which contained both paid advertising as well as relevant search results, would always be displayed on top. As the default option, it looked more attractive to consumers. As a result, Google's practices were in fact a combination of two different practices creating anti-competitive effects in their combination, as now emphasised by the ECJ. First, Google created ‘(i) the more favourable positioning and display of Google’s own specialised results within its general results pages than the positioning and display of results from competing comparison-shopping services’.[5] Secondly, Google ensured ‘(ii) the simultaneous demotion of results from competing comparison shopping services by the application of adjustment algorithms’.[6]

1.2.  A long journey

As already indicated, the case has a long history, which might continue with damages actions.[7] Originally, the case was initiated by third-party complainants suffering from Google's practices. Amongst 24 others,[8] Foundem, a UK-based vertical search service and competitor of Google, initially originated the case at the EC. After the commitment proceedings initiated in March 2013, failed in September 2014,[9] the EC moved on to render its infringement decision in 2017. The EC found that Google had abused its dominant position by on the one hand decreasing traffic from its general results page to competing to comparable shopping services but on the other hand increasing traffic to its own, i.e., self-preferencing.[10] The EC based the decision inter alia on the enhanced attractiveness of the boxes, the click rate of customers, and the accompanying traffic diversion of the above-described practices.[11] In its decision, the EC did not refer to types of abuses previously recognised like a refusal to deal or tying and bundling.

In 2021, the GC largely backed the EC and nearly dismissed all of Google’s pleas. According to the GC, Google’s self-preferencing practice distorted competition and could be seen as a deviation from competition on the merits.[12] The GC categorised the practice of Google as self-preferencing and a form of leveraging discrimination and its own type of abuse, distinct from a refusal to supply and the Bronner[13] case law[14] – one of the main arguments of Google –, thus limiting the scope of the essential-facility doctrine.[15] The GC underlined the special circumstances of the case and even relied on Google’s ‘super-dominance’[16] or the obligation of non-discrimination arising from the provisions for internet access providers under Regulation (EU) 2015/2120[17]. The GC further backed the EC in its finding of effects but for those on 13 national markets.[18] According to the GC, it sufficed that Google’s practices could potentially create a negative impact on competition.[19] Furthermore, the GC held that Google failed to establish a convincing counterfactual for the reduction in traffic, which was in the centre of the EC effects analysis.[20] Also, the as-efficient-competitor (AEC) test was judged not to be mandatory to establish abuse under Article 102 TFEU.[21] An objective justification was not granted by the GC.[22] Lastly, Google disputed the calculation of the fine due to the novelty of the theory of harm.[23] Conversely, the GC saw the uniqueness in this novel theory of harm, which deserved a corresponding serious penalisation.[24]

Against this decision, Google appealed before the ECJ on four grounds. First, Google argued that the EC failed to prove failure to provide access to Google’s search algorithm and, second, to sufficiently substantiate the leveraging of market power to the downstream product market. Third, Google further aimed for judicial review of the applied standard of causality between the alleged abusive behaviour and its market effects. Lastly, Google argued for the invalidity of the judgment based on an understanding of the mandatory application of the AEC test. These will be dissected in the following.

2.   Dispensability of the indispensability criterion for independent discriminatory leveraging abuses

2.1.  We do not need Bronner...

For the entirety of the fourteen years of the whole Google Shopping proceedings, the competition world argued about the correct legal test and, in particular, the possible applicability of the so-called Bronner-criteria for duty to access or supply cases to self-preferencing abuses.[25] According to the Bronner-criteria, a refusal to access or supply constitutes an abuse, (i.) if that refusal is likely to eliminate all competition in the market in question on the part of the entity applying for access, (ii.) such a refusal is incapable of being objectively justified, and (iii.) the infrastructure, in itself, is indispensable to carrying on that undertaking’s business, inasmuch as there is no actual or potential substitute in existence for that infrastructure.[26]

It therefore came as no surprise that the appeal demanded a final word by the ECJ on the matter in its first ground of appeal. The applicants’ arguments can be summarised in the following: Google’s conduct, in essence, constituted a refusal of a duty to supply and consequently be assessed against the Bronner-criteria, because Google ‘failed to provide competing comparison shopping services with access to prominent, dedicated boxes on its results pages, which had rich display features, and which were not prone to demotion by algorithms such as Panda’.[27]

On the contrary, the ECJ sided with the GC. The case is not about ‘access of competing comparison shopping service to boxes, but to access to Google’s general results pages under non-discriminatory conditions’.[28] While the practice in Google Shopping is not unrelated to the issues of access and discrimination, so are many exclusionary practices.[29] Indeed, practices such as rebates, exclusivity agreements, bonus schemes, price reductions, tying, bundling, and other forms of predatory behaviour can lead a dominant company to treat downstream customers differently by offering them varying conditions.[30]

The ECJ hereby followed its line of reasoning known from TeliaSonera,[31] Slovak Telekom[32] and Lithuanian Railways:[33] the Bronner-criteria are only relevant in the ‘specific circumstances’ of a refusal to access or supply situation and cannot be transferred to other abusive conduct.[34] The ECJ repeated that only a finding of a refusal to access or supply abuse ‘has the consequence of forcing that undertaking to conclude a contract with that competitor’, which constitutes a severe interference with the dominant company’s ‘freedom of contract and the right to property’.[35] In Google Shopping, according to the ECJ, the situation is different and does not amount to the same level of interference with Google’s freedom of contract and the right to property, since access to the general results page was given, albeit, on discriminatory conditions.[36]

2.2.  ...because we have a distinct form of abuse

The ECJ rightly gets to the heart of the distinction and paves the way for the recognition of an independent discriminatory leveraging abuse: not concluding a contract at all or breaking off contractual relationships and consequently being forced to conclude a contract represents a significantly more intensive infringement of Google’s freedom of contract than if Google grants access to its general results page but failed to provide similar visibility and ensure equal treatment when it comes to the display of comparison shopping services.[37] The decisive factor is therefore the relevance to fundamental rights, such as freedom of contract, and the rule remains that only an explicit refusal to access or supply if you have not previously contracted or have broken off the contractual relationship falls under the Bronner category. Therefore, discriminatory conditions of access are a different category.[38] In fact, the ECJ sides – rightfully so – with the GC and is even more explicit by stating that there is such a thing as a distinct independent form of discriminatory leveraging abuse.[39]

3.   Standard of the discriminatory leveraging abuse: what is the plus?

Consequently, the heart of the appeal circled around the question: what constitutes an independent discriminatory leveraging abuse, especially in the form of self-preferencing? When does discriminatory leveraging depart from competition on the merits?

3.1.  Competition on the merits

As a prerequisite, a definition by the ECJ of competition on the merits or behaviour that no longer constitutes competition on the merits would indeed be useful. The ECJ kept it rather vague by reiterating its rather circular Superleague[40] definition of competition on the merits by stating that it may ‘lead to the departure from the market or the marginalisation of competitors which are less efficient and so less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation’.[41] While this definition might be used as a starting-point, a ‘departure from the market’ or ‘marginalisation’ of less-efficient competitors does not offer any positive definition of competition on the merits. Rather, it seems to only refer to what competition on the merits does not entail.

Recently, the great Heike Schweitzer, in her last article published after her premature death together with her doctoral student Simon de Ridder, provided for a much more workable definition of competition on the merits, according to which ‘it stands for the general idea that, first, dominant undertakings shall compete on account of their skills, abilities, and superior performance, striving to offer consumers the best value for money or most attractive product or deal instead of competing by obstructing their competitors’ effective access to markets, resources, data, or customers or by otherwise creating competitive disadvantages for actual or potential rivals; and, secondly, that they must consider the effects of their conduct, due to their ‘special responsibility’ not to allow their conduct ‘to impair genuine undistorted competition on the common market’’[42] Rooted in the Ordoliberal idea of Leistungswettbewerb,[43] Schweizer and de Ridder show that competition on the merits is directed at market-normal behaviour centred around a consumer-oriented model of competition.

At least and in line with their negatively connoted definition, by determining when discriminatory leveraging constitutes an abuse under Article 102 TFEU, the ECJ reasoning allows the establishment of a test to determine when discriminatory leveraging departs from competition on the merits. Still, one has to put the puzzle pieces together that the ECJ left[44] in its assessment of the first and second ground of appeals, to determine what we call an independent discriminatory leveraging abuse plus.

3.2.  Discrimination needed

Firstly, an independent discriminatory leveraging abuse needs a form of discrimination.[45] In the case at hand, on many occasions, the ECJ clearly flags the special discriminatory conduct of Google, which amounted to (i) giving preferential positioning and visibility to its own specialised results on general results pages compared to those of competing comparison shopping services, and (ii) simultaneously lowering the ranking of results from these competing services through the use of adjustment algorithms.[46] Thus, the discrimination in Google Shopping entails not only a self-favouring conduct alone but at the same time, a demotion of competitors.

3.3.  Exclusionary effects plus specific circumstances

However, any form of discrimination is not enough, according to the ECJ, who sides with the GC.[47] Rather, it must secondly – based on specific, tangle points of analysis and evidence – give rise to at least potentially anticompetitive or exclusionary effects and all the specific circumstances and accompanying practices need to be taken into account.[48] According to the ECJ, the conduct, market(s) in question or the functioning of competition on that or those market(s), including obstacles to entry or growth or other blocking measures regarding entry or growth on that or those market(s), the position and competition on related and neighbouring markets, i.e., the leveraging of market power, and other factual circumstances, have to be taken into account.[49]

For the case at hand, the ECJ specifically backed the GC by accepting three circumstances: (i.) the importance and irreplaceable of traffic generated by Google’s general search engine for comparison shopping services, i.e., positive network effects, (ii.) the behaviour of users when searching online, i.e., users concentrating on the first three to five and generally only the visible search results, and (iii.) the impact of diverted traffic, i.e., ‘that traffic represented a large proportion of traffic to competing comparison shopping services and that it could not be effectively replaced by other sources.[50] The (potential) anticompetitive effects of the described discriminatory conduct taken together with these circumstances were, according to the ECJ, ‘capable of characterising the existence of practices falling outside the scope of competition on the merits’.[51] At the same time, the ECJ clarified that issues brought forward by the GC, such as the mentioned ‘super-dominance’ or the obligation of non-discrimination arising from the provisions for internet access providers under Regulation (EU) 2015/2120, were not necessary and decisive to confirm the abusive conduct of Google in the case at hand.[52]

3.4.  What about other forms of discrimination?

We now know the abstract requirements for an independent discriminatory leveraging abuse: a discrimination plus (potential) anti-competitive effects plus an assessment of the relevant circumstances at hand.[53] Yet, the judgment leaves room for interpretation.

Firstly, the ECJ is ambivalent whether every discriminatory conduct needs to consist of both elements – self-favouring and demotion. In some parts of the judgment, the Court mentions a combination of both[54] while in other parts, the sole reliance on self-favouritism or discrimination[55] seems to be sufficient.

Here, one needs to consider that the self-preferencing conduct in the case at hand – the placement of the Google shopping service separately from the general search results in above-placed highlighted boxes – would have already included a demotion of competitors. Every self-preferencing conduct necessarily entails that a competitor’s services is treated less favourably and is, therefore, as compared to the favoured own service, demoted. However, in Google Shopping, the competitor comparison shopping services were not only included in a lower ranking in the highlighted boxes but on the general search results page, which is subject to the described further demotion.

However, this does not mean that mere self-preferencing without further demotion cannot be considered as abusive. The ECJ highlighted time and again the specific factual circumstances of Google Shopping and corresponding case-specific assessment.[56] We know that once the – quite severe – form of discrimination consisting of both self-favouritism and further demotion comparable to Google Shopping is given in a case, an independent discriminatory leveraging abuse is more likely assumed. In that sense, further demotion serves as a high indicator. However, the anti-competitive nature anyway depends on the plus-factors. Should other self-preferencing conduct – already constituting discrimination and anyway entailing a certain element of demotion – fulfil the effects-based and all-circumstances plus-factors, a discriminatory leveraging abuse cannot be ruled out from a theoretical standpoint.[57]

Secondly, on the plus-factors itself, Google Shopping is in line with former case law requiring more than a simple discrimination.[58] The judgment makes clear that even discrimination in the form of a combination of self-preferencing with competitor-demotion is not always harmful to competition and therefore does not qualify as a ‘prima facie’ or in the literature also called ‘by-object’ abuses[59] that, according to ECJ jurisprudence,[60] goes in hand with a rebuttable presumption of anticompetitive effects, only requiring an effects-analysis on the part of the authority in case of a rebuttal.

Rather, Google Shopping is in line with case law taking an effects-based approach, considering all relevant circumstances for conduct, which is not by nature so harmful that it can only be rationalised to harm competition.[61] Only when these conditions are fulfilled, does the dominant company depart from competition on the merits. This finally puts a nail in the coffin of both the argument that self-promoting is always common business practice and the often-brought infamous example of supermarket shelf allocation for their own-brand products, which is seen as unproblematic conduct.[62] The ECJs effects-based and all-circumstances approach distinguishes common business practices of self-promotion from harmful self-preferencing.[63] Transposed to the supermarket example: Supermarkets must do more (the plus) than allocate favourable shelf space for their own-brand products to commit an independent discriminatory leveraging abuse in the form of self-preferencing. A further demotion of competitors other than positioning the own products more favourably, similar to Google Shopping, serves as a strong indicator. Even without such demotion, the effects and surrounding circumstances decide on the anti-competitiveness.

Still, the judgment in Google Shopping only provides a direction into what the plus must entail, i.e., the relevant threshold when the circumstances allude to an abuse. Naturally, the effects-based and all-circumstances approach amounts to a case-by-case analysis. The circumstances may but are not required to resemble Google's actions and standing within the pertinent and adjacent markets. However, the case of Google Shopping provides illustrations of factors that may be significant, especially in digital markets, including user behaviour and the role of gateways.[64] Consequently, the threshold for consideration should be established at a relatively elevated level, although, e.g., a ‘super-dominance’ is not necessary. Therefore, national and upcoming EU-case law[65] will help to further clarify the relevant threshold of circumstances for independent discriminatory leveraging abuses in the form of self-preferencing.

4.   The what-if-dilemma: causation and the burden of proof

Similar to the GC proceedings, the applicants claimed an error in the analysis of the causal link between the alleged abuse and its likely effects as well as the wrongful attribution of the burden of proof regarding a counterfactual analysis on Google instead of on the EC.

At this stage, also interesting admissibility questions came up. The ECJ once again outlined the distinction between points of law and points of fact. The analysis is simple and straightforward: The question of whether a counterfactual analysis is necessary and which (abstract) criteria apply are questions of law which are amenable to judicial review in an appeal before the ECJ.[66] On the other hand, the concrete application of such a counterfactual would amount to questions of fact which can only be subject to an appeal if the facts were distorted. As the arguments were all about whether a counterfactual analysis by the EC was required by law and which criteria the EC must assess under the relevant legal test, the claimants alleged errors of law which are admissible in the appeal procedure. Otherwise, the appeal would be ‘deprived of parts of its purpose’.[67]

4.1.  Burden of proof vs standard of proof for the counterfactual analysis

In the judgment, the ECJ reaffirms the almost obvious statement that the burden of proof for the existence of an infringement of competition law lies with the EC.[68] However, more important than the question of the attribution of the burden of proof was the issue how the GC dealt with the different elements of evidence and which evidence it regarded as sufficient for a fact to be proven. Naturally, the burden of proof and standard of proof interfere in this regard. If the standard of proof is relatively low, it might be easier for one party (here the EC) to discharge the burden of proof.

Yet, when evaluating the lawfulness of the EC decision and the GC’s appraisal of it, it is important to distinguish burden of proof and standard of proof. Both elements, if assessed wrongfully, might amount to errors of law. An allegedly wrongful assessment of the respective standard of proof does, however, not lead to a wrongful attribution of the burden of proof or vice-versa. In the present case, it is rather clear from the GC’s judgment that it correctly attributed the burden of proof to the EC.[69] The fact that the GC regarded it as sufficient for the EC to engage in an analysis of different indicia and context of a potential infringement without necessarily requiring the EC to engage in a full counterfactual analysis and application of an AEC test, is a matter of standard of proof. It relates to the relevant level of conviction which is required to assume a potential violation of Article 102 TFEU. The conviction of an authority or a court can be reached through different means of analysis. A counterfactual analysis involving econometrical models or similar means could clearly be one option. Proof through indicia is simply another option to reach a sufficient level of conviction. Generally, the proof of a fact through indicia is well known in civil procedural law and sometimes also referred to as ‘indirect proof’.[70] Whereas proving a certain fact directly, e.g., through documentary evidence or witnesses, might be superior to indirect evidence, very often in complex litigation, it will be necessary to rely on indirect evidence. Direct evidence is often not available or not sufficient to prove all necessary elements.

Especially when the matter to be proven concerns a hypothetical scenario, e.g., the relevant counterfactual for causation in a case like this one, but also elements of prognosis in merger control, indicia is particularly important. Besides, the counterfactual analysis which Google wanted the EC to conduct does not constitute direct evidence either. An analysis of the counterfactual with the help of econometrical modelling is equally based on certain assumptions which are usually based on indicia itself. Therefore, it is not necessarily less valuable or less convincing to infer a potential restriction of competition from indicia than to infer it from econometrical modelling. The Court, when holding different types of indicia admissible,[71] does not establish a priority of either of the different methodologies over the other. It ‘merely found that it is permissible for the Commission to rely on a range of evidence, without being required systematically to use any single tool to prove the existence of such a causal link’.[72]

From the perspective of the law of evidence, both constitute indirect evidence and must be assessed against their cogency and plausibility. In the concrete case, the EC was able to plausibly demonstrate that Google’s practices led to a decrease in generic search traffic of competing comparison-shopping services, basing its findings on observations of the market developments,[73] even without having needed to apply further counterfactual analysis. One likely explanation is that the effects were so strongly visible in the given case that simple observation of the market developments sufficed already to provide sufficiently convincing indicia for a potential restriction of competition.[74] It is very clear from the context and wording of the Court’s judgment that the EC could have alternatively also applied a full counterfactual analysis. It is simply not obliged to do so in case sufficiently cogent evidence can be provided otherwise.[75]

4.2.  Counterfactual analysis can be used as counterevidence by the claimant

Consequently, in the view of the ECJ, a counterfactual analysis as a means of evidence can equally be used as evidence in favour of the claimant.[76] However, in the present case, this would serve as counterevidence aimed at undermining the original evidence presented by the EC, without shifting the burden of proof, which remains on the EC that initially discharged it through sufficient circumstantial evidence.[77] It simply refers to the possibility that discharging the burden of proof through the provision of circumstantial evidence by the EC can be challenged by the claimant attacking either the basis of the evidence which the EC provided or by providing contrary evidence themselves. This approach seems reasonable and coherent with common standards of evidence procedure. Furthermore, it is in line with the ECJs view in the Intel Renvoi appeal.[78] Just like in Google Shopping, the Court underlined the burden of proof on the Commission.[79] However, it found that the standard of proof and methodology to be applied by the Commission can be altered by substantiated submissions of the undertaking based on supporting evidence.[80] Such substantiated submission could be seen in an AEC-analysis in the context of Intel or in a counterfactual analysis in the case of Google Shopping.

4.3. Relevant criteria for the counterfactual analysis

However, that counterevidence must also be sufficiently cogent either. In particular, a counterfactual analysis must exclude all relevant behaviour which is part of the alleged anticompetitive practice. Otherwise, the counterfactual analysis risks to lose its validity and significance. Studies, as provided by Google,[81] showing no effects in a counterfactual where only the ‘boxes’ were removed but the algorithms remained unchanged do not constitute meaningful counterfactual analyses as these do not cover the nature of the alleged anticompetitive conduct as described above.[82]

That requirement is not contradicting the finding of the GC and ECJ that ‘none of the practices at issue, taken separately, had given rise to any competition objections’.[83] This can, again, be understood when looking at the issue from the perspective of the law of evidence. A simplistic counterfactual analysis, only involving one of the elements of the alleged anticompetitive practice would simply not be sufficiently convincing regarding the absence of effects in the scenario of both practices applied together. This seems reasonable as a counterfactual analysis is naturally facing a risk of imprecision and can only be regarded as an estimate of how markets would have hypothetically had developed. If the potential anticompetitive effects are based on the reasonable assumption that the two elements of the anticompetitive practice (promoting its own boxes and demoting competitors’ results) are inextricably linked, a counterfactual analysis of only parts of that practice is too limited to mirror the hypothetical development of the market. It is, therefore, despite being admissible, no cogent evidence.[84]

5.   The never-ending tale of the as-efficient-competitor-test

Ever since the beginning of the Intel-saga,[85] much discussion has taken place as to whether the EC could be obliged to always examine the requirements of the AEC test whenever establishing an abuse of dominance. In contrast, considerable resources and time of the EC could be saved if the EC would only be obliged to apply the AEC test in order to prove abusive behaviour on the basis of foreclosure effects if – and only if – the undertaking elaborated on the absence of such effects during the administrative procedure based on supporting evidence.[86] It is regrettable though that the ECJ did not use the opportunity to further clarify the standard of such arguments of the parties. It is – for now – at the discretion of the EC to evaluate if the presented evidence is sufficient for the EC to be obliged to apply the AEC test. This decision is then subject to judicial review the intensity of which is also still an open question as illustrated by the Intel-saga.[87]

5.1.  Scope of the as-efficient-competitor-test

Much of the discussion was born from the fact that Article 102 TFEU and its unwritten categories of abuse require the EC to establish that the unilateral conduct in question is abusive and does not constitute competition on the merits. The purpose of this being that the application of Article 102 TFEU is not indicated in cases where a weaker, less competitive company seeks shelter under Article 102 TFEU (the AEC principle). This is in line with the pre-Intel jurisprudence stating that ‘the objective of Article 102 TFEU is not to protect less-efficient undertakings’.[88]

The first choice one needs to make is whether – for the conduct in question – a (potential) per se restriction of competition is indicated or if based on an effects analysis the EC has to prove harmful effects to competition on the relevant market. As indicated above,[89] discriminatory leveraging abuses are not per se harmful, but the EC has to consider all relevant circumstances of the conduct and take an effects-based approach.

Second, to make a distinction between harmful and non-harmful conduct under an effects-based approach, the AEC test, in contrast to the AEC principle, is one valid instrument because it requires the EC to prove that the undertaking produces different effects on the competition on the relevant market than an undertaking not acting wrongfully. Such additional effects lay, according to the ECJ, in the foreclosure effects proven by actual or potential economic data to show ‘a strategy aiming to exclude competitors that are at least as efficient’.[90] Such a strategy goes beyond the effects every efficient competitor would cause on the market. This, however, requires a considerable amount of time and resources spent on behalf of the EC. The ECJ is therefore right in assuming that the undertaking itself has to indicate such circumstances, pointing at the fact that foreclosure effects on the market would have occurred in any case. The Commission has to prove a ‘strategy’ behind the market behaviour aiming to create such effects.[91] If no such circumstances are indicated, the EC is not obliged to acquire data to test its assumption of foreclosure effects.[92]

The ECJ thus dismisses an argument put forward by the parties that the GC had confused the AEC test and the more general requirement to establish an abuse.[93] The ECJ ruled that, while it is true that the unilateral behaviour derogates from competition on the merits giving grounds to analyse the extent of the undertaking’s dominant position, the EC is under no obligation to prove a ‘strategy aiming to exclude competitors that are at least as efficient as the dominant undertaking from the market’[94] if not indicated. Should accused undertakings want to initiate the ECs obligation to apply the AEC test, the undertakings need to state specifics, not just generalised allegations, the ECJ held.[95] This is convincing since the undertaking is better informed about its own market behaviour than the EC and should be able to point the finger at where the EC needs to look for proof that the application of Article 102 TFEU could be misplaced.

5.2.  Wave goodbye to the more economic approach?

This recent judgment is thus in line with a silent withdrawal of the more economic approach that was once favoured by the European Commission[96] but lead to a series of pushbacks by the Courts.[97] However, the recent case-law is arguably indecisive and recent legislative acts do not seem to clearly put efficiency first. For one, the DMA does not include any efficiency defence[98] even though proclaiming ‘contestable and fair markets’ (Article 1 DMA).[99] For another, some recent case-law arguably seems to again confirm a more economic approach,[100] the Google Shopping Case, however, points the finger in another direction: By restricting the applicability of the AEC test to cases where either the EC or the parties invoke it, the ECJ clearly refrains from declaring a mandatory application of the AEC test in Article 102 TFEU cases. Such mandatory application of an AEC standard would have fostered a more ‘economic’ application of competition law provisions to establish an abuse of market power. Also, the new draft exclusionary guidelines from the EC shy away from such an approach.[101] Overall, one could argue that the requirements of a judicial review proven application of a more economic-based or effects-based analysis in a broader sense seemed simply too high to be workable.

In light of such considerations, it seems favourable to lift the burden for the EC by declaring the AEC test to be only compulsory if either the EC or the undertaking is willing to apply it. For now, we can safely say that there is no AEC standard, e.g., a standard always applicable when establishing abuse under Article 102 TFEU (more economic approach plus), necessarily required to establish an abuse under Article 102 TFEU.[102] The AEC test remains a tool of choice by either the EC or the parties to the infringement. The negative impact of market behaviour can be proven by anticompetitive effects, allowing the EC an analysis of all the circumstances of the case. The Commission’s discretion is though limited. Its methodological choice how to prove anticompetitive effects must be ‘based on a concrete logical pattern, defined and made explicit’ by the Commission.[103] Should the Commission choose the AEC-test as a method to demonstrate anticompetitive effects, the ‘AEC test must be able usefully to support the conclusion’ of such effects.[104] The ability of the AEC test to deliver such results is subject to judicial review as can be seen in the Intel Renvoi proceedings.[105]

6.  Conclusion

From a purely procedural standpoint, the ECJ showed very sound and balanced reasoning in Google Shopping. It emphasised the burden of proof on the EC but applied a manageable standard of proof for the EC which allowed it to discharge the burden of proof in different ways, including with circumstantial, indirect evidence. This does not preclude that the undertakings concerned provide – without being obliged to – a more detailed counterfactual analysis but in a procedural situation such as the given, this would only constitute possible counterevidence.

From a substantive side, the ECJ judgment has legitimised a certain form of self-preferencing as abusive: discriminatory leveraging in the form of self-favouritism combined with competitor demotion that create (potential) foreclosure effects, depending on the relevant circumstances of the case. The latter were quite severe in Google Shopping, relying on the substantial market power, gateway status, and market circumstances. Other self-preferencing practices indubitably must reach a certain comparable magnitude in order to meet this standard. Yet, the breaking point has not yet been fully determined.

Regarding remedies, the Google Shopping case certainly highlights the need for more effective and timely solutions. The procedural delay and the fact that the market dynamics shifted in favour of Google over time underscore the limitations of ex-post enforcement in rapidly evolving digital markets. Future cases, such as Google’s complaint against Microsoft regarding cloud services,[106] may offer insights into the evolving nature of self-preferencing and the adequacy of current remedies.

Generally, the question arises as to what role the self-preferencing theory of harm under Article 102 TFEU will play for digital companies in the context of growing EU and national digital regulation. In reflecting on the implications of self-preferencing within the framework of the DMA or the German § 19a German Competition Act, it is important to recognize the broader regulatory trends.

The inclusion of self-preferencing in Article 6(5) DMA suggests a more targeted approach to gatekeeper practices, to some degree distinct from the Google Shopping interpretation of Article 102 TFEU. It focuses on specific self-preferencing strategies of gatekeepers, namely self-preferencing in ranking and does not require an effects-based analysis. It is, thus, limited not only in the personal sense to gatekeepers but also in terms of self-preferencing practices. Yet, while Article 6(5) DMA highlights only a more favourable treatment, mentioning also ‘related indexing and crawling’ pinpoints to an accompanying demotion similar to Google Shopping. Further, albeit the DMA’s disconnect to the effects-based approach, the personal scope of gatekeepers suggests at least a connection to the mentioned necessary specific circumstances highlighting the crucial gateway status in the present case. The fact that gatekeepers under the DMA provide core platform services constituting important gateway(s) per Article 3(1)(b) DMA, is at least loosely comparable to the three specific circumstances surrounding the gateway status highlighted by the ECJ in the present case.[107] Still, the per se approach of the DMA prohibitions entails a more straightway prohibition and likely a much swifter procedural handling in the future.

At the same time, national regulation, such as § 19a German Competition Act, takes a much broader stance than the DMA and covers self-preferencing of undertakings of paramount significance for competition across markets. This could invite more expansive enforcement strategies.[108] This divergence raises questions about the balance between sector-specific regulations like the DMA and more general competition-like laws at national levels, such as § 19a German Competition Act.

Additionally, the revision of Regulation 1/2003 and the possibility of stricter national enforcement, such as through economic dependency or essential facility rules, presents an interesting dynamic in that context. A recent study from the German Competition Authority shows pertinent interest in such national endeavours.[109] On the one hand, these national measures tend to fill a gap, since they go beyond Article 102 TFEU and are often rooted in a long-standing tradition in the Member States.[110] On the other hand, these rules are capable of further complicating the regulatory landscape for dominant firms, as feared by the EC.[111]

Generally, the implications for Article 102 TFEU and the ongoing draft guidelines revision, also following the newest Intel Renvoi judgment,[112] indicate a recalibration of the overall abuse of dominance framework. Although presumptions of abuse remain relevant,[113] the pushback against a purely economic approach signals a partial return to more formalistic competition law principles. The EC's flexibility in proving abuse, even with indirect evidence, suggests that dominance cases might continue to evolve, but the fundamental standards for Article 102 TFEU remain intact. While also Intel Renvoi shows us that the AEC test is not abandoned,[114] Google Shopping shows us the AEC test’s limited scope as a voluntary instrument of either the EC or the undertakings. The intensity of judicial review of the Commission’s application of the AEC test and the standard of proof the parties have to comply with during the administrative procedure though remain unresolved.

While the public enforcement stage of Google Shopping might be over, private enforcement is not. Private actions have already been brought on national level, most of them having been on hold, waiting for this final ECJ decision.[115] Famously, Heureka’s damages action brought against Google in the Czech Republic already made it to the ECJ before the actual EC originated Google Shopping case did.[116] The preliminary reference judgment from April 2024 concerned the statute of limitation for Google Shopping related damages actions. This is a topic of highest relevance for practice, given the overall long duration of the proceedings on EU level before we now have a final decision, binding on national courts according to Article 16 Regulation 1/2003.[117] In Heureka, the ECJ clarified that the limitation period for these types of claims can only commence once the breach of competition law has occurred and the injured party is aware of the necessary information to initiate a legal action.[118] Additionally, the ECJ determined that the limitation period must be paused during a Commission investigation and any follow-up court proceedings,[119] leaving ample room for private follow on actions now. Still, the potential for private actors to seek redress post-judgment is uncertain due to practical challenges of proving harm and causality in self-preferencing cases.[120]

Finally, while the ECJ’s substantive reasoning and procedural handling of Google Shopping appear sound, it also illuminates a structural tension in the overall enforcement system of EU competition law. The EC's victory may have come at the cost of third-party rights, as the complainant faced procedural expenses while market dynamics evolved unfavourably for them over the long duration. This case signals a cautious retraction from the economically heavy-handed approach of the last two decades, hinting at a judicial preference for more traditional competition law principles, even if that benefits the regulator over market participants. Thus, while the Google Shopping judgment might be viewed as a success for the EC, its long-term impact on market fairness, the effectiveness of competition law and on third-party rights remains contentious.

-------------------
European Papers, Vol. 10, 2025, No 1, pp. 25-44
ISSN 2499-8249 - doi: 10.15166/2499-8249/823

* Academic assistant and PhD candidate, LMU Munich, eva.fischer@jura.uni-muenchen.de.
** Tenure track professor for European law, University of Vienna, and postdoctoral research-er (Habilitandin), Heidelberg University, lena.hornkohl@univie.ac.at (corresponding author).
*** PhD candidate, University of Göttingen, nils.imgarten@jura.uni-goettingen.de.

In accordance with the declaration of ethics of the Academic Society for Competition Law (ASCOLA), we have nothing to disclose. The paper was finished on 31 October 2024, later de-velopments had to be disregarded.