Anthropic Claude Skills Marketplace: A New AEO Surface for B2B SaaS

On March 18, 2026, the European Commission opened a formal market investigation under the Digital Markets Act to determine whether the major generative AI assistants — ChatGPT, Gemini, Microsoft Copilot — meet the gatekeeper thresholds that would subject them to the regulation's most stringent obligations. Six weeks later, the US Department of Justice filed an amended complaint in the long-running Google search monopoly case that added specific allegations about AI Overviews self-preferencing and Gemini default placement. Three weeks after that, the UK Competition and Markets Authority designated OpenAI, Anthropic, and Google DeepMind with strategic market status under the new Digital Markets, Competition and Consumers regime.

These are not isolated proceedings. They are coordinated parts of a regulatory wave that, by mid-2027, will fundamentally restructure how AI assistants choose citations, how AEO operators measure citation share, and how the distribution surface that has emerged over the last three years actually functions. The eighteen-month window between now and the effective date of the first conduct requirements is the most important period in the history of AI search regulation, and operators who treat it as a compliance side-quest rather than a primary input to content strategy will be playing defense for the rest of the decade.

This piece walks through the four parallel regulatory tracks now in motion — EU DMA, US DOJ, UK CMA, and the FTC's AI partnership inquiry — examines what is likely to change at the citation surface, and provides an operator-focused playbook for AEO programs that need to be defensible under a 2027 regulatory regime that does not yet exist but is coming into focus fast.

The Four Regulatory Tracks Now Converging

The European Commission, the US Department of Justice, the UK Competition and Markets Authority, and the US Federal Trade Commission are pursuing four formally separate but substantively overlapping inquiries that share three goals: citation transparency, default-recommendation neutrality, and competitive access to user attention. The proceedings, the alleged conduct, and the remedy paths are different in each jurisdiction, but the trajectory is the same.

Each of these tracks is independently consequential. Together, they represent the most aggressive coordinated regulatory action against a tech-industry category since the late-1990s Microsoft case. AEO operators reading this section should not treat the proceedings as adjacent to their work. The remedies that emerge — transparency obligations, neutrality requirements, citation auditability — will define the surface area that AEO competes for over the next five years.

The proceedings also matter because the operators they target are the same operators whose answer products determine where AEO citation share is won and lost. If ChatGPT becomes subject to DMA conduct rules in mid-2027, every AEO program optimizing for ChatGPT citation share will operate under a different set of incentives than they do today. If Google must publish AI Overviews citation source distributions on a quarterly basis under a DOJ consent decree, the measurement infrastructure that AEO teams rely on changes overnight. If the CMA's draft conduct requirements take effect in the UK, the rest of the world is likely to see equivalent rules within twenty-four months.

EU DMA: From Gatekeeper Designation to AI Conduct Rules

The Digital Markets Act, in force since 2022, was designed to address the structural market features that emerged from the first wave of platform consolidation — app stores, search engines, social networks, online marketplaces, browsers. The framework was deliberately drafted to be extensible. In March 2026, the European Commission used that extensibility to open the most consequential market investigation in the DMA's short history.

The investigation has two parts. The first asks whether generative AI assistants, considered as a standalone core platform service, meet the gatekeeper thresholds when integrated into existing gatekeeper products. The second asks whether the integration of AI assistants into search engines, browsers, and operating systems constitutes a separate conduct issue under the existing gatekeeper obligations of Alphabet, Apple, Meta, Microsoft, and Amazon.

Both parts are expected to conclude with designation by the end of 2026. The Commission has signaled, through both formal communications and informal statements at sector conferences, that the question is not whether AI assistants will be designated but what conduct obligations will apply. The current best-guess set, based on published guidance documents and industry consultation responses, includes the following.

Self-preferencing prohibition. Designated AI assistants will be prohibited from preferring their own services or those of related entities in AI-generated answers when responding to queries that have a competitive answer landscape. The clearest operational example is Google's AI Overviews citing Google Shopping, Google Maps, and YouTube at rates that materially exceed the citation share those properties would receive under neutral source selection. The CMA's preliminary analysis of equivalent UK queries, published in its phase two report, found that Google properties were cited 2.4x more often in AI Overviews answers than would be predicted by relevance models trained on independent ranking signals.

Data portability for business users. Designated gatekeepers will be required to provide content creators with the data their content generated during the AI answer construction process — which queries produced citations, what excerpts were used, how often the citation appeared, and where in the answer the citation was placed. This is the single most operationally important provision for AEO operators, because it converts citation measurement from a third-party scraping exercise into an audited regulatory disclosure.

Interoperability for AI assistants. Designated gatekeepers will be required to permit users to select non-default AI assistants within their integrated products. The clearest example is Chrome users being permitted to set Claude or Perplexity as their default browser assistant, instead of Gemini, without functional degradation. The expected effect on citation share is to reduce the default-driven advantage that integrated AI products currently enjoy and to increase competitive pressure on citation neutrality.

Dispute resolution mechanism. Designated gatekeepers will be required to operate a structured dispute resolution mechanism through which business users can challenge citation or ranking outcomes. The mechanism must produce reasoned decisions within statutory timelines and is subject to oversight by the Commission's dedicated DMA enforcement team.

The Commission has stated that conduct requirements will be finalized in the second half of 2026 and will take effect in mid-2027. The eighteen-month window is short by any measure of corporate policy adjustment but long by the measure of AEO program adaptation. Operators who begin preparation now will compound an advantage through the entire implementation period.

For broader context on how AI distribution dynamics are evolving toward a more concentrated future, see our analysis of AI search 2030: a five-prediction distribution forecast.

US DOJ: Why the Google AI Investigation Matters More Than the Original Search Case

The Department of Justice's amended complaint, filed in the Eastern District of Virginia in February 2026, added three substantive allegations to the existing search monopoly case. The amended complaint alleges that Google leveraged its search distribution agreements with Apple and Android OEMs to entrench Gemini as the default AI assistant on a majority of mobile devices in the United States. It alleges that AI Overviews self-preference Google properties at rates inconsistent with neutral source selection. And it alleges that Google's data agreements with publishers — under which publishers receive limited compensation for use of their content in AI Overviews — exploit Google's dominant position in search distribution to extract terms publishers would not accept in a competitive market.

The remedy phase of the case, expected to begin in earnest in late 2026 after the Court rules on Google's motion to dismiss the amended complaint, is the single most consequential AI search proceeding in the United States. The remedies that DOJ has signaled it will seek include the following.

Mandatory citation source distribution disclosure. Google would be required to publish quarterly reports of the citation source distribution in AI Overviews answers, broken down by query category. The reports would be subject to audit by an independent monitor appointed by the Court.

Permission for third-party citation neutrality measurement. Google would be required to permit accredited third-party measurement firms to query AI Overviews at scale for the purpose of measuring citation distribution and to provide a public API for that purpose. The measurement firms would publish independent quarterly assessments of citation neutrality.

Default assistant choice screens. Google would be required to present Android users with a non-preferential choice screen for AI assistant selection on first device setup, similar to the browser choice screen mandated under the original 2009 European Commission consent decree against Microsoft.

Publisher content compensation framework. Google would be required to negotiate compensation frameworks with publishers whose content is materially used in AI Overviews answers, with a backstop arbitration mechanism for cases where negotiation fails. The Australian News Media Bargaining Code is the apparent template.

The probability that all four remedies are adopted is low. The probability that at least two are adopted is high, based on the patterns of recent DOJ consent decrees and the apparent preferences of the trial judge. The implications for AEO operators in the United States are direct: by 2028, the AI Overviews citation surface will be more transparent, more neutral, and more contested than it is today. Operators whose citation share is currently propped up by self-preferencing within the Google ecosystem will see that advantage erode. Operators whose citation share is earned on substantive citation quality grounds will benefit from the more neutral citation environment.

Reuters and the Wall Street Journal have both reported extensively on the DOJ proceeding; the Reuters tracker on the Google antitrust case is the most useful single source for following the docket. For operators thinking about how content compensation might evolve under any of these remedies, the broader trend toward monetized content access is examined in the crawler permission economy: how publishers will monetize training data in 2026.

UK CMA: Likely First to Enforce Citation Transparency

The Competition and Markets Authority has moved fastest of the three major Western jurisdictions on AI conduct regulation, in part because the UK's Digital Markets, Competition and Consumers Act 2024 gives the CMA broader and more flexible powers than the DOJ or the European Commission. The Act, in force since January 2025, created the strategic market status (SMS) designation framework and authorized the CMA to impose tailored conduct requirements on designated firms without the lengthy market investigation process that the DMA requires.

In April 2026, the CMA published the phase two findings of its foundation models market investigation. The report concluded that the foundation model and AI assistant markets exhibit four structural features that warrant ongoing intervention: high switching costs for downstream users, default-driven distribution that entrenches incumbent assistants, vertical integration between model developers and cloud providers that forecloses competitor access, and opaque citation logic that prevents downstream content creators from understanding how their content is selected.

In May 2026, the CMA designated three firms with strategic market status: OpenAI for ChatGPT, Anthropic for Claude, and Google DeepMind for Gemini. Microsoft Copilot was not designated as a standalone product because the CMA concluded that Microsoft's AI assistant capability is functionally a distribution of OpenAI's models; the CMA's view is that conduct requirements imposed on OpenAI will reach Copilot through the underlying model provider relationship.

The CMA has stated that draft conduct requirements will be published in Q3 2026 with final requirements taking effect in early 2027. The draft is expected to include the following.

Quarterly citation source distribution disclosure. Designated firms will be required to publish quarterly reports of the citation source distribution across their AI assistant answers, broken down by query category and jurisdiction. The reports will be subject to audit by the CMA.

Per-answer citation rationale. Designated firms will be required to maintain a per-answer log of which sources were considered, which were cited, and why — and to provide that log to the CMA's dispute resolution authority on request.

Citation policy disclosure. Designated firms will be required to publish a citation policy explaining the principles governing source selection, the relative weight of different signal types, and the conditions under which the firm will downgrade or exclude particular source categories.

Equal access to citation surface. Designated firms will be required to ensure that citation eligibility is not conditioned on commercial relationships with the firm or its affiliates, except where a transparent and non-discriminatory commercial offering is available to all market participants.

The CMA's conduct requirements are likely to be the strictest of the three jurisdictions at the point of taking effect, in part because the UK regulatory framework is the most recently drafted and is therefore the most calibrated to the specific conduct issues that the AI assistant market has presented. The Financial Times has reported, citing CMA officials, that the agency intends the UK regime to set the international benchmark for AI conduct regulation and expects the EU and US to converge on similar requirements over a twenty-four-month period.

For AEO operators, the UK CMA framework matters even if their primary market is the US or the EU, because the disclosure obligations imposed on designated firms will produce data that is globally observable. Citation source distribution reports published by OpenAI under UK conduct requirements will reveal the global citation behavior of ChatGPT, not just the UK-specific behavior. The disclosure obligation will functionally serve as a global transparency mechanism even if the formal regulatory authority is jurisdictional.

FTC AI Partnership Inquiry: The Slowest Track With the Largest Surprise Potential

The Federal Trade Commission's Section 6(b) inquiry into AI partnerships, opened in January 2024 and continuing through 2026, is the slowest-moving of the four major regulatory tracks but has the largest range of potential outcomes. The inquiry is examining the Microsoft-OpenAI relationship, the Google-Anthropic relationship, and the Amazon-Anthropic relationship, with particular focus on whether the cloud compute provision arrangements and equity investments constitute anticompetitive conduct under existing antitrust law.

The inquiry has not yet produced a public report. The FTC's commissioners have given conflicting public signals about the likely conclusions, and the change in administration in January 2025 has affected the agency's enforcement priorities in ways that are still being clarified. The most likely outcomes, in order of decreasing probability based on patterns of recent FTC inquiries:

A report identifying competitive concerns without immediate enforcement action, paired with a recommendation that Congress consider legislative remedies. This outcome would be the most likely under the current commissioner composition and would have limited near-term impact on AEO operators.

A consent order with one or more of the partner firms requiring structural separation of cloud compute provisioning from model investment, with carve-outs for existing commercial relationships. This outcome would have significant industry impact and would likely accelerate the entry of additional foundation model providers into the assistant market, increasing citation diversity over time.

A full enforcement action seeking divestiture or dissolution of the partnerships. This outcome is unlikely under the current commissioner composition but is the explicit preference of the agency staff that has been driving the inquiry. If pursued, it would have transformative effects on the AI assistant market structure and would create a multi-year period of uncertainty about which firms would be operating which assistants.

The FTC inquiry intersects with the other three regulatory tracks in important ways. The CMA's SMS designations of OpenAI and Anthropic implicitly assume that those firms are structurally independent from their cloud compute partners; if the FTC determines otherwise, the UK conduct requirements would need to be redesigned. The DOJ's Google case is, in part, a response to Google's incomplete vertical integration with Anthropic — Google is both a major Anthropic investor and a competitor through DeepMind. The DMA's AI assistant designations will be substantially affected by the structural conclusions the FTC reaches about the AI partnership market.

AEO operators should track the FTC inquiry because the surprise outcome — a serious enforcement action — would reshape the assistant landscape more dramatically than any of the other three regulatory tracks combined.

EU AI Act Category Extensions: The Slow-Building Risk

The EU AI Act, in force since 2024 and progressively phasing in through 2027, was originally drafted to address general-purpose AI capabilities through a tiered risk framework. The category extensions consultation, opened by the European Commission in Q2 2026, is examining whether to add specific categories addressing AI search and AI assistant functions to the high-risk classification.

If adopted, the extensions would impose three additional obligations on AI assistant providers operating in the EU.

Content provenance disclosure. AI assistants would be required to disclose, for each generated answer, the categories of sources from which the answer was constructed and to provide users with the ability to drill down to specific citations. The disclosure obligation would be technical and standardized, allowing third-party tooling to verify compliance.

Synthetic content labeling. AI-generated content would be required to be labeled as such in machine-readable form, with a chain of provenance back to the originating system. The labeling requirement would extend to text, images, and audio generated by AI assistants.

Bias and fairness assessment. AI assistant providers would be required to conduct annual bias and fairness assessments of their citation behavior and publish summary results, with full results made available to the AI Office and national competent authorities.

The consultation period closes in Q3 2026 with a likely Commission proposal in late 2026 and adoption by the Council and Parliament expected in 2027. The earliest effective date for new obligations is Q1 2028.

The AI Act category extensions are the slowest-building of the regulatory risks but also the most far-reaching, because they apply uniformly to all general-purpose AI providers, not just designated gatekeepers. The DMA conduct rules will apply to a handful of firms; the AI Act category extensions will apply to dozens.

What Mandated Citation Transparency Actually Changes

If the regulatory trajectory plays out as described above, the AEO operating environment in 2028 will differ from the current environment in four specific ways.

Citation data becomes a regulated disclosure. The CMA quarterly reports, the DMA business user data portability provisions, and the DOJ remedy disclosures will all produce auditable citation distribution data that operators can access without third-party scraping. The citation measurement firms — Profound, SerpRecon, Bluefish — will pivot from synthetic measurement to analysis layered on top of regulated disclosures. Internal AEO measurement programs will be able to verify their citation share against authoritative source data for the first time.

Self-preferencing erodes as a citation share advantage. The current Google-owned property citation premium in AI Overviews will compress. Microsoft properties cited in Copilot will face similar compression in the EU and UK. Independent operators competing on substantive citation quality will gain measurable citation share in categories where self-preferencing currently distorts the distribution.

Default-driven distribution loses some leverage. The DMA interoperability provisions, the DOJ default-assistant remedy if adopted, and the CMA conduct requirements will collectively reduce the citation share advantage that integrated AI assistants enjoy through default placement. Citation share will be redistributed toward assistants that users actively choose, which over time tends to favor assistants that produce more useful answers, which over time tends to favor assistants that draw on more diverse citation sources.

Dispute resolution becomes an operational function. AEO operators in regulated markets will gain access to formal dispute resolution mechanisms for citation outcomes they believe to be incorrect or biased. The mechanisms will impose real cost on assistant operators when they make errors, which will create downstream pressure to improve citation accuracy.

The combined effect is to convert the AI search citation surface from an opaque, self-preferencing-prone, default-dominated environment into a more transparent, more neutral, more contested environment over a three-year period. AEO operators who treat the transition seriously will compound an advantage. Operators who continue to optimize against the current opaque environment will find their measurement frameworks, content strategies, and competitive positioning increasingly misaligned with where the surface actually moves.

The Eight-Step Playbook for Regulatory-Defensible AEO

The following playbook is for AEO operators who want their programs to be defensible under a 2027 regulatory regime that does not yet exist but is taking shape. The investments are not optional. They will define which operators retain citation share through the regulatory transition and which operators see citation share erode as the surface restructures.

1. Stand up a regulatory monitoring function. Assign one person on the AEO team — or contract one specialist — to monitor the DMA, CMA, DOJ, FTC, and AI Act proceedings on a weekly basis. The pace of rulemaking is accelerating and the divergence between jurisdictions is widening. Operators without dedicated monitoring will discover changes too late to adapt content strategy in time. The function does not need to be large. It needs to be consistent and to feed a quarterly internal briefing that reaches the head of content, the head of SEO, and the head of legal.

2. Audit your citation defensibility. For your top 200 cited pages, conduct a defensibility audit. Are the factual claims primary-sourced and dated? Are the people behind the analysis named? Is the methodology disclosed? Are the data extractable? Pages that pass the defensibility audit are likely to be favored by regulated citation logic. Pages that fail are likely to lose citation share as transparency requirements take effect. Use the audit results to drive a remediation roadmap.

3. Build a structured factual claims layer. Implement a structured data layer that exposes the factual claims in your content in extractable, dated, sourced form. Schema.org, JSON-LD, or a custom layer all work. The point is that when dispute resolution mechanisms emerge, your content will be able to demonstrate the underlying factual integrity that supports each claim. Operators who do this systematically will be able to challenge citation errors effectively when the mechanisms open.

4. Document your source citation policy publicly. Publish a public source citation policy explaining how you select sources, weight evidence, handle conflicts of interest, and update outdated information. The policy serves two purposes. First, it signals to AI assistants that your content is itself produced under a citation discipline equivalent to what the assistants will be required to disclose. Second, it serves as a comparison standard against which you can hold assistant citation policies accountable when they are published.

5. Diversify your jurisdictional footprint. Build separate content variants or, at minimum, separate optimization passes for the EU, UK, and US markets. The regulatory regimes are diverging, the conduct rules will apply differently, and the citation distributions in each jurisdiction will move differently as remedies take effect. Operators who optimize for the global average will lose to operators who optimize jurisdiction-by-jurisdiction. The cost is real but the upside is durable.

6. Engage in formal consultations. The CMA, the European Commission, and to a lesser extent the DOJ all open formal consultations as part of the rulemaking process. AEO operators who engage in those consultations — through industry associations or directly — have a meaningful ability to shape the conduct requirements that take effect. The cost of engagement is a few staff days per quarter. The downside of disengagement is that the rules will be written by parties whose interests are not aligned with yours.

7. Instrument disclosure-ready citation measurement. Build your internal citation measurement infrastructure to be ready to ingest the disclosure data that the regulatory regimes will produce. This means investing in citation tracking platforms whose roadmaps include integration with CMA quarterly reports, DMA business user data, and DOJ disclosure data. Operators with infrastructure ready to absorb authoritative citation data on the day it becomes available will compound an advantage over operators whose measurement is still based entirely on synthetic scraping.

8. Coordinate AEO and legal functions. The regulatory transition will produce operational decisions — what to dispute, what to document, what to disclose, what to contest — that span AEO and legal. Establish a recurring coordination mechanism, with clear ownership, before the conduct requirements take effect. Operators who only set up coordination after enforcement begins will lose the first eighteen months of disputes to operators who have functional processes already in place.

For broader context on how content monetization and revenue models are being restructured under these same dynamics, see the publisher revenue models for zero-click survival playbook.

What Could Derail the Regulatory Trajectory

Three scenarios could materially slow or reshape the regulatory wave described in this piece. AEO operators should track each as a counter-indicator.

Court reversals. The DMA gatekeeper designations have been contested in EU courts and several appeals are ongoing. The CMA's SMS designations are likely to be appealed by the designated firms. The DOJ's amended complaint is subject to the trial judge's discretion. A series of court reversals could materially slow the regulatory trajectory and shift the effective dates of conduct requirements by twelve to twenty-four months.

Administrative changes. Regulatory enforcement is sensitive to administration changes. A change in the leadership of the DOJ Antitrust Division, the FTC commissioners, or the European Commission could shift enforcement priorities in ways that slow or accelerate specific tracks. The 2027 election cycles in both the EU and the US are meaningful inflection points.

Voluntary commitments by the firms. OpenAI, Anthropic, and Google have all signaled, through public statements and through behavior during the CMA proceeding, that they may offer voluntary citation transparency commitments to avoid more burdensome regulated disclosures. The Verge has reported on internal OpenAI policy discussions about pre-emptive transparency releases. Voluntary commitments would shift the timeline forward — operators would gain access to disclosure data sooner — but would also reduce the regulatory leverage that AEO operators can use to challenge specific citation outcomes.

None of these scenarios eliminate the regulatory trajectory. They affect the pace and the specific shape of the conduct requirements that emerge. The strategic posture for AEO operators is the same in each scenario: prepare for a transparent, neutral, contestable citation environment to take effect within a two-to-three-year horizon, and structure content strategy accordingly.

Anomalies Worth Watching

A few specific developments are worth watching over the next six months because they will materially affect how the regulatory wave actually plays out.

The European Commission's response to the formal consultation on AI Act category extensions, expected in Q4 2026, will signal whether the Commission intends to push the high-risk classification aggressively or to defer to the DMA framework for assistant-specific obligations. The two approaches produce different operational outcomes for AEO operators.

The CMA's publication of its draft conduct requirements, expected in Q3 2026, will be the first time a major Western jurisdiction publishes specific text on AI assistant conduct rules. The text will set the benchmark against which the EU and US frameworks are measured. Operators should read the draft carefully and submit consultation responses if any provisions create operational issues.

The trial judge's ruling on Google's motion to dismiss the amended DOJ complaint, expected in late summer 2026, will set the procedural posture for the remedy phase. A favorable ruling for DOJ will accelerate the timeline for remedy negotiations. An unfavorable ruling will narrow the scope of the case and may delay the remedy phase by twelve months.

The FTC's publication of the Section 6(b) inquiry report, expected in late 2026 or early 2027, will provide the clearest signal of US enforcement direction on the AI partnership question. The report may recommend legislative remedies, consent orders, or full enforcement actions. Each path implies different industry dynamics and different citation environment outcomes.

Takeaway: The regulatory wave converging on AI search in 2026 is the most consequential shift in the AEO operating environment since AI assistants entered the market. The DMA, the DOJ Google case, the CMA's SMS designations, the FTC partnership inquiry, and the AI Act category extensions will collectively restructure the citation surface around three principles — transparency, neutrality, and contestability — within an eighteen-to-thirty-month window. AEO operators who treat the transition as a compliance afterthought will find their measurement frameworks, content strategies, and competitive positioning misaligned with the surface that emerges. Operators who build citation defensibility, jurisdictional differentiation, regulatory monitoring, and disclosure-ready measurement into their programs over the next two quarters will compound an advantage that holds through the rest of the decade.