CMA Fair Ranking Conduct Requirement: OpenAttribution Response

To: Competition and Markets Authority Re: Google Search Conduct Requirements Consultation, Fair Ranking CR From: Alex Springer, OpenAttribution Date: February 2026

1. Introduction

Targeted response to the Fair Ranking CR, focused on AI features and the interaction with the Publisher CR. Separate response submitted for the Publisher CR. We do not address questions 6.3 or 6.7.

2. Scope Must Be Defined by Function, Not Product Label (Q6.2)

The inclusion of AI Overviews and AI Mode in paragraph 2(b) is essential. But the CR treats them as equivalent. They are not.

AI Overviews augment a search results page. They appear on the SERP alongside organic listings. AI Mode is a conversational interface generating multi-turn responses from web content, closer to Gemini than to a results page.

The SMS decision excluded Gemini from general search services (SMS Decision, paras 4.85-4.86). If AI Mode continues evolving toward a conversational product (and Google has commercial incentives to do so), the CMA faces a classification problem: AI Mode looks like search today but could look like Gemini tomorrow.

The interpretative notes should define scope by function: if a feature uses search-crawled content to generate responses to user queries, it is in scope regardless of branding. Without this, Google can move content-dependent activity outside the CR’s reach by restructuring features.

There is also a potential gap between the two CRs. The Fair Ranking CR covers ranking within and relative to AI features on the SERP. It does not explicitly cover how content is selected inside an AI-generated response. If an AI Overview draws on five sources but cites two, that selection has commercial consequences for publishers. The CMA may consider this inherently a ranking decision, or covered by the Publisher CR’s attribution requirements. We would encourage explicit clarification on which instrument governs citation selection.

3. Non-Discrimination and the Publisher CR (Q6.4)

Paragraph 4(a)(iii), prohibiting downranking of publishers who opt out of AI features, is the single most important provision for making the Publisher CR’s opt-out controls work.

The Fair Ranking CR already addresses organic results’ positioning relative to AI features on the SERP (paragraph 2(c), interpretative note 1(c)). The specific case of opt-out publishers warrants explicit attention.

Google is unlikely to implement an explicit ranking penalty for opt-outs. But as AI features take more above-the-fold SERP space, publishers who opt out lose visibility by not appearing in the dominant feature. No ranking signal is needed. The presentation does the work. If a publisher opts out of AI Overviews and appears only in organic results pushed below the fold, the practical outcome is a ranking penalty. Paragraph 4(a)(iii) should cover these presentation effects.

Separately, paragraph 4(a)(ii) should explicitly cover AI content licensing deals. As publishers begin negotiating paid agreements with Google for AI training data, a category of commercial arrangement that did not exist when Google’s Honest Results Policy was drafted, there is a foreseeable risk that licensed content receives preferential treatment in AI responses.

4. Material Changes Must Include AI Feature Changes (Q6.5)

The concept of “material change” (interpretative note 9) should explicitly include changes to how AI features select, weight, and present content. Changes to source citation logic, attribution prominence, or SERP space allocation are all material to publishers. They may not fit neatly into “ranking algorithm change” but the effect is identical.

AI features change differently from traditional ranking algorithms. Model updates, grounding logic adjustments, and prompt changes can be individually minor but cumulatively significant. The interpretative guidance should address how cumulative incremental changes are assessed against the materiality threshold.

5. AI Feature Complaints Should Be Separately Tracked (Q6.6)

The quarterly summaries to the CMA (paragraph 7) should break out AI feature complaints as a separate category. If AI Overviews are systematically reducing traffic to specific content verticals, that pattern needs to be visible in the data, not buried in aggregate ranking complaints.

Smaller publishers are reluctant to complain formally against a platform they depend on for traffic. The CMA should consider supplementing the complaints process with its own monitoring of AI feature impacts, drawing on the transparency data required by the Publisher CR.

6. Summary

Define scope by function, not product label. AI Mode is converging with Gemini. The CR must capture content-dependent AI features regardless of branding.
Clarify which CR governs citation selection inside AI responses. Neither instrument explicitly covers source selection within generated text.
Extend opt-out protection to presentation effects. Displacement below the fold by AI features is functionally a ranking penalty.
Cover AI licensing deals in the non-discrimination provision. A new category of commercial arrangement that should not influence ranking or citation.
Define “material change” to cover AI feature changes. Model updates, grounding logic, and SERP real estate allocation should be in scope.
Track AI feature complaints separately in quarterly summaries to the CMA.

Contact: Alex Springer, OpenAttribution

GitHub: https://github.com/openattribution-org
Email: alex@openattribution.org