Operator brief

UK CMA publisher conduct requirement: operator brief for marketing, legal, and content

On 3 June 2026 the CMA imposed a Publisher Conduct Requirement on Google under the Digital Markets, Competition and Consumers Act 2024. It gives UK-facing publishers stronger controls, transparency, metrics, and attribution expectations for how Search Content is used in Google generative AI. This brief is an operator SOP: what changed, when obligations bite, and what marketing, legal, and content teams should do. It is not legal advice.

Last updated 13 September 2026

Opening answer

If you publish for people in the UK, treat Google's generative AI use of your Search Content as a governed control problem, not a vibes debate.

The CMA's Publisher Conduct Requirement, the Publisher CR, requires Google to give publishers effective controls over use of their Search Content in generative AI, publish clear explanations of how that content is used, provide engagement metrics for search generative AI features, take reasonable steps on clear attribution with a path back to the source, and not retaliate in classic general search for using those controls.

It applies to generative AI outputs shown to persons located in the UK, natural or legal persons, per the Notice. It is a conduct requirement on Google, not a new statute you file. Your job is to decide opt-outs deliberately, document them, and operationalise the metrics and attribution once Google's compliance surfaces appear.

What changed, 3 June 2026

After designating Google as having strategic market status, SMS, for general search services, with the SMS decision published 10 October 2025, the CMA imposed the Publisher CR on 3 June 2026.

In plain operator language, Google must:

Controls: give publishers effective controls to withhold Search Content from use in generative AI, covering training and grounding of both broader generative AI services outside general search, such as Gemini AI Assistant and the Vertex AI API, and search generative AI features inside general search, such as AI Overviews and AI Mode.
Granularity for Search AI grounding: for grounding of search generative AI features, enable opt-out at directory level and page level.
No circumvention: not try to get around a publisher's withhold choice by acquiring that Search Content through other sources, with interpretative nuance on open-source datasets.
Transparency: publish clear, comprehensible, user-friendly Explanatory Information on how Search Content is used for training and grounding, and on what the controls do.
Metrics: give publishers clear, detailed metrics on user engagement with their Search Content in search generative AI features.
Attribution: take reasonable steps so Search Content in search generative AI features is attributed clearly and accurately, with a clear means for users to access that content, and publish how Google approaches attribution and factuality measurement.
Anti-retaliation: not prevent or penalise publishers for using the controls, for example by downranking opted-out content in general search outside search generative AI features.

Search Content means publisher content collected through Googlebot, or any Google crawler fulfilling the function of crawling open web content for Google's general search.

When it comes into force, and what not to invent

From the Publisher CR Notice, 3 June 2026:

The Publisher Conduct Requirement comes into force six months after the date of the Notice, except
Paragraph 4.b, page-level controls for grounding of search generative AI features, comes into force nine months after the date of the Notice.

Planning aids only, derived by counting six and nine months from the Notice date. These month names are not labelled in the Notice, so confirm against any later CMA update before client deadlines:

Notice date: 3 June 2026
Six months after Notice: about 3 December 2026, covering most obligations, including directory-level grounding controls under Notice paragraph 4.a
Nine months after Notice: about 3 March 2027, Notice paragraph 4.b only, for page-level grounding controls for search generative AI features

In legal or client counsel copy, quote the Notice's "six months / nine months after the date of this Notice" language first.

Press against Notice: the CMA press release's framing of nine months to implement all changes, with important parts earlier, is secondary. Do not overwrite the Notice's split timetable with a single nine-month deadline.

The Notice continues, subject to variation or revocation, through the SMS designation period ending 10 October 2030, subject to Act provisions on extension and revocation.

How the Interpretative Notes change the Google-Extended conversation

The Interpretative Notes, 3 June 2026, are the CMA's reading of what effective controls likely require. They are not a substitute for the Notice, and Google may take a different compliant approach if it can demonstrate compliance to the CMA.

Key operator implications:

Expect changes to the existing Google-Extended control so publishers can opt Search Content out of training of broader generative AI and search generative AI features, and out of grounding of broader generative AI services.
Where a publisher withholds via Google-Extended for training of search generative AI features, Google should ensure that content is not used for that training going forward. Google is not required to remove Search Content already added to such models.
For grounding of search generative AI features, expect directory-level and page-level opt-outs, with page-level on the later nine-month clock.
UK location for scope may reasonably be assumed from device-reported location.
Training, for interpreting the CR, includes fine-tuning.
Metrics expectations include impressions with attribution position, engagement and click-throughs identifiable as from search generative AI features, and click-through rate, disaggregated per publisher and from classic organic search, ideally via a commonly accessible platform such as Search Console.
Anti-retaliation: opting out should not, for the purpose of punishment, cost you snippets or classic general-search presentation relative to opted-in peers. Promoting the benefits of staying in, or negotiating agreements, is not itself retaliation under the Notes' examples.

This is why the Contradiction Decoder warning still matters: Google-Extended is not a Google Search ranking off-switch. Under the CMA package it becomes part of a wider UK control story for training and grounding, while Search AI grounding also needs directory and page controls.

Contradiction Decoder, Google's guide against ChatGPT, Perplexity and Claude →

Operator SOP

Legal and compliance, own the decision

Map whether you are a publisher making content available on the web whose Search Content is collected for Google general search.
Decide, in writing, your posture for UK audiences across four boxes: training of broader generative AI, for example Gemini Apps and Vertex-class services; grounding of broader generative AI; training of search generative AI features; and grounding of search generative AI features, directory against page.
Do not treat a single robots token as the whole answer. Track Google's Explanatory Information when published, and record which control you used, when, and why.
Separate UK CMA and DMCCA Google obligations from non-Google bots such as OpenAI, Perplexity and Anthropic. Those are different instruments.
Escalate disputes on inaccurate attribution using whatever publisher pathways Google publishes, and keep a log.
This brief is general information, not advice for your facts. A named legal approver is required before regulated or client-specific counsel language ships.

Content, SEO and AEO, own the assets

Inventory high-value URLs and directories likely to appear in AI Overviews and AI Mode.
Prepare for directory-level and page-level grounding decisions, with page-level on the later commencement. Decide which money pages stay in for citation and attribution, and which should be withheld.
Keep pages crawlable and snippet-eligible if you want to remain in Search AI features. Opting out is a choice, not a free brand strategy.
When metrics land, impressions, clicks and CTR from search generative AI features, fold them into AEO reporting separately from classic organic. Do not blend them into one vanity AI traffic number without labels.
Watch attribution quality: are you named, linked, and accurately described when your Search Content grounds an answer?

Marketing, comms and growth, own the narrative

Questions before pitches. "Do we want to be in Google's UK search generative AI features for this cluster?" comes before any "we are fully opted out of AI" slogan.
Do not promise clients that Google-Extended alone removes AI Overviews. It is not a Search ranking or inclusion kill switch.
Do not claim the CMA banned AI scraping, or forced paid licensing for every citation. The CR is a control, transparency, metrics, attribution and anti-retaliation package on Google.
If you publish case studies, quote only what the GOV.UK Notice and Notes say, with dates. No invented commencement folklore.
Align brand storytelling with AEO. AEO leads how we talk about answer visibility, and this CR is the UK Google-specific governance layer under DMCCA.

Do and don't card

Do	Don't
Plan around six-month and nine-month commencement from the 3 June 2026 Notice, using derived calendar dates only as a planning aid	Treat derived month names, for example March 2027, as CMA-labelled deadlines
Split training against grounding, and broader AI against Search AI features	Collapse everything into “block AI”
Expect Google-Extended changes plus directory and page grounding controls	Assume Google-Extended turns off AI Overviews ranking
Demand and use disaggregated Search AI engagement metrics when provided	Mix Search AI clicks into undifferentiated organic without labels
Keep classic SEO and AEO quality if you stay opted in	Opt out and still expect the same Search AI citation outcomes
Treat anti-retaliation as a CMA obligation on Google	Assume zero second-order ranking effects in every edge case, since the Notes recognise limits
Keep ChatGPT, Perplexity and Claude robots decisions on a separate track	Copy CMA Google controls into non-Google bot myths

FAQ

What is the CMA Publisher Conduct Requirement?

A conduct requirement imposed on Google on 3 June 2026 under the DMCCA digital markets regime, requiring effective publisher controls, transparency, metrics, attribution steps, and anti-retaliation for use of Search Content in generative AI shown to UK users.

When does it apply?

Six months after the 3 June 2026 Notice for most obligations, and nine months after the Notice for page-level grounding controls for search generative AI features.

Does this cover ChatGPT?

No. It is a CMA requirement on Google's general search services and the related generative AI definitions in the Notice. Other engines have separate operator controls.

Is Google-Extended enough?

The Interpretative Notes expect Google-Extended changes for certain training and grounding opt-outs, plus directory-level and page-level controls for grounding search generative AI features. Do not treat Google-Extended as the whole UK package, or as a Search ranking off-switch.

Will opting out hurt our classic Google rankings?

Google must not penalise publishers for using the controls in the ways prohibited by the anti-retaliation clause, for example purposeful downranking in general search outside search generative AI features. The Notes also recognise that Google may not control every second-order algorithmic effect.

Closing

For UK operators, June 2026 is the moment Google's generative use of Search Content stopped being only a product FAQ and became a DMCCA conduct regime. AEO still leads the visibility craft. This CR leads the UK compliance craft for Google. Run both, with receipts.

How we measure AI visibility →

Sources

Notes

The commencement calendar dates, about 3 December 2026 and about 3 March 2027, are derived by counting six and nine months from 3 June 2026. Prefer quoting the Notice's month language in legal copy, and use calendar anchors only as planning aids.
The CMA press framing of nine months for all changes, with important parts earlier, must not replace the Notice's split timetable: six months by default, and nine months for the page-level paragraph 4.b only.
Compliance Reporting Notice timings cover Google's plan, officer and report deadlines to the CMA. Those are Google compliance process, not publisher force dates, and should not be presented as publisher opt-out deadlines.
The final decision PDF and the 4 August 2026 CR summary are linked on GOV.UK and are not fully narrated here. Re-check them before quoting consultation history.
This brief is general information, not legal advice. Client-specific opt-out decisions need a named legal approver.
SMS designation runs from 10 October 2025 to 10 October 2030 per the Notice. Confirm the SMS Decision Notice before expanding on designation mechanics.
We do not assert that the controls were live on 3 June 2026, that Google-Extended alone is a full Search AI grounding opt-out, that an organic penalty is mandatory if you opt out, or that the CR extends to non-Google engines or non-UK output scope.