Google AI Search Controls: Measure Before You Touch the Toggle

Google has just made AI Search inclusion feel a lot less theoretical.

On 3 June 2026, the UK Competition and Markets Authority announced a conduct requirement for Google Search. The practical version: publishers need proper tools to stop their content being used to power AI features in Search, and Google has to improve attribution with clear links.

Google also published its own update the same day. It is beginning to test a Search Console toggle that lets website owners decide whether their site appears in and helps ground generative AI Search features, including AI Overviews, AI Mode, and AI Overviews in Discover. It is also rolling out new Search Console insights for a subset of UK site owners, with impressions and page/country visibility data for generative AI features.

This is useful. It is also exactly the kind of admin toggle that looks like a light switch and behaves more like a fuse box.

So before anyone clicks it in a mild panic, make a measurement plan.

What changed

The CMA says publishers will now have effective tools to prevent their content being used to power AI features in Google Search, such as AI Overviews. Google also has to make sure publisher content is properly attributed with clear links in AI-generated search results, and the requirement covers content use for AI model fine-tuning too.

Google's own post says the new Search Console control will let website owners opt out of appearing in and grounding generative AI Search features. Google says sites that opt out will not receive traffic or impressions from those generative AI features, and that the control will not be used as a ranking signal for search results outside those features.

There are two extra bits worth noticing:

• Google says AI Overviews now has more than 2.5 billion monthly active users, and AI Mode has passed one billion monthly users.
• The new controls and insights are starting with a subset of UK website owners before a wider rollout.

That makes this both a policy story and a marketing operations story.

The wrong question is "should we opt out?"

At least as a first question.

The better first question is: what job does Google AI Search currently do for this site?

For a publisher whose content is the product, the answer may involve licensing, attribution, bargaining power, and whether AI answers are replacing paid or ad-funded reading. That is a serious commercial decision, not a tweak for Friday afternoon.

For a SaaS, agency, ecommerce, local service, charity, or public information site, the answer may be different. Your content might be less about protecting a paid article archive and more about being found by someone who is comparing options, checking trust signals, or looking for a credible next step.

Opting out could be sensible for some sites. It could also remove a growing discovery surface just when the reporting is finally getting clearer.

That is why the toggle needs a runway.

A practical measurement plan before changing anything

Here is the version I would use.

1. List the pages that matter

Do not start with the whole site. Start with the pages where the business would actually care about AI Search visibility:

• homepage,
• pricing,
• product or service pages,
• high-intent comparison pages,
• help docs or guides,
• newsroom or editorial pages,
• evergreen explainers that already win search traffic.

If a page would never be useful inside an AI answer, it does not need to lead the decision.

2. Separate revenue pages from rights-sensitive content

Put each page into one of three buckets:

• Discovery pages: pages you want surfaced as often as possible.
• Rights-sensitive pages: content where reuse, summarisation, or licensing matters.
• Private or low-value pages: pages that should not be part of public discovery anyway.

Most messy decisions come from treating those buckets as one blob.

3. Capture the current signals

Before you touch controls, record the basics:

• Is the page indexable?
• Is crawling allowed in robots.txt and by the CDN/WAF?
• Is the canonical tag correct?
• Is the page eligible for snippets?
• Does the page have clear title, description, article/product structure, and visible text?
• Are AI crawler rules, Content Signals, llms.txt, and snippet controls saying compatible things?

Google's own guidance is still fairly grounded here: for AI Overviews and AI Mode, foundational SEO still matters, pages need to be indexed and snippet-eligible, and there is no special schema or AI text file required for Google generative AI Search.

That does not make files like llms.txt pointless across the wider AI ecosystem. It just means you should not confuse "useful cross-engine context" with "a magic Google ranking lever". Small distinction, large number of LinkedIn posts avoided.

4. Watch Search Console, but do not stop there

The new Search Console data is helpful because it starts to separate AI visibility from the soup of normal performance reporting.

But early AI Search reporting should not be treated as a complete ROI dashboard.

Pair it with:

• landing page sessions,
• conversion events,
• branded and non-branded query trends,
• server logs for crawler access,
• CRM or lead-source notes where available,
• week-by-week page changes.

The question is not just "did we get impressions?" It is "did the right pages get discovered, did the right people arrive, and did anything useful happen after the click?"

5. Decide by page type, not by vibes

A reasonable first policy might look like this:

• Keep high-intent commercial pages eligible for AI Search while you measure.
• Keep useful public guides eligible if they support acquisition and trust.
• Review rights-sensitive editorial content separately.
• Keep private, account, checkout, and internal pages excluded through normal indexing and access controls.
• Avoid broad robots changes unless you understand every bot and product affected.

The dangerous version is one global rule copied from someone else's business model.

6. Write the rollback note before you click

If you do opt out, write down:

• who approved it,
• which property or section it applies to,
• the baseline date range,
• what metric would make you reverse it,
• when you will review the decision.

That sounds overly formal until three people ask why AI Search impressions vanished and nobody remembers who touched the setting.

What Scavo should keep checking

This change makes AI visibility less abstract, but the underlying operational checks stay familiar.

The important failures are still boring:

• crawl policy says one thing, edge rules do another,
• robots.txt and sitemap disagree,
• AI crawler access is blocked accidentally,
• snippet controls suppress the useful part of a page,
• llms.txt exists but points at stale or low-value URLs,
• structured data says something the visible page does not,
• important content only appears after fragile client-side rendering,
• pages have no clear authorship, date, price, evidence, or next step.

Those are the things that make a site hard to trust, quote, preview, or click.

The new Search Console control does not remove that work. It just means more teams will finally have a visible place where the consequences show up.

A sensible default for most non-publisher sites

For most small SaaS, service, ecommerce, and agency sites, I would not rush to opt out.

I would do this instead:

1. Keep the key public pages eligible.
2. Fix the crawl, indexability, snippet, and citation basics.
3. Track the new Search Console AI visibility data when it appears.
4. Compare AI visibility pages with conversion and lead quality.
5. Revisit the policy once there is evidence, not just mood music.

There will be exceptions. If your content itself is the monetised product, or if legal/commercial policy says reuse is not acceptable without a deal, that is a different conversation.

But for a business trying to be found by people with messy, high-intent questions, the boring answer is probably best: measure first, decide second, monitor forever.

What to do next in Scavo

1. Run a fresh scan on your homepage and one high-intent content page.
2. Open the AI Visibility checks first: crawler policy, AI bot parity, snippet control safety, citation readiness, llms.txt, Content Signals, and markdown negotiation.
3. Fix contradictions before changing any Google Search Console AI control.
4. If you get access to the new Search Console AI insights, record the baseline date range before changing anything.
5. Re-scan after policy, CDN, CMS, or robots changes so the live site matches the decision you thought you made.

Sources

• CMA: Fairer deal for publishers and Google search services in the UK
• Google: New opportunities, control and insights for website owners
• Google Search Central: AI features and your website
• Google Search Central: Optimizing for generative AI features on Google Search
• Google: 5 new ways to explore the web with generative AI in Search