Google’s Generative-AI Search Toggle: Build an Opt-Out Decision Record
Google is testing a Search Console control for generative-AI Search visibility. Save the scope, baseline, approval, observation window, and rollback rule first.
Direct answer: Do not change Google’s new generative-AI Search control without a written decision record. Google says the tested Search Console toggle can remove an eligible site from appearing in or grounding AI Overviews, AI Mode, and AI Overviews in Discover. An opted-out site will not receive traffic or impressions from those features.
The announcement describes a test for a subset of sites in the United Kingdom. It is not a globally available setting, a ranking control for ordinary Search, or evidence that opting in or out will improve business results.
Confirm the control and its scope
Capture the exact Search Console property, verified owner, market, date, interface label, current state and surfaces named in the control. Do not infer availability from a screenshot belonging to another publisher.
| Surface | Documented effect of opting out | Do not infer |
|---|---|---|
| AI Overviews | No appearance or grounding | A change to ordinary rankings |
| AI Mode | No appearance or grounding | A guaranteed referral loss |
| AI Overviews in Discover | No appearance or grounding | All Discover visibility ends |
| Search outside these features | Not used as a ranking signal | Every result remains unchanged |
The final row needs care. Google states that the control is not used as a ranking signal outside the generative-AI features. That does not mean the entire search ecosystem is static or that no indirect business effect can occur.
Save a baseline before the decision
Record the longest comparable pre-change window available. Separate ordinary Web Search, the generative-AI report where eligible, Discover, referral analytics, conversions, subscriptions and direct audience measures. Preserve query, page, country and device dimensions where privacy and reporting limits permit.
The baseline is not a forecast. Seasonality, product releases, news cycles, ranking changes, demand, consent, measurement definitions and other site releases can move the same numbers. Add those confounders to the record before anyone interprets a change as the toggle’s effect.
Use the site’s generative-AI reporting guide for the relevant fields and the measurement crosswalk to keep impressions, citations and referrals separate.
Write the decision record
Name the business objective, affected content, rights concern, expected benefit, accepted cost, accountable owner, approver, change time, observation window, rollback condition and next review date. A publisher protecting licensed reporting may make a different choice from a merchant seeking product discovery.
Separate principle from measurement. A publisher may opt out because it does not permit this use even if referrals are positive. Another may participate while monitoring attribution. The record should make that governing choice explicit instead of pretending one traffic number answers every policy question.
If the site is not in the test, publish no speculative implementation instructions. Record the announcement, current ineligibility and what evidence would trigger a future review.
Observe without claiming causation
- Annotate the change in the analytics and editorial release ledger.
- Verify the saved state in Search Console with the correct property owner.
- Check the named generative-AI surfaces and ordinary Search separately.
- Compare the declared windows without hiding missing or anonymized rows.
- Review legal, editorial and audience outcomes alongside traffic.
- Apply the rollback or review rule written before the result was known.
A result can support a decision within the observed property and period. It cannot establish a universal effect for every site, market or Google surface. If the control changes during the test, start a new observation rather than merging incompatible states.
Ask a question or join the discussion