Search Console Generative AI Reports: How to Read AI Impressions Correctly

The short answer

Since 3 June 2026 Google Search Console has shipped dedicated reports for generative AI surfaces, and since 11 August 2026 they are rolled out worldwide for every property. For the first time you can read impressions from AI Overviews and AI Mode separately from classic results. It is real progress — but it does not answer the question most teams actually ask: "are we being cited?"

What the report gives you

• Separate impressions for generative AI features in Search and Discover instead of the old blended totals.
• URL level: which pages appeared inside AI answers.
• Clicks and position using the same logic as the classic report.
• A time series starting at your property's rollout date, not backfilled for the full year.

What it does not give you

These gaps cause most misreadings:

1. No query completeness. Long, conversational prompts are frequently anonymised or omitted. The very questions asked in AI Mode are therefore underrepresented.
2. No citation signal. An impression means the URL appeared in the surface — not that your brand was named in the answer text.
3. No split between AI Overviews and AI Mode. Both sit in one bucket even though user behaviour and click probability differ sharply.
4. No before/after. Without a pre-rollout baseline you cannot cleanly quantify the format shift.

The three most common false conclusions

How to read it properly

Step 1 — Segment. Compare AI impressions and classic impressions per URL cluster (product, guide, glossary). The split differs massively: explanatory content wins AI impressions, commercial pages often do not.

Step 2 — Keep CTR separate. Never compute a blended CTR across both reports. AI CTR is structurally lower; blended, every content strategy looks like decay.

Step 3 — Pair it with citation testing. The Search Console report says "visible"; your own prompt test series says "cited". Only both together produce a metric you can steer on. Fix a set of 30 to 50 buying-intent questions and run it monthly.

Step 4 — Decide on opt-out deliberately. Google offers controls to keep content out of generative surfaces. For most brands that is the wrong reflex: opting out loses mentions without recovering clicks. Document the decision instead of inheriting a default.

A realistic reporting setup

A monthly AI visibility board with four metrics is enough to start:

• AI impressions per content cluster (from the new report)
• Citation share on a fixed question set (your own measurement)
• Referral traffic from AI surfaces (analytics, consent dependent)
• Branded search volume as a lagging indicator

Teams that keep reporting sessions alone will cut exactly the content that carries the brand inside AI answers.

Next steps

Pull the first complete weeks from the report, build a per-cluster baseline and add citation measurement on top. How we run that operationally is on our GEO agency page; the underlying concepts are in the glossary under answer engine optimization and zero-click brand equity.