Yes, and the surprise is where. As of early August 2026, Google confirmed that AI Mode query data lands in your main Search Console performance report — the same table as your classic Search queries — not in the separate Generative AI performance report. Because AI Mode counts every follow-up turn as a new query, conversational fragments like "yes, pricing" or "yes go on" are AI Mode turns hiding in plain sight in your keyword data.
This is part of the measuring AI visibility series, and a direct follow-on to what the Generative AI performance report shows and can't. That report answers "did I appear in an AI feature?"; this note is about a different data home — where the AI Mode question text actually surfaces, and how to read it honestly.
Where does AI Mode query data actually show up in Search Console?
In the main Performance report's query table, alongside classic Search queries — not in the Generative AI report. Google confirmed in early August 2026 that the Generative AI performance report has no query dimension (it exposes impressions, pages, countries, devices, and dates only), so AI Mode's query strings flow into the general performance data instead. There are now two AI-relevant data homes in Search Console, and they hold different things:
| Search Console surface | What it isolates | Query text? | Clicks? |
|---|---|---|---|
| Generative AI performance report (launched 3 Jun 2026) | Impressions inside AI Overviews and AI Mode | No | No |
| Main Performance report (Search) | All Search queries, AI Mode turns included but unlabeled | Yes | Yes |
The practical consequence: the report that tells you how often you appeared in AI features (the Generative AI report) is silent on which questions surfaced you, while the report that holds the AI Mode question text (the main report) doesn't tell you which rows are AI Mode at all. The data is split across two views, and neither joins them for you.
Why does each follow-up turn count as a new query?
Because AI Mode treats a conversation as a sequence of independent queries, not one session. Google's own documentation states that if a user asks a follow-up question within AI Mode, they are essentially performing a new query — and the impression, position, and click data in that new response are counted as coming from this new user query. So a single AI Mode conversation of five turns can generate five separate query rows in your data, each attributed to the words of that turn.
That design is why the query text looks so strange. A person mid-conversation doesn't re-type the full context; they reply in shorthand. Those replies become the "query."
Each follow-up turn in an AI Mode conversation is logged as its own query — so a five-turn conversation shows up as five separate, conversational-looking rows in your Search Console data, not one.
How do you recognise an AI Mode conversation turn in your data?
Look for queries that only make sense as replies, not as things someone would type into a search box cold. They are short, dependent, and conversational. Common shapes practitioners have reported spotting in mid-2026:
- One-word confirmations — "yes", "sure", "ok"
- Continuation prompts — "yes go on", "tell me more", "and then"
- Refinements that assume prior context — "yes, pricing", "the cheaper one", "for a small team"
- Pronoun-heavy fragments with no named subject — "does it integrate with that"
None of these carry a label saying "AI Mode." You are inferring from the conversational shape, so treat the pattern as directional evidence, not a confirmed segment. The tell is dependency: if a query is meaningless without a preceding turn, it almost certainly is a following turn.
What can't you do with this data? (the honest limits)
The limits matter as much as the discovery, and naming them keeps you from over-reading the data:
- You can't break AI Mode out. There is no filter, dimension, or flag that isolates AI Mode queries from classic ones. You can pattern-match the fragments, but you can't measure AI Mode as a clean segment.
- You can't reconstruct the conversation. Turns aren't threaded or grouped by session, so you see scattered fragments, not the journey they belonged to.
- Position is unreliable here. As with AI Overviews, AI-feature placement doesn't map cleanly onto the classic "average position" metric, so read presence and query text, not rank.
- Volume is understated and noisy. Only turns that produced an impression for your site appear, and only when they clear Google's query-sampling thresholds — so this is a keyhole view, not a census.
This is the same honesty the Generative AI report demands: a real first-party signal and a narrow slice at the same time. It does not contradict the standing caveat that Search Console isn't a clean AI-prompt log — the regex query-mining method still treats GSC as a proxy for AI intent. What's new is narrower and specific: AI Mode's own follow-up turns do land in the main report, as conversational fragments you can spot but not cleanly segment.
What should brands do about it?
Read the conversational fragments as conversion-journey breadcrumbs, then structure content to be citable at every turn — not just the opening question. A practical loop:
- Pattern-match the fragments. Scan your query table for conversational, dependent strings and group them by likely intent (confirm, continue, refine, compare, price).
- Map them to journey stage. A "yes, pricing" turn is a late-funnel signal; "tell me more" is mid-funnel. The mix tells you where AI Mode conversations about you are heading.
- Answer every branch extractably. Query fan-out already means one question spawns many sub-queries; AI Mode conversations add sequential turns on top. Cover the predictable follow-ups — pricing, integrations, comparisons, who-it's-for — as self-contained, answer-first chunks so you can be cited at each turn, not only the first.
- Corroborate in the engines. Because you can't segment AI Mode in GSC, verify with answer-level checks whether you're actually named and cited across engines, the way the five-metric reporting stack prescribes.
The strategic read is the one worth keeping: AI Mode behaves less like a search box and more like a multi-turn conversation, and each follow-up turn is a measurable query. If your brand isn't structured to be cited accurately at each turn, you simply won't appear in those rows — and you won't know the conversation happened.
Turning scattered, unlabeled query fragments into an answer-level view of whether you're cited across every AI surface, conversation turn by conversation turn, is exactly what Buffy Intel is built to do. Questions: [email protected].