Glossary
AI Overviews
AI Overviews are Google's AI-generated summaries that appear above traditional results for many queries, synthesizing an answer from multiple sources with citation links. For brands they are a high-stakes surface: an Overview can answer a buyer's question — accurately or not — before any organic listing gets the chance to earn a click.
AI Overviews matter disproportionately because they sit on top of the largest search surface in the world. A category query that once sent traffic to ten comparison pages can now resolve in the Overview itself, with only the cited sources getting meaningful visibility and everyone else losing the impression entirely.
Appearing in an Overview is not the same as ranking. Google composes them from pages it judges reliable and extractable for that specific question, which rewards clear standalone answers, entity clarity and structured data — classic AEO — sometimes over pages that outrank them organically. The gap between those two states is the most useful thing to measure, because it is diagnosable: a page that ranks but is never drawn from is failing extractability, not authority.
Two properties make Overviews awkward to monitor, and both are worth knowing before you start. They do not trigger on every query, so your denominator is not your whole keyword set — it is the subset that produces an Overview at all. And what appears varies by location, device and signed-in state, so a single check from your own browser is a data point about you rather than about the query.
What it is made of
AI Overviews is an umbrella over several distinct measurements. Conflating them is what makes the term mean different things to different people, so each one is defined below with the arithmetic to compute it.
- Trigger rate
-
The share of your target queries that produce an AI Overview at all. This is the denominator for everything else, and it is the number most people skip — absence from an Overview that never appears is not a problem to solve.
How to compute: Queries returning an Overview ÷ total queries checked, recorded per run because triggering changes over time.
- Presence rate
-
How often your brand is named in the Overviews that do appear. Measured against triggering queries only, never against the full keyword set, or the figure flatters or punishes you for reasons unrelated to your content.
How to compute: Overviews naming your brand ÷ Overviews that appeared.
- Citation share
-
The proportion of the sources an Overview links to that belong to you. Being named without being cited means Google composed the claim about you from somewhere else — which tells you where to go and fix it.
How to compute: Citations pointing at your domain ÷ all citations across the Overviews in the run.
- Rank–Overview divergence
-
Queries where you rank well organically but are absent from the Overview. This is the most actionable metric on the page: authority is evidently not the blocker, so the failure is extractability — the page does not contain a liftable answer to that specific question.
How to compute: Count of queries where organic position ≤ 10 and brand presence in the Overview is no. List them; do not average them.
- Competitor presence
-
The same presence rate measured for your named rivals on an identical query set. The absolute number matters less than who occupies the Overview on your highest-intent comparison questions.
How to compute: Per competitor: Overviews naming them ÷ Overviews that appeared, on the same run.
How to monitor whether AI Overviews mention you
There is no rankings page for Overviews, and Search Console does not separate them out, so the only way to get a number is to check the queries yourself under controlled conditions. The point of the controls is not rigour for its own sake — it is that an uncontrolled check measures your browser rather than the query.
-
Fix a query set and record organic position alongside
Pick 20–30 buyer queries and freeze the list. Record where you rank organically for each one in the same run — that second column is what turns the exercise from a presence check into a diagnosis, because rank-without-presence and no-rank-no-presence call for completely different work.
-
Check from a clean, consistent context
Incognito or a fresh profile, signed out, with the same location and device type every run. Overviews vary by all three, so a set checked from a signed-in account in one city and later from a phone in another produces a difference that means nothing. Consistency matters more than which context you pick.
-
Record whether an Overview appeared at all, first
Before looking for your brand, log trigger or no trigger. Overviews do not appear for every query and the pattern shifts over time. Without this column your presence rate silently drifts as triggering changes, and you will read that drift as your own performance moving.
-
For each Overview, log presence, citations and competitors
Was your brand named; which URLs were cited and how many were yours; which rivals were named and in what order. Paste the Overview text rather than summarising it — the specific claim it makes about you is the thing you will want later, and it is not reconstructable from a yes or no.
-
Isolate the divergence list
Filter for queries where you rank in the top ten and are absent from the Overview. That list is your work queue, and it is short enough to act on. For each one, open your ranking page and ask whether it contains a sentence that answers that exact question and could be quoted on its own. Usually it does not — and that is the fix.
-
Repeat monthly and compare like for like
Same queries, same context, same scoring. Read the change rather than the level: an Overview that stopped citing you, a competitor newly named on your top comparison query, a claim about your pricing that went stale. Monthly is enough; Overviews shift, but not so fast that weekly checking tells you anything a month of data would not.
All of it is a spreadsheet and a controlled browser. The tedious half — running the set on a schedule and diffing what changed — is what monitoring tools automate, but the divergence analysis in step five is judgement about your own pages and worth doing by hand at least once.
Frequently asked
What determines whether Google AI Overviews mention you?
Two filters in sequence. Google has to consider your page a credible source for the question, and it has to find a passage on it that answers the question and can be lifted intact. The second filter is where most pages fail, which is why a page can rank in the top five and still never appear — it is about the topic without ever stating the answer outright near the top.
Why do I rank well but never appear in the Overview?
Because ranking and Overview selection test different properties. Ranking rewards a page that comprehensively covers a topic; Overview composition needs one self-contained, quotable statement resolving one question. Long narrative guides frequently outrank everything and offer nothing liftable. Track this as rank–Overview divergence: it is the most fixable gap you can measure, because authority clearly is not the obstacle.
How do I track competitor performance in AI Overviews?
Run the identical query set and log which rivals are named and in what order, alongside your own presence. The useful output is not a league table but the specific queries where a competitor holds the Overview and you do not — particularly on comparison and pricing questions, where the Overview reaches a buyer who is close to deciding.
Can I see AI Overview performance in Search Console?
Not separately. Search Console reports clicks and impressions but does not break out whether an impression came from an Overview, so you cannot isolate the surface there. That is precisely why the manual check exists — the query set, the controlled context and the trigger column are standing in for a report that does not exist.
How do I prove the value of AI Overview work?
Not through traffic alone, because an Overview that answers the question may produce no click at all — the value shows up as brand awareness rather than a session. The defensible measures are presence rate on your highest-intent queries over time, citation share, and the divergence list shrinking. Pair those with branded search volume, which tends to move when a brand starts appearing in answers people read without clicking.
Do AI Overviews appear for every query?
No, and this matters for measurement. Triggering varies by query type and shifts over time, so your denominator is the subset of queries that actually produced an Overview in that run — not your full keyword list. Logging trigger or no trigger before anything else is what keeps the presence rate meaningful between runs.
See what AI currently says about your brand
Perciva runs buyer-intent prompts across ChatGPT and Gemini and shows you the answers verbatim — including which competitor gets recommended instead of you.
Get a free snapshot