AEO monitoring
AI Overviews ≠ answer-engine citations
Two surfaces, two metrics, one common conflation, and what each one is for.
There is a slide in a lot of monthly decks with one number on it and the words AI visibility underneath. Two people look at that slide. One of them is picturing a block on Google's results page. The other is picturing ChatGPT.
They are not looking at the same thing, and the number is an average of two worlds with almost nothing in common. Teams blend them weekly, and the blend measures neither. The distinction takes two minutes to learn and saves quarters of misdirected work.
What is an AI Overview, exactly?
An AI Overview is the AI-written block Google shows on top of a classic results page, on some queries and not on others.
The metrics that describe it are presence metrics. How often an Overview shows up on the keywords you track, and whether you are inside it when it does. It lives entirely inside Google's ecosystem, responds to work you do on the results page, and competes with your own blue links for the same click.
That competition has a cost you can measure. When an Overview appears and you are not in it, the same ranking position earns fewer clicks, which is why striking-distance math applies a haircut for exactly this case.
Presence also varies by market and by query class, so you read it on your tracked keywords rather than in general. An Overview that shows on your informational queries but never on your commercial ones is a completely different situation from the reverse. Only the per-keyword view tells you which of the two you have.
What counts as an answer-engine citation?
A citation happens somewhere else entirely: inside ChatGPT, Perplexity, Gemini or AI Mode, when the assistant composes an answer and lists your page among the sources it used.
The metrics here are citation rate and citation share, meaning how often an AI answer names one of your pages as a source across the questions you track, read engine by engine. These respond to different work altogether. Whether AI crawlers can fetch you. Whether your passages survive being lifted out of context. Whether the answer economy treats you as a source at all. None of that is results-page work.
Beside the citation sits the mention, the weaker relative: your name in the answer without your page in the sources. Citations count above mentions in the roll-up, and that ordering applies only on this surface, because an AI Overview has no equivalent of a mention at all.
Ask it yourself
How often do AI Overviews show on our tracked keywords, and separately, what's our citation rate across AI assistants?
See also: how citations, mentions and share of voice get measured →
Why does blending them cost money?
Because the blended number sends the wrong team to work.
Overview inclusion moves with your rankings and your presence on the results page. Citation share moves with fetchability and passage quality. So a blended figure that dips can dispatch a content team to fix what was actually a robots.txt problem, or send an engineer after a passage-quality problem. Kept apart, each number names its own workstream, which is the entire reason metrics exist.
The error compounds in a trend line. The two metrics move on different rhythms, Overviews with Google's rollouts and your rankings, citations with engine behavior and your fetchability, so a blended trend is two unrelated stories fighting over one slope.
The tell that a team has blended them is the phrase AI traffic used with no surface attached to it. Ask which surface. In our experience the workstream sorts itself out in the same meeting.
The two ledgers stay separate
In Quattr they live in different families of analysis on purpose. One family reads AI Overview inclusion beside featured snippets, on the results-page side. Another reads citations, share of voice and sentiment, on the answer-engine side. No analysis sums them.
The example below shows the results-page half of that ledger: AI Overview visibility benchmarked against competitors, on its own surface, with its own metric.
The separation shows up in the deck too. The results-page slide and the answer-engine slide are different slides, each with its own scope note and its own as-of date, which is the date that data was last complete. No executive summary adds them together.
The workflow that does this: AI Overview visibility vs competitors →
A third surface makes it matter more
Google's AI Mode is a full conversational search surface with its own behavior, and it deserves its own metric rather than a seat inside either blend.
The habit that scales to all three is small. Name the surface before you read the number. Which surface, which metric, which piece of work. Any conversation about AI visibility that starts there ends better than one that does not.
It also future-proofs your reporting. Surfaces will keep multiplying, and a team that names its surfaces survives each new one with a new row rather than a new argument.
See also: why AI Mode is a third surface and not a bigger AI Overview →
Frequently asked
- Which number should we put in the board deck?
- Both, on separate lines, each labeled with its surface. An AI Overview show-rate answers whether Google is summarizing your queries and whether you are inside the summary. A citation rate answers whether assistants treat your pages as sources. Averaging them produces a figure that nobody can act on.
- Where does AI Mode go, results page or answer engine?
- Its own row. Its reach behaves like Google and its answers behave like a conversation, so it is measured with citation-style metrics but reported separately from ChatGPT and friends.
- Our AI Overview inclusion fell and our citations rose. Which one is right?
- Both are, and that is the point of keeping them apart. Those two surfaces respond to different work, so opposite moves in the same month are ordinary rather than contradictory. A blended line would have shown you a flat quarter and hidden both stories.