Everyone wants a number for AI search. How often does my brand show up in ChatGPT? In AI Overviews? In AI Mode? Tools now happily hand you a score.
The problem, as Search Engine Journal argues in a new piece built on Ahrefs research, is that a score going up or down rarely tells you what to do next.
Demand is exploding, clarity isn’t
Ahrefs data cited in the report shows US searches for “AI search tracking” up 184% year over year, and “AI rank tracking” up 175%. That’s appetite, not adoption ” but the direction is obvious. Every growth team is being asked the same question by leadership: are we visible in AI or not?
Before you answer, understand what’s inside the number.
A mention and a citation are different animals
Most visibility scores blend at least two events:
- Mention ” your brand name appears in the answer, with no link.
- Citation ” a URL of yours is credited as a source, sometimes while the answer discusses something else entirely.
Roll both into one metric and you lose the diagnosis. A drop could mean you lost share of voice on a topic, or simply that a page stopped being cited. Those need completely different fixes.
Retrieval muddies it further. Ahrefs looked at 1.4 million ChatGPT prompts and found Reddit URLs were retrieved at scale yet cited in just 1.93% of cases. Retrieval isn’t citation, and citation isn’t influence.
Classic rankings still do heavy lifting
Anyone declaring SEO dead should sit with this: across 863,000 keywords and 4 million AI Overview URLs, 37.1% of cited URLs also sat in Google’s organic top 10 for the same query. Another 26.2% ranked between 11 and 100, and 36.7% were nowhere in the top 100.
So rankings are a meaningful input, not a requirement. Ahrefs points to query fan-out as the explanation for the outliers: Google breaks a query into sub-queries, and pages strong on adjacent topics get pulled in even without ranking for the head term. Which means head-term rank tracking will systematically miss a chunk of your AI presence.
The schema reality check
Here’s the finding that should recalibrate a few GEO decks. Pages cited by AI were nearly three times more likely to contain JSON-LD. Correlation looked strong.
Then Ahrefs tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 control pages. AI Mode citations rose 2.4%, ChatGPT 2.2% ” both statistically insignificant. AI Overview citations actually fell 4.6%, a small but significant decline, roughly 12 fewer citations a day on pages already pulling hundreds. Both groups were already trending down beforehand, and Ahrefs is careful to say it can’t prove causation.
Note the sample: pages with at least 100 prior AI Overview citations. So this says nothing about whether schema helps an unknown page get discovered in the first place.
More interesting from the same ChatGPT analysis: cited pages had titles that matched the sub-queries ChatGPT generates more closely than non-cited pages did ” and that gap was wider than against the original prompt. URLs with natural-language slugs were cited more often. Write for the questions behind the question.
The score can rise while clicks fall
This is the trap. Ahrefs analysed CTR across 300,000 keywords using desktop Search Console data, comparing December 2023 to December 2025. For keywords with an AI Overview, top-result CTR was 58% below expectation ” an additional decline on top of an already falling baseline, drawn from informational queries.
Translation: you can hold rankings, win citations, and still lose traffic. Appearing in an answer and getting a visit are separate outcomes.
What to actually do
Treat the score as a triage signal, not a verdict. AI answers are probabilistic ” the same prompt returns different brands on different runs ” so use large prompt sets and check whether a “change” is really a prompt edit or a model swap.
Then read the metrics diagnostically. Mentions down with prompts flat? Look at competitor share, which topics slipped, and whether it’s one platform. Citations down with rankings stable? Check who got cited instead and whether your coverage matches how queries decompose. Rankings down? That’s a traditional SEO problem wearing an AI costume.
Pair all of it with outcome data: Search Console’s AI performance reports (impressions only for now) plus AI referral tracking in analytics. The score points at the question. Revenue answers it.
Source: Search Engine Journal
