HOW TO MEASURE AI SEARCH VISIBILITY: ARE YOU IN THE ANSWERS OR NOT?
By Ben Fildes · 7 July 2026 · 4 min read
- Rankings and clicks now measure a shrinking slice of reality, because influence increasingly happens inside an AI answer before any click
- The metric that matters is share of AI voice: how often AI names you versus competitors for the queries in your space
- You can be winning the new game and losing on the old dashboard, so a citation-blind report will tell you to cut what works
- The measurement tools are new and imperfect, but a rough read on the right thing beats a precise read on the wrong one
To measure AI search visibility, you track how often AI answers actually name your business, for the questions that matter in your market, against the competitors who currently get named instead. That number, sometimes called share of AI voice, is the real scoreboard now. Rankings and traffic still have a place, but they miss a growing share of what decides who wins, because a buyer can read an AI answer, form an opinion and act without ever generating a click you could measure. If you only watch clicks, you are measuring the wrong thing more every quarter.
Here is how to think about measurement when the influence has moved into the answer.
Why the old dashboard lies to you now
For twenty years the search scoreboard was clean: your ranking position, your click-through rate, your sessions. It worked because influence and clicks were the same event. To influence a buyer, you had to get the click. Measure clicks and you measured influence.
The AI answer breaks that link. Now a buyer can be influenced inside the answer, before any click, and sometimes instead of one. They ask, they read a synthesised response that names a few sources, they trust it, they move on already leaning one way. If your business was named, you won something real, and none of it appears as a session.
So here is the uncomfortable situation you can end up in. Your AI citations are climbing, your brand is getting named in more answers, more people are arriving already convinced, and your rankings-and-traffic report looks flat or down, because some of those people got what they needed from the answer and never generated a tracked click. Read only the old dashboard and you will conclude the thing that is working is failing, and cut it. That is the single most expensive measurement mistake being made right now.
When influence moves into the answer, a rankings-only report measures a shrinking slice of reality. The new scoreboard adds the thing that now decides who wins.
- Ranking position — where you sit on the links page
- Click-through rate — assumes influence needs a click
- Sessions and traffic — misses zero-click answers entirely
- Flat or down when you are winning — cited but not clicked reads as failure
- Share of AI voice — how often AI names you vs competitors
- Citation rate per query — tied to the questions that matter
- Which pages get cited — tells you exactly what to fix
- Trend over time — a leading indicator that moves before revenue
| Option | Detail |
|---|---|
| Old scoreboard: Ranking position | where you sit on the links page |
| Old scoreboard: Click-through rate | assumes influence needs a click |
| Old scoreboard: Sessions and traffic | misses zero-click answers entirely |
| Old scoreboard: Flat or down when you are winning | cited but not clicked reads as failure |
| New scoreboard: Share of AI voice | how often AI names you vs competitors |
| New scoreboard: Citation rate per query | tied to the questions that matter |
| New scoreboard: Which pages get cited | tells you exactly what to fix |
| New scoreboard: Trend over time | a leading indicator that moves before revenue |
The metric that matters: share of AI voice
The number to build your new scoreboard around is share of AI voice: across the questions that matter in your space, how often does AI name you, compared with how often it names each competitor?
It answers the question that actually decides your visibility now, "when my buyer asks, who does the machine recommend, and is it me?" It is a share, not an absolute, because being cited once is meaningless if a rival is cited ten times for the same questions. And it is tied to specific queries, because you can dominate answers for questions nobody asks and still be invisible where it counts.
Treat it as your leading indicator. It moves before revenue does, because it reflects who is winning the recommendation moment that precedes the enquiry.
What to actually track
You do not need a hundred metrics. A useful AI-visibility scoreboard has a few things on it.
Your citation rate on the queries that matter. Pick the real questions your buyers ask AI, then check how often each major answer engine, ChatGPT, Google AI Overviews, Perplexity, names you when asked. That is your baseline.
Your competitors' citation rate on the same queries. The gap is the story. Being cited twice feels fine until you see the rival cited nine times for the identical questions.
Which of your pages get cited. The pattern is usually obvious once you can see it. Cited pages state clear, specific answers. Invisible ones bury the point in adjectives. This tells you what to fix.
Movement over time. A single snapshot is a photo. The value is in the trend: is your share of AI voice climbing as you do the work, or flat because you are answering questions nobody asks?
And keep your classic metrics running alongside. Rankings, traffic and conversions still matter, especially for transactional and branded searches that still drive clicks. The point is to add the new number, not replace the old ones. I made that both-games argument in full in SEO vs GEO.
Be honest about the limits
I will not oversell the precision here. The tools that measure AI citations are a couple of years old, not twenty. AI answers vary between users and phrasings, so this is sampling and estimation, not the tidy determinism of a rankings report. Anyone claiming a perfect, exact AI-visibility number is overselling it.
But imperfect measurement of the right thing beats perfect measurement of the wrong thing, and ranking-only reporting has quietly become the wrong thing. I would rather know roughly how often AI names you than know precisely where you sit on a links page fewer people scroll to. The rough number points you at the right work. The precise one points you at a shrinking slice of reality.
Start with a baseline
Everything above stays abstract until you see your own number. "Am I in AI answers?" has no obvious answer, so it stays a worry rather than a task, and nothing gets done.
So make it concrete. Get a baseline of how often AI names you against your competitors for the questions your buyers actually ask. That is the honest first step, and it is exactly where our AI Search product begins, with a free scan. See the number, see the gap, and then decide how much closing it is worth. For most operators I speak to, seeing a competitor cited where they are invisible is the moment measurement stops being a spreadsheet chore and becomes the reason to act.
Founder of Neon Gorilla. First Class BA in Marketing and an MSc in Enterprise and Innovation (Distinction) from Keele. Previously co-founded Beast Biltong with Eddie Hall, stocked in 2,000+ stores. Everything here is written from our own campaign logs, not theory.
More about Ben →Run a free scan on your own domain: see how often ChatGPT and Google AI Overview cite you versus your competitors. Then AI Search is how you close the gap. We owe you 1 client in 90 days, or it's free until you get one.