AI search visibility: what it is, whether it matters for you, and how to track it

AI visibility tracking is the practice of checking whether AI tools - ChatGPT, Google's AI Overviews, Perplexity and the rest - mention or cite your business when someone asks them a question you would want to rank for. That is the whole idea in one sentence. The harder questions are whether it matters for you, and how to do it without quietly draining hours you would rather spend elsewhere. Let me walk through both.

What AI visibility actually means

For twenty years the game was simple to describe: rank on Google, get clicks. You could open a rank tracker and see position 3 for 'plumber Hamilton' and know roughly what that was worth. AI search breaks that tidy picture, because the answer often appears without a click at all. Someone asks an AI assistant 'who makes portable spectrometers for field geology' and it names three companies in a paragraph. If you are one of the three, you have visibility. If you are not, you are invisible - and no rank tracker will tell you, because there was never a blue link involved.

So AI visibility is really asking a different question than classic SEO. Not 'where do I rank on this page of results' but 'when a machine summarises my whole category, does it include me, and does it say something accurate'. Those are two separate things, and the second one matters more than people expect. I have seen AI answers name a business but describe a service it stopped offering years ago.

Is it worth tracking? It depends who your buyers are

Here is the honest answer, and it has nothing to do with how big your business is. It has to do with where your traffic came from in the first place. If your customers historically found you by reading things - guides, comparisons, technical explainers, the sort of research-heavy content that answers a question before anyone is ready to buy - then AI is sitting directly in that path and you should be watching it closely. If your customers found you by searching 'auto electrician near me' and calling the first decent-looking result, AI barely touches that, and tracking it is mostly a curiosity. I have written more on that split in who actually loses traffic to AI search.

I built proper AI visibility tooling for one client - a B2B business selling into a technical field, the kind of buyer who reads three articles and a spec sheet before they email anyone. For them it was clearly worth it. A meaningful slice of their pipeline started with an informational search, and those searches are exactly what AI Overviews now intercept. If they were being cited, great. If a competitor was being cited and they were not, that was a real commercial problem worth a real response.

AI visibility matters most for the businesses whose customers do their homework before they buy - not the ones whose customers just want a phone number.

- Tom McSherry

How to actually track it

The mechanics are less mysterious than the tool vendors would like you to think. You are doing three things: deciding what to ask, asking it repeatedly, and recording what comes back.

• Build a set of prompts that mirror how your buyers actually ask - not your keywords, their questions. 'Best X for Y in Australia', 'who supplies Z', 'is A or B better for C'. Twenty to fifty good prompts beats five hundred lazy ones.
• Run them across the assistants your buyers actually use, on a schedule. Answers drift week to week, so a single snapshot tells you almost nothing. The signal is in the trend.
• Record two things per answer: were you mentioned, and was what it said correct. A wrong mention can be worse than no mention.
• Note who else gets named. Your AI competitor set is often not your Google competitor set, and that is genuinely useful intelligence.

You can do a scrappy version of this by hand in a spreadsheet for next to nothing, which is honestly where I tell most people to start. You only need real tooling once the volume makes manual checking silly - and by then you will know whether it is earning its keep.

The discipline problem

This is the part the excitable articles skip. AI visibility tracking can quietly eat your entire monthly retainer if you let it. There is always another prompt to test, another assistant to add, another wobble in the results to investigate. The work expands to fill whatever hours you give it, and most of that expansion produces no decision you would not have made anyway.

So I put hard edges around it. A fixed prompt set that changes rarely. A fixed cadence. A fixed, small time budget. And a rule that a tracked change only earns action if it points to something we would actually do - fix an inaccurate answer, strengthen a page that keeps getting a competitor cited instead of us, chase authority in a gap. Tracking that does not change behaviour is just an expensive anxiety subscription.

How this fits your reporting

Treat AI visibility as one input, not the headline. It sits alongside the things that actually pay the bills - rankings that still exist, traffic, enquiries, revenue. If you are unsure how it slots into a sensible scorecard, I have laid out my broader approach in how to measure SEO results, and it is worth understanding what GEO is before you spend a dollar chasing it.

My overall take: for research-led B2B, start tracking now, but keep it on a tight leash. For everyone else, a quarterly manual check is plenty. I hold that view loosely - the landscape is moving fast and I would be happy to be proven wrong within a year. But I would rather you tracked deliberately than either ignored it or drowned in it.

Whether tracking this is worth your time depends on what kind of business you are - see AI search for local business.