LLM Visibility Tracking: Know What AI Says About You
LLM visibility tracking tells you whether ChatGPT, Gemini, Perplexity, Copilot and Google's AI answers name your business, for which questions, and who they name instead. It runs on tools you could buy yourself. What we add is the reading of the numbers and the work that follows. Nobody can promise a mention, and we will not.
What LLM visibility tracking actually measures
Five countable things, and none of them is a ranking position.
LLM visibility tracking measures one thing: whether an AI assistant names your business when someone asks it a question you sell against. Not your position in a list. Your name, inside the answer.
The method is plain once you see it. A tool asks the same set of questions on a schedule, across several assistants, and stores every answer word for word. Then it counts. Over a month that gives you a picture of how often you appear, and who appears in your place.
The five things worth counting
- Whether you are named at all, on each assistant, for each question.
- Which questions bring you up, so you learn what you are known for.
- Which assistants name you, because ChatGPT, Gemini, Perplexity, Copilot and Google's AI answers behave differently.
- How you are described, since a wrong description does more damage than no mention.
- Who else is in the answer, which is usually the number that gets a reaction in the room.
That last one matters more than people expect. Boards rarely ask about share of voice. They ask why a competitor keeps coming up. LLM visibility tracking gives you the answers in full, quoted, so the conversation stops being a guess and starts being evidence.
None of this is a rankings report with a new coat of paint. There is no position one inside a generated answer. There is named, or not named, and how often that holds up across many runs.
Why LLM visibility tracking is not rank tracking
There is no Search Console for ChatGPT, and that changes what any number can mean.
Google hands you a free tool that reports real impressions and real clicks. ChatGPT gives you nothing of the kind. No impression count, no click data, no list of the questions people actually typed. So LLM visibility tracking cannot report what happened out there. It reports what happens when a tool asks the question again, on your behalf.
That is sampling, and sampling comes with rules. Ask the same assistant the same question twice and the wording can change. Answers also vary between users, because location, account history and the model version all feed in. A model update can move your figures overnight while nothing on your website has changed at all.
Why the numbers are directional
Directional means the shape is reliable and the exact figure is not. Say you move from one mention in ten runs to one in three. Something real has shifted. A move from three in ten to four in ten tells you very little. We label which is which, because a false decimal point sends people off spending money on the wrong thing.
Two further differences are worth knowing. A page can sit at number one on Google and never get quoted by an assistant. And a page almost nobody finds on Google can be quoted daily, because it answers one question cleanly and early. Our guide to appearing in ChatGPT and Google AI Overviews explains why that happens.
The tools behind LLM visibility tracking, and what they cost
Real products, real prices, in the currency their makers sell in.
This service runs on third-party tools. We did not build a tracker and we will not pretend we did. Every LLM visibility tracking product works the same way underneath: it asks, it records, it counts. The prices below are the vendors' entry tiers, quoted in US dollars because that is how they sell.
- Otterly.ai from $29 a month on Lite, covering 15 prompts, and $189 a month at Standard.
- Peec AI from $95 a month for 50 prompts, with Claude and GPT-5 held back for enterprise plans.
- Profound from $99 a month on annual billing, $399 at Growth, and custom pricing above that.
- Semrush bundles AI visibility inside its wider suite, from $139 a month.
- Ahrefs Brand Radar at $199 a month for one platform, $398 for a selection, $699 for all six.
- Scrunch AI at $250 a month on annual billing, or $300 month to month, covering seven engines.
So why hire anyone
Because the subscription is the cheap part. A tool will tell you that you were named in eighteen runs out of sixty. It will not tell you which of those questions is worth chasing. Nor why a rival owns the one that converts, or what to change on the site this month. That reading is the job, and a dashboard cannot do it.
Our work starts at £495 a month, rolling monthly, with no lock-in. The tool licence sits on top at whatever the vendor charges. You can hold it in your own name if you would rather own the data outright. For the wider market picture, our guide to how much SEO costs in the UK covers what agencies charge.
How we run it, month by month
Three steps, a monthly rhythm, and a short read written by a person.
We start with the questions. This is the part that decides whether any of it is useful. We want the ones your customers actually ask an assistant, in their own words. That is rarely the keyword list a search tool would hand you. Twenty to fifty questions is usually enough to show a pattern.
Then we check the assistants can reach you at all. A site that blocks the wrong robot cannot be quoted, however good the writing is. That check takes about an hour and most reports never mention it. Our public teardown of Bupa's website shows the kind of technical reading behind it.
The crawler check most reports skip
Training bots and retrieval bots are different animals. GPTBot gathers pages that may train future models. OAI-SearchBot is the one that surfaces sites inside ChatGPT's search, and OpenAI's own documentation states that the two settings are independent. Block the wrong one and you remove yourself from answers without meaning to.
Google is built differently, which catches people out. Google-Extended is a robots.txt setting rather than a crawler, and Google's own announcement describes it as a control over Gemini and Vertex AI. Ordinary Googlebot still does the fetching. Switching Google-Extended off will not take you out of AI Overviews.
After that it settles into a rhythm. We run the set monthly, since that matches how fast answers and model versions move. Weekly runs cost more and mostly show noise. You get a short written read: what moved, what did not, and what we are changing next. LLM visibility tracking only earns its money when something happens because of it.
We agree the questions
Twenty to fifty real customer questions, in their words. Get this wrong and every number afterwards is measuring the wrong thing.
We run them every month
Same questions, same assistants, same schedule. We also check that retrieval crawlers can reach your pages.
You get a written read
What moved, what is noise, which rival is taking your place, and the one thing we are changing next month.
A mention is not a citation, and the difference matters
One builds reputation. The other sends you visitors you can count.
A mention is your name in the answer text. A citation is a linked source the assistant pulled from and shows to the reader. They look alike in a report and they do very different jobs.
Citations bring visitors. Perplexity lists its sources, Google's AI answers link out, and Copilot footnotes what it used. Plain ChatGPT often answers with no link at all, so you can be recommended to thousands of people and see nothing in your analytics. That gap annoys finance teams, though it is real rather than a fault in the measurement.
The one honest number in the whole field
Where an assistant does link you, the visit lands in your analytics as a referral from that assistant. It is the only figure here that counts real people instead of sampled runs. We pull it every month and set it beside the tracking data, so you can see both halves of the picture.
The distinction also changes what you do next. Named but never linked usually points at reputation: reviews, mentions on sites the models trust, consistent facts about your business. Linked but rarely named usually points at the writing on your own pages. LLM visibility tracking separates the two, and that separation is most of its value.
What LLM visibility tracking cannot tell you
The honest list, including the promise nobody in this trade can make.
Start with what is missing. LLM visibility tracking cannot tell you how many people saw an answer about you, because no assistant publishes that figure. It cannot tell you what real users typed, only what our sample asked. And it cannot tie a mention to a sale, unless the visitor arrived through a link and then did something you can see.
Nor can anyone promise a mention. Not us, not the tool vendors, not the agency with the confident graph. A model decides what to say at the moment it is asked, and that decision is not for sale. Where a proposal implies otherwise, that is the sentence to ask about.
What it is genuinely good for
It works as a diagnostic, in the same way analytics is a diagnostic. It tells you where you stand, whether that is moving, and which rival keeps turning up in your place. Those three answers are usually enough to decide where next month's effort goes, which is what a board is really asking for.
We will also say when tracking is the wrong purchase. If nobody has yet written a clear answer to a single customer question, measuring the absence every month is an expensive way to confirm it. Do the work first, then track it.
No guarantee, from us or from anyone
We cannot promise that ChatGPT, Gemini, Perplexity or Google's AI answers will ever name your business, and neither can any tool or agency selling you tracking. We report what we find, and we say plainly when nothing is moving.
Where tracking sits beside the rest of the AI work
Four pages, four different jobs, so you can pick the right starting point.
This page is measurement. It answers where you stand today and whether that is changing. The neighbouring pages do other jobs, so here is the split in plain terms.
An AI visibility audit is the one-off version. One run, one written report, no subscription, and a sensible place to start if all you need is the current position. Ongoing LLM visibility tracking makes sense once you are actually changing things and want to know whether the changes worked.
The work that follows the numbers
Improvement lives on AI search optimisation. That is where the writing, the schema markup and the crawler fixes happen. The retainer version, with a named person on it, sits on our AI SEO agency service. That starts at the same £495 a month, rolling, with no lock-in.
If the vocabulary is new, read generative engine optimisation, explained first. It defines the terms without the acronym soup. Then come back to this page, because LLM visibility tracking is the part that tells you whether any of the rest is working.
Related: AI visibility audit: the one-off snapshot · AI search optimisation: the work itself · Generative engine optimisation, explained
Fair questions, straight answers
The things marketing leads ask us before they take this to a board.
What is LLM visibility tracking?
LLM visibility tracking means asking AI assistants the same set of questions on a schedule and recording whether they name your business. It covers ChatGPT, Gemini, Perplexity, Copilot and Google's AI answers. You learn how often you appear, for which questions, how you are described, and which competitors appear in your place.
How is it different from Google rank tracking?
Rank tracking reads a fixed list of results, so a position means something. AI answers are written fresh each time, and two identical questions can produce different wording minutes apart. There is no position to record, only how often you are named across many runs. That makes the numbers directional rather than exact.
Which tools do you use, and could I buy them myself?
We run established tools rather than something we built. Otterly.ai starts at $29 a month, Peec AI at $95, Profound at $99 on annual billing, Semrush's bundled AI toolkit at $139, Ahrefs Brand Radar at $199 and Scrunch AI at $250. You can buy any of them yourself. What we sell is the reading of the data and the work that follows.
More questions about LLM visibility tracking
How much does it cost?
Our work starts at £495 a month for Starter, rolling monthly, with no lock-in. The tool subscription sits on top at whatever the vendor charges, and you can hold that licence in your own name if you would rather own the data. Larger programmes are scoped after we have seen the site, and we will not invent a figure before then.
Can you guarantee ChatGPT or Gemini will mention my business?
No, and nobody can. A model decides what to say at the moment somebody asks, and no agency or tool controls that. We can improve the odds by making your site easy to reach, easy to quote and consistent about the facts. We cannot promise the outcome, and we will not pretend otherwise.
How often should we check?
Monthly suits most businesses, and monthly is how we sell LLM visibility tracking. It matches how quickly answers and model versions move, and it gives changes time to show. Weekly runs cost more and usually show noise rather than progress. We run more often only around a launch or a rebrand, when answers genuinely shift week to week.
Find out what AI says about your business
Send us your web address and three questions your customers ask. We will run them across the main assistants and tell you what came back, in plain English, with no promise we cannot keep. If LLM visibility tracking is the wrong purchase for you right now, we will say that too.
Raoul (Alex) Müller
Email: office@digitalmovement.uk
Phone: +44 7865 064463