Do you need an AI visibility tool?

Somebody will try to sell you an AI visibility tool this year, if they have not already.

The category is real and it is moving quickly. In May a UK platform called Searchable raised $14 million to track brand visibility across ten AI search engines, with American Express, KPMG and Siemens named among its clients. In the same month Berlin-based Peec AI was reported to have reached $10 million in annualised revenue, roughly doubling in six months. At the entry end of the market, tools like GEOAnalyzer Pro launched at $99 a month. The reason for the rush is not mysterious. Google now reports more than 2.5 billion monthly users for AI Overviews and over a billion for AI Mode, so whether a business appears inside those answers has become worth money to somebody.

Here is the part worth sitting with if you run a hotel, a campsite, a restaurant or an attraction. Almost none of that money was raised to help you. It was raised to help marketing teams at large brands. These tools are built for people whose job title contains the word "search", who report to somebody on a monthly cycle, and who have a team to hand the findings to.

That does not make them bad tools. It makes them a different product for a different reader. The useful thing is knowing which product you actually need.

What a visibility tool measures

Most of the category does one thing, and does it well: monitoring.

You give it a list of prompts a customer might type. "Family hotels in north Devon." "Dog friendly cottages near Bude." It runs those prompts across ChatGPT, Google's AI answers, Perplexity and others, repeatedly, then reports how often you were mentioned, how you were described, and who was mentioned instead of you. Some call the result a share of voice. Some call it a visibility score.

That is genuinely useful information in one specific situation: when somebody's job is to act on it. A score tells you where you stand. It does not tell you what to change. The gap between those two things is the whole of the work, and at a large brand that gap is filled by an agency or an in-house team. If nobody in your business is going to be handed a task off the back of the report, the report is where it stops.

There is a second thing worth knowing before you pay for a score, which is how much the underlying answers move on their own. Rand Fishkin ran 2,961 brand-recommendation prompts across ChatGPT, Claude and Google AI. Fewer than one run in a hundred returned an identical list of brands, and fewer than one in a thousand returned the same list in the same order. Ask the same question twice in a week and you can get different businesses, with nobody having changed anything.

The monitoring tools know this, which is why the serious ones measure how often you appear across many repeated runs rather than where you ranked once. That is a sound method. It is also why a single spot check tells you very little, and why watching a number between a changeover and a delivery will read as a verdict when it is not one.

The question that comes first

Before "how often do I get mentioned" there is a plainer question: can these systems read my website at all?

It is not the same question, and it is the one small operators tend to skip. An AI system can only describe your business from what it can actually retrieve. If your prices sit inside a booking widget, if your best line lives inside a photograph, if your pages arrive empty to anything that is not a browser, then no amount of monitoring will explain the result you are looking at. You will watch a number and never see the cause.

Readiness is also the part you can do something about. Whether ChatGPT mentions you next Tuesday is outside your control, and outside everyone else's. Whether your opening hours exist in text a machine can lift, whether your policies are stated plainly, whether your pages serve real content to a crawler, all of that is inside your control this week.

This is the honest split in the category, and it is worth naming plainly because glassmark sits on one side of it. Monitoring tells you the result. A readiness audit tells you the causes. If you are only ever going to pay for one, pay for the one that produces a list of things to fix.

A short test

If you have never checked what an AI assistant currently says about you, do that before you buy anything at all. Ask three assistants the questions your guests ask. Write down what comes back. Twenty minutes, no cost, and it will teach you more than the first month of any subscription.

If the answers come back thin, wrong, or about somebody else, the next step is finding out what those systems can actually see when they visit your site. That is a readiness question and it has concrete answers.

If you already have somebody doing marketing work every week, with a budget behind them, monitoring earns its place. You have the second half of the loop, which is what makes the first half worth paying for.

If you do not, a dashboard is a subscription that makes you feel behind once a month for a year.

What to ask of anything you pay for

Three questions, whichever side of the split the tool sits on.

Does it name specific findings on specific pages, or does it only give you a number? A score without a location is not something anyone can act on.

Is the output forwardable? Most of this work gets done by a web person who is not in the room. If you cannot send them the report and have them understand what to do, the report has not finished its job.

What does it actually promise? Nobody can guarantee that an AI assistant will recommend you, and the research above is a good reason why. Any tool implying otherwise is selling a certainty this market does not contain. What is honestly available is narrower and more durable: your information, readable, consistent and present, so that when a system does answer, it answers with yours.

Read more

• Searchable raises $14M for AI search visibility (TechStartups)
• Peec AI more than doubled revenue to $10M (TechCrunch)
• Why AI brand recommendations vary between runs (Search Engine Land on Rand Fishkin's study)
• Google I/O 2026: AI Overviews at 2.5 billion monthly users (TechCrunch)