How to Measure AI Visibility Without Expensive Tools: A Manual Approach

You can measure AI visibility perfectly well without spending a single euro on software: define a fixed set of questions your audience asks, enter them manually into ChatGPT, Perplexity and Google AI Overviews, and note in a spreadsheet whether and how your brand appears. Not scalable, but honest and concrete. In this article you will learn how to set up that manual measurement, which columns your spreadsheet needs, how to extract a trend from it and when you are better off switching to a paid tool.

At Customer Impact we are the GEO specialist for B2B in the Benelux. Our stance: start small and measure what generates leads, not what looks good on a dashboard. Want guidance on this? Take a look at our generative engine optimization (GEO).

Why measure AI visibility manually?

Because you do not need software to know whether an AI model mentions your brand, only discipline. AI answers usually happen without a click: someone asks ChatGPT a question, gets an answer and closes the window. Your Google Analytics never sees that interaction, even though something important just happened, namely that your brand was or was not recommended. A paid tool automates the tracking of that, but the underlying measurement you can do yourself.

For a small B2B team, the manual approach is the right starting point for three reasons. You do not have to free up budget before you know whether there is anything to measure. By running the prompts yourself, you learn how the models see your market, which no dashboard shows you quite so rawly. And you keep the scope small, which is exactly what a first baseline should be. The manual route is therefore not a second-rate solution, but a logical first step.

Which prompts should you test?

Test the questions with buying intent in your niche, not general terms you will never show up for anyway. The difference between a useful and a useless measurement lies entirely in your choice of prompts. “What is marketing” produces noise. “Which B2B marketing agency in Flanders specialises in lead generation” is a question a prospect genuinely wants answered.

Build a list of ten to twenty prompts across a few categories:

• Category prompts: “best [service] for [audience]” or “which agencies do [specialisation]”. Here you want to be on the list.
• Problem prompts: the question the way a customer phrases their problem, before they even know your solution.
• Comparison prompts: “[your brand] vs [competitor]” or “alternatives to [competitor]”.
• Brand prompts: “what does [your brand] do” to check whether the model describes you correctly.

Write down the exact wording and freeze it. The value of this method lies in repeatability, so the prompt you use in month one must be word for word the same in month three. Following ten prompts properly is worth more than a hundred scattered attempts. The underlying indicators you can derive from this are covered in the five core indicators of AI visibility.

How do you build the measurement spreadsheet?

A simple spreadsheet with one row per prompt and per platform is enough. You do not need to build anything complicated; a shared sheet in Google Sheets or Excel does the job. Give every measurement these columns:

• Date of the measurement.
• Prompt (the exact text).
• Platform (ChatGPT, Perplexity, Google AI Overviews, Gemini).
• Mentioned? yes or no.
• Position or context: are you named as the first option, somewhere in the middle of the pack, or only in passing?
• Cited source: which of your pages was quoted, or an external page about you.
• Competitors that show up in the same answer.
• Sentiment or accuracy: is what the model says about you correct, or is something wrong?

Those last columns make the difference between counting and understanding. “Mentioned: yes” is nice, but if the model positions you as the most expensive option or gives wrong information, you immediately know where to adjust your content. Also watch which sources do get cited, because those are the pages you would want to be.

How do you run the measurement consistently?

Run every prompt the same way, in a clean session, without earlier chats colouring the answer. A few practical rules keep your numbers reliable:

• Always start a new chat or use an incognito window, so memory and history do not influence your result.
• Keep your location consistent. AI Overviews and other engines factor in location, so measure from the same region as your audience.
• Test every prompt on every platform. Citation logic differs strongly per engine: a brand that is strong in ChatGPT can be invisible in Perplexity, because the platforms quote different sources. That is why you measure broadly.
• Note the cited links too. With Perplexity and Google AI Overviews, sources are explicitly listed, which immediately shows you which content the model trusts.

Want to dig deeper per platform into what makes you visible? Read how to get found in ChatGPT and get found in Perplexity. They really are different rules of the game.

How do you turn it into a usable number?

Calculate a simple share: on how many of your prompts does your brand appear, divided by the total, per platform. That gives you a rough visibility score you can compare month after month. If you appear in four out of ten category prompts in ChatGPT, that is forty percent, and you note it down.

Do the same for your most important competitor, with exactly the same prompts. That gives you a simple share-of-voice picture: who gets mentioned more often on the questions that matter to your business. The absolute numbers matter less than the direction. If your share rises after you have reworked your content, your GEO approach is working. If it stagnates, you know something else is needed.

Important and honest: this is an indication, not exact science. AI models do not always give the same answer to the same question, so do not count small fluctuations as a trend. Look at the pattern across several months, not at a single measurement. And always tie visibility back to what counts: leads and revenue, not a number for its own sake.

When do you switch to a paid tool?

Switch as soon as the manual measurement costs more time than it delivers. The method above is strong as a baseline and for a small number of prompts, but it does not scale. A few signals that you are hitting the limit:

• You are tracking multiple clients or brands and the monthly manual work is getting out of hand.
• You want daily tracking or an alert when you suddenly drop out, which is not feasible by hand.
• You need sentiment analysis or larger prompt volumes to support your decisions.

At that point, automation pays for itself. Which categories of tools exist and how to choose without overpaying, you can read in Measuring AI visibility: tools and methods for B2B. Our level-headed line stays the same: for most Belgian B2B brands, an expensive enterprise tool is overkill, and a spreadsheet gets you further than you think.

The short summary

Measuring AI visibility without tools comes down to three things: a fixed set of buying prompts, a simple spreadsheet and the discipline to measure the same way every month. Do not just count whether you are mentioned, but also how and with which sources, and translate everything back to the question of whether it generates leads. Only move to a paid tool once the manual work becomes too heavy. Want to set this up together and find out which prompts matter for your market? Book your free intake