How do you measure whether GEO pays off?

Anyone who puts money and time into visibility through AI wants to know what it delivers. That turns out to be harder than with traditional search engines. There is no fixed ranking to track. An AI answer also differs from one user to the next. According to an eMarketer survey from January 2026, summarised in this report by The Digital Bloom, only 23% of marketers invest in measuring GEO. Yet measuring is possible, with a simple and repeatable method.

Start with a baseline built on real customer questions

Every measurement starts with the questions your customers ask. Generic search terms such as "accountant" say little. A customer is more likely to type into ChatGPT: "which accountant in the Mechelen area works a lot with construction companies?". Draw up a list of twenty to thirty such questions, spread across:

• questions asking for a provider in your region
• questions about a problem you solve, before the customer is even thinking of a solution
• comparison questions between you and a competitor
• questions about your name, to see what AI services know about your business

Put those questions to ChatGPT, Perplexity, Gemini and Google with AI Overviews. For each question, note whether your business is mentioned and in which position. Also note whether the information is correct and which sources the service cites. That is your baseline: the starting point against which you set every later measurement.

Two practical rules make the measurement more reliable. Always use the same wording, so you compare like with like. Ask the questions in Dutch and include your region, because those are the questions your customers type.

Repeat the measurement every month

A single measurement is a snapshot. AI services constantly adjust their models and sources, so a mention today can be gone next month. Repeat the same list of questions in the same way every month. After a few months, a trend becomes visible.

Track a few key figures each month:

• the share of questions in which your business is mentioned
• the share of answers in which your details are correct
• which competitors are mentioned more often than you
• which external sites appear as sources, because that is where your opportunities for new mentions lie

Recognise AI traffic in your analytics

The second line of measurement lies in your own figures. Visitors who land on your site through an AI service leave a trace in the referring domains of your analytics tool. Look for domains such as chatgpt.com, perplexity.ai and gemini.google.com. In most packages you can create a separate channel group that bundles this traffic, so you can see it evolve month after month.

For ChatGPT this is especially easy. OpenAI states in its FAQ for publishers that ChatGPT automatically adds the parameter utm_source=chatgpt.com to links in its search answers. That traffic is therefore recognisable in any common analytics tool.

Google AI Overviews is harder. According to Google itself, in its explanation of AI features, traffic from AI Overviews and AI Mode is included in the total search traffic in Search Console. You cannot filter it out separately there. Here the monthly list of questions remains your best measuring instrument.

Keep an eye on volume. For many SMEs, traffic from AI services is still small today compared with Google. So look mainly at what those visitors do: how long they stay, which pages they view and whether they get in touch.

What a score out of 100 does and does not tell you

Many analyses, ours included, summarise the situation in a score out of 100. Such a score combines technical points, such as crawler access and structured data, with the quality of your texts and your presence on other sites. It is useful for quickly seeing where the biggest gap lies and for tracking progress over the months.

A score also has limits. It measures the conditions that increase the chance of a mention. Whether AI services actually name you, you only see in the list of questions. A site with a high score can still be missing from the answers in a sector with strong competitors, while a site with a lower score in a niche market is already mentioned often. Read the score together with the mentions and the traffic.

Our monthly rhythm

In our monthly follow-up for SMEs across Flanders, from Ghent, we work to a fixed rhythm:

• Week 1: technical. We check crawler access, structured data, llms.txt and site speed. Whatever is wrong, we fix straight away.
• Weeks 2 and 3: content and mentions. We rewrite or write pages that answer the questions on the list and work on mentions on sites that AI services use as sources.
• Week 4: measure again. We repeat the list of questions, review the AI traffic in your analytics and discuss the results with you.

In the Starter package this follow-up costs 2.000 to 3.000 euros per month. In the Growth package it costs 3.000 to 5.000 euros per month. Both prices are excl. VAT. If you first want to get the basics right, you can start with a one-off optimisation of 500 to 1.500 euros excl. VAT.

What you can do now

• Write down twenty questions your customers really ask and enter them into ChatGPT and Perplexity this week. Save the answers as your baseline.
• In your analytics tool, look for chatgpt.com, perplexity.ai and gemini.google.com among the referring domains.
• Put a fixed day each month in your calendar to ask the same questions again.

Would you like us to run that measurement for you every month and adjust straight away? Take a look at our monthly follow-up. If you want the leads that come from AI traffic followed up automatically in your CRM, our sister agency hetaibureau.be can help you further.