Skip to content
ceeme

How much of the AI market does a measurement cover?

That depends entirely on how many engines you measure, and the gap is wider than most figures suggest. In Belgium, 4.2% of people opening an AI assistant use Claude. Measuring only on Claude therefore measures 4.2% of the market and presents it as “your AI visibility”. The four engines ceeme measures by default — ChatGPT, Gemini, Claude and Perplexity — cover 93.0% together. What falls outside that is almost entirely Copilot, at 7.0%, and we do not measure it today.

Measure your own company → Rather talk it through first

In short

4.2% on one engine
That is how much of the Belgian AI market you see when you only query Claude. The resulting figure can be perfectly correct and say almost nothing.
93.0% on four
The four core engines together. Not a hundred, and that difference belongs in the report rather than being left out.
The gap has a name
7.0% is Copilot, and we do not measure it. A coverage figure without the gap beside it is advertising.
Different per market
Netherlands 95.1%, France 95.9%, Germany 95.6%. A single coverage figure for “Europe” does not exist.

What does the study say?

That a visibility figure without a coverage figure is incomplete, and by how much. The calculation itself is simple: take each AI assistant's market share in the client's market, add up the shares of the engines you actually queried, and set that sum beside the figure. For Belgium that comes to 93.0% with four engines and 4.2% with Claude alone.

The interesting part is not the sum but what sits underneath it. Two tools can produce a completely different figure for the same company without either one calculating, measuring or reporting anything wrongly — they simply queried a different set of engines. Without a coverage figure that difference is invisible, and then the tool with the cheeriest number is the one that appears to be right.

What was measuredHowWhat the study does not say
Our own calculation, across four markets, with shares from July 2026the shares of the four core engines added up per market and set against total measured AI use. Source of the shares: Statcounter · AI chatbot market sharenothing about where BUYING questions get asked. A share of usage is not a share of purchase decisions, and nobody measures that distinction reliably today — us included
What falls outside the measurement, with name and shareevery engine with a known share that is not in the core, named separately rather than left outthat those engines do not matter. Copilot sits inside Office and is therefore used in places no other assistant reaches; that we do not measure it is a limit of ours and not a verdict on it

Our own calculation. The engines come from the measurement's core list, the shares from Statcounter · AI chatbot market share (July 2026). Every percentage on this page is computed at build time and appears nowhere as text — if a share changes, the page changes with it.

Why does this matter to you?

Because it is the first question you can put to any supplier, and because the answer is usually not ready. Which engines did you query, and what share of my market is that together? Anyone who cannot give a number has an impression rather than a measurement. And anyone who can must also be able to say where the share comes from and which month it is from — a market share from last year is a different number in this field.

And because it differs per market. The same four engines cover 95.1% in the Netherlands, 95.9% in France and 95.6% in Germany. Anyone selling in several countries who gets one coverage figure is getting an average of markets that do not behave the same way.

What does ceeme do with it?

The coverage figure sits in the report, next to the score and not in a footnote. And the engines we do measure but cannot weight — because no share is known for that market — appear in the table as measured-but-not-weighted. Setting them to zero would lower the figure for no reason; giving them an invented percentage would pollute it. The third way is to name them.

The weighting itself is one formula, in one place, and the report and the dashboard both read that same one. That sounds obvious and is not: it is precisely where things once went wrong for us, and that story sits on the other method study next to this one.

Frequently asked questions

So why do you not measure Copilot?

Because we cannot query it today in the same way as the other four, and measuring an engine halfway is worse than not measuring it — you would get a figure that gets compared with the rest while it came about differently. As long as that holds, it sits in the coverage figure as what it is: the largest piece we do not see.

Is a higher coverage figure always better?

No, and that is the trap under this whole piece. Querying more engines costs money and yields little on a small engine; it is a trade-off, not a contest. The point is that the number is THERE, so that you can make that trade-off instead of having to take ours on trust.

Why is there no figure for “Europe”?

Because it does not exist. The shares are measured per country and they diverge; adding them into one European figure produces an average that holds in no country. Sell in three countries and you get three coverage figures — that is less convenient and it is correct.

Read on

Free, no account and no card.