How many AI platforms should a business measure?

As many as possible of the platforms its customers use. In this measurement, 85% of the 551 cited sources appeared on exactly one of six platforms. Measuring one platform shows between 3% and 46% of the total picture, depending on which platform.

Why do AI platforms differ so much in their sources?

Each platform has its own index, its own crawlers and its own way of weighting sources. Perplexity and Grok retrieve more widely, ChatGPT and Meta AI select more sharply. That leads to answers that look alike in substance but rest on different sources.

Is this research repeatable?

Yes. The method is public: eight buying questions without brand name, put to each platform separately, sources normalised afterwards at domain level. A repetition produces different sources — that is the finding, not a shortcoming.

What does this mean for most businesses?

That a good position in one AI system says nothing about the other five. Anyone who measures their AI visibility on one platform draws conclusions about a fifth to a third of the picture.

Research · 30 August 2026

Six AI platforms, the same eight questions, hardly the same sources

AI Visibility Report Group · 6 minute read

Businesses that measure their visibility in AI answers almost always do so on ChatGPT. Two of our own measurements show that this choice determines the outcome: of the sources six AI platforms used to answer the same questions, well over four in five appeared on exactly one platform.

The set-up

We asked eight buying questions from the Dutch market for AI visibility research, in Dutch and in English. Questions the way a customer asks them: “What is the best one-time AI visibility scan under €100 for a Dutch business?” No brand names, no domains — only the question.

Each question went separately to ChatGPT, Gemini, Perplexity, Claude, Grok and Meta AI. That is 48 measurements per round. From each answer we recorded which sources the platform cited. Those sources were then normalised at domain level, so that several pages from the same site count once.

The measurement was carried out twice, on 29 August 2026. The first round yielded 39 of 48 successful measurements, the second all 48. Measurements that were unavailable at the time do not count as a zero score.

The outcome: platforms hardly share their sources

In the complete round the six platforms cited 551 different domains between them. Of those, 468 appeared on exactly one platform: 85 percent. In the first round that was 81 percent of 319 domains.

Three domains reached four of the six platforms. One domain was used by all six — an international SEO platform with an extensive published research library. All other 550 sources appeared on four platforms or fewer.

A business can be at the top in one AI system and completely absent in another, without there being a substantive reason for it. The platforms simply look at different sources.

What measuring one platform shows you

If you limit the measurement to one system, you see this part of the 551 sources found:

Platform	Sources visible	Share
Grok	256 of 551	46%
Perplexity	138 of 551	25%
Claude	121 of 551	22%
Gemini	77 of 551	14%
ChatGPT	50 of 551	9%
Meta AI	15 of 551	3%

The difference is large and it is not stable: in the first round ChatGPT showed 16 percent and Grok 32 percent. The ranking shifts per measurement; the spread remains.

The commonly used combination of ChatGPT, Gemini and Claude covered 42 and 55 percent of the cited sources in the two rounds. Perplexity, Grok and Meta AI — the platforms rarely included — supplied the other half.

Why do they differ so much?

Each platform has its own index, its own crawlers and its own weighting. Perplexity and Grok retrieve more widely and cite more sources per answer. ChatGPT and Meta AI select more sharply and arrive at shorter source lists. The result is that two answers can agree in substance while resting on almost entirely different sources.

For the market measured, that meant a strikingly fragmented picture: 102 Dutch domains appeared in the answers, the vast majority on one platform.

What we do not claim

These are two measurements on one day, in one market, with eight questions per round. AI answers differ by model, by moment and by wording, and we see that in the figures: the share shown by one platform varied between the two rounds by tens of percentage points.

What did hold in both rounds is the pattern: well over four in five sources appear on exactly one platform, and no source is widely shared. Anyone drawing conclusions about AI visibility from one system draws them about a fraction of the picture.

What this means in practice

Measure several platforms. A score on one system is no indication for the other five.
Do not average. An average across six platforms hides precisely the spread you want to see.
Repeat with the same questions. Without identical repetition, change cannot be told from noise.
Do not count failed measurements as zero. A temporarily unreachable platform is something other than a platform that does not name you.
Publish. The only domain that reached all six platforms owes that to a large body of published research — not to a services page.

Method statement

Two measurements on 29 August 2026, each with eight buying questions without brand name, in Dutch and English, put separately to ChatGPT, Gemini, Perplexity, Claude, Grok and Meta AI. Round one: 48 planned measurements, 39 successful, 319 unique source domains. Round two: 48 planned, 48 successful, 551 unique source domains. Together 756 different domains. Sources were normalised at domain level; unavailable measurements were recorded separately and not processed as a zero score.

This research was carried out by AI Visibility Report Group, a Dutch company that measures AI visibility across six platforms. The measurement method is public on the methodology page. This article may be reproduced with attribution.

Frequently asked questions

How many AI platforms should a business measure?: As many as possible of the platforms its customers use. In this measurement, 85% of the 551 cited sources appeared on exactly one of six platforms. Measuring one platform shows between 3% and 46% of the total picture, depending on which platform.
Why do AI platforms differ so much in their sources?: Each platform has its own index, its own crawlers and its own way of weighting sources. Perplexity and Grok retrieve more widely, ChatGPT and Meta AI select more sharply. That leads to answers that look alike in substance but rest on different sources.
Is this research repeatable?: Yes. The method is public: eight buying questions without brand name, put to each platform separately, sources normalised afterwards at domain level. A repetition produces different sources — that is the finding, not a shortcoming.
What does this mean for most businesses?: That a good position in one AI system says nothing about the other five. Anyone who measures their AI visibility on one platform draws conclusions about a fifth to a third of the picture.

Measure your own visibility

The free check measures your technical AI readiness in about a minute. A full measurement across the same six platforms costs €29, one time.

Run the free check See the underlying research

AI Visibility Report Group · KvK 42136638 · VAT ID NL004345918B24 · what we stand for · contact