Skip to content
SellerChat
Knowledge base

What the published figures actually say

Marketing pages quote AI statistics constantly and rarely say where they came from. This page does the opposite: two datasets, both from a public statistical office, both stored as a dated snapshot, with what they do not tell you written next to what they do.

None of these figures come from SellerChat, and none of them is evidence that this product works.

A person studying printed charts on paper, pen in hand

The figures on this page are a stored snapshot, not a live feed. They were taken on the date printed under each chart, and they have not been re-checked since. Read them as history rather than as the current state of anything.

European Union, enterprises with ten or more employees

Share of EU enterprises using at least one AI technology

Four survey rounds. The 2022 round did not take place, which is why the series has four points and not five.

What this says: Adoption in this measure roughly doubled between 2023 and 2025, having barely moved in the two years before that.

What this does not say: It does not say what the technologies are used for, whether they stayed in use, or whether any of them face a customer. A company counted here may be using AI for stock forecasting and nothing else.

The same data as a table
Share of EU enterprises using at least one AI technology. Unit: % of enterprises
Series 2021 2023 2024 2025
Use at least one AI technology 7.65% 8.06% 13.48% 19.95%

Population: Enterprises with ten or more employees, all activities except agriculture, forestry, fishing, mining and the financial sector. Indicator E_AI_TANY, use of at least one of the listed AI technologies. The list of technologies in the questionnaire changed between rounds, so the levels of different years are comparable only approximately, while the direction of change is stable.

Source: Eurostat, dataset isoc_eb_ai. Reference period: 2021, 2023, 2024, 2025. Geography: European Union, 27 countries. Snapshot taken: 2026-07-26. Dataset last updated by the source: 2026-06-15

European Union, by enterprise size

AI use by company size

The same indicator, split by how many people the enterprise employs, at the start and the end of the observed period.

  • 2021
  • 2025

What this says: Every size group grew, and the gap between the smallest and the largest stayed roughly the same in proportion.

What this does not say: It says nothing about which tools were adopted or what they cost, and a small company in this data is still one with at least ten employees.

The same data as a table
AI use by company size. Unit: % of enterprises in that size group
Series 10 to 49 employees 50 to 249 employees 250 employees or more
2021 6.12% 12.55% 28.41%
2025 17% 30.36% 55.03%

Population: Enterprises of the corresponding size, all activities except agriculture, forestry, fishing, mining and the financial sector. Indicator E_AI_TANY. Each size group has its own denominator: this is the share within the group, not the share of all EU enterprises. Intermediate aggregates were left out so that the groups do not overlap.

Source: Eurostat, dataset isoc_eb_ai. Reference period: 2021 and 2025. Geography: European Union, 27 countries. Snapshot taken: 2026-07-26. Dataset last updated by the source: 2026-06-15

What is missing here

The figures we would like and do not have

Being clear about the holes is more useful than filling them with something that sounds close enough.

Nothing worldwide

Both datasets cover the European Union. There is no comparable statistical series here for other markets, so nothing on this page should be read as a global figure.

Nothing about chat specifically

The indicator counts any of a list of AI technologies. It does not isolate customer-facing chat, so it cannot be used to argue anything about this category.

Nothing about outcomes

Adoption is not results. No dataset here says whether the companies counted got anything out of it.

Nothing of our own

We do not publish figures from our own users on this site. When there is something checkable to say, it will be said with its method attached.

How to read a figure like this

The habits that keep a statistic from turning into a claim it cannot support.

Check the denominator

A share is meaningless until you know what it is a share of. Here it is enterprises above a size threshold, in one region.

Check the period

The most recent point in a survey series is usually a year or more old by the time anybody quotes it.

Check what changed in the question

When the questionnaire changes between rounds, the direction of a series survives and the exact levels do not.

Do not let it become a promise

Adoption rising elsewhere is not a statement about your website. It is context for a decision, not evidence for one.

Numbers are context, your own conversations are evidence

The most useful data about your website is what your visitors actually ask.