"X% of companies use AI": you read this sentence in press releases, funding applications and presentations, sometimes with 20%, sometimes with 57%. Both figures can be correct, because they come from different surveys asking different questions. On this page you will first find the official figures from Destatis and Eurostat, followed by the most important surveys from Bitkom, ifo, KfW Research and the German Economic Institute (IW), each with its population, sample and period. At the end there is a comparison table, a guide to citing the figures correctly and all values as a CSV.
I put this page together because these numbers are often passed on without their origin. A Bitkom rate for companies with 20 or more employees turns into a statement about "the German economy", a value from the KfW SME Panel into a figure for all businesses. Here, every number comes with whom it describes and when the question was asked.
As of: All sources retrieved on 08.10.2026. Official figures for the reference year 2025 (Destatis tables of 24.11.2025, Eurostat database as of 15.06.2026), surveys up to the Bitkom study of 14.09.2026. Destatis published the AI results for 2025 on 24.11.2025, so following that rhythm the figures for 2026 can be expected at the end of November 2026. Eurostat plans the next update of its overview article for December 2026. I will add both afterwards.
- of companies used AI in 2025
- 26%
- 10 or more employees, 2024: 20%
- of small companies
- 23%
- 10 to 49 employees
- of large companies
- 57%
- 250 or more employees
- EU-27 average 2025
- 19.95%
- Germany according to Eurostat: 25.97%
- Germany among the 27 EU member states
- 8th place
- top: Denmark with 42.03%
- of AI-using companies use AI for marketing or sales
- 35.05%
- most common purpose in Germany
The official figure: Destatis and the ICT survey
The most robust figure on AI adoption in Germany comes from the survey on the use of information and communication technologies in enterprises (ICT survey) run by the Federal Statistical Office (Destatis). It follows the same rules in all EU member states, so the German values are directly comparable with those of other countries. The AI questions were asked for the reference years 2021, 2023, 2024 and 2025, not for 2022.
For any citation it matters what the survey covers. According to the quality report 2025 (German), these are companies with at least 10 persons employed in manufacturing, energy and water supply, construction, trade, transport and hospitality, information and communication, real estate and professional and technical services, plus repair of computers and communication equipment. Not included are, among others, agriculture, banks and insurers, health and social work and education. Medical practices, care services or banks therefore do not appear in this rate, while tax advisory firms, as professional service providers, do.
Participation is voluntary. In 2025, Destatis contacted about 98,800 companies to reach the net sample of 20,000 companies set out in law, and the response rate was 24.6%. The questionnaire is online, and the reference period is January of the survey year. The results are extrapolated to all companies in the population, and according to Destatis the relative standard error of the key indicators is below 5%.
A company counts as "using AI" if it uses at least one of eight specified technologies: text mining, speech recognition, natural language generation (including program code), generation of images, video or audio, image recognition, machine learning for data analysis, automation of workflows or decision making, and autonomous movement of machines.
AI adoption by company size
In 2025, 26% of companies with 10 or more employees used AI, compared with 20% a year earlier. The larger the company, the more common it is: 23% of small companies with 10 to 49 employees used AI, 36% of medium-sized companies with 50 to 249 employees and 57% of large companies with 250 or more employees. Figure 1 shows both years side by side.
- 2024
- 2025
- Total2024: 20%2025: 26%
- 10 to 49 employees2024: 17%2025: 23%
- 50 to 249 employees2024: 28%2025: 36%
- 250 or more employees2024: 48%2025: 57%
Source: Destatis, ICT usage in enterprises 2024 and 2025
Share of all companies with 10 or more employees. 2024 legal units, 2025 enterprises under the EU enterprise definition, so the year-on-year comparison is limited.
Data as table
| Company size | 2024 | 2025 |
|---|---|---|
| Total | 20% | 26% |
| 10 to 49 employees | 17% | 23% |
| 50 to 249 employees | 28% | 36% |
| 250 or more employees | 48% | 57% |
Source: Destatis, table on AI use by size class 2025 (as of 24.11.2025) and Destatis, press release no. 444 of 25.11.2024 (German) for 2024. The press release also gives 12% for 2023 and 11% for 2021.
Two points limit the year-on-year comparison. First, Destatis changed the unit of presentation in 2025: up to 2024 it used legal units, since 2025 it uses enterprises under the EU statistical units regulation, which can consist of several legal units. Second, the generation of images, video and audio was added as an eighth technology in 2025. Part of the increase may therefore come from counting more applications.
Which AI technologies companies use
Among companies that use AI, text mining and the generation of images, video or audio are the most widespread at 52% each. They are followed by speech recognition (42%) and natural language generation, which includes chatbots and program code (35%). Autonomous machines hardly play a role at 6%.
- Text mining52%
- Image, video, audio generation52%
- Speech recognition42%
- Natural language generation35%
- Automation of workflows27%
- Machine learning23%
- Image recognition17%
- Autonomous movement of machines6%
Source: Destatis, ICT usage in enterprises 2025
Share of companies that use AI. Multiple answers possible. Image, video and audio generation surveyed for the first time in 2025.
Data as table
| Technology | Values in % |
|---|---|
| Text mining | 52% |
| Image, video, audio generation | 52% |
| Speech recognition | 42% |
| Natural language generation | 35% |
| Automation of workflows | 27% |
| Machine learning | 23% |
| Image recognition | 17% |
| Autonomous movement of machines | 6% |
By size class, the differences show mainly in the more demanding methods. Machine learning is used by 21% of small AI users but by 39% of large ones:
| Technology (share of AI users) | Total | 10 to 49 | 50 to 249 | 250 or more |
|---|---|---|---|---|
| Text mining | 52% | 51% | 53% | 57% |
| Generation of images, video, audio | 52% | 51% | 55% | 57% |
| Speech recognition | 42% | 41% | 43% | 50% |
| Natural language generation, incl. program code | 35% | 32% | 40% | 50% |
| Automation of workflows or decision making | 27% | 26% | 25% | 37% |
| Machine learning | 23% | 21% | 25% | 39% |
| Image recognition | 17% | 16% | 17% | 24% |
| Autonomous movement of machines | 6% | 5% | 6% | 12% |
Table scrolls sideways
Source: Destatis, table on AI use by size class 2025, as of 24.11.2025.
Why companies do without AI
Destatis also asks companies without AI whether they have ever considered using it. In 2025, this applied to 20% of non-users (2024: 18%), to 53% of large companies, 32% of medium-sized and 18% of small ones. This group then gives its reasons, and these have shifted: lack of knowledge remains at the top with 72%, while legal uncertainty (62%) and data protection (60%) have risen markedly. Only just under one in three of these companies names excessive costs.
- 2024
- 2025
- Lack of knowledge2024: 71%2025: 72%
- Unclear legal consequences2024: 58%2025: 62%
- Data protection and privacy2024: 53%2025: 60%
- Incompatible with systems2024: 44%2025: 45%
- Data missing or poor2024: 45%2025: 44%
- Costs too high2024: 28%2025: 32%
- Ethical concerns2024: 23%2025: 23%
- AI not useful2024: 21%2025: 18%
Source: Destatis, ICT usage in enterprises 2024 and 2025
Share of companies that do not use AI but have considered it. Multiple answers possible.
Data as table
| Reason | 2024 | 2025 |
|---|---|---|
| Lack of knowledge | 71% | 72% |
| Unclear legal consequences | 58% | 62% |
| Data protection and privacy | 53% | 60% |
| Incompatible with systems | 44% | 45% |
| Data missing or poor | 45% | 44% |
| Costs too high | 28% | 32% |
| Ethical concerns | 23% | 23% |
| AI not useful | 21% | 18% |
Source: Destatis, table on reasons against using AI 2025 (German, as of 24.11.2025) and press release no. 444 (German) for 2024. The Destatis table labels the share of companies that have considered AI as a share "of all companies". The Eurostat data show that the 20% refers to companies without AI (20.27%), while relative to all companies it is 14.95%.
What the official figure does not measure
The 26% is a lower bound for a specific part of the economy, not a statement about all businesses. Micro businesses with fewer than 10 employees are missing entirely, as are the sectors named above. What is counted is also what the company itself reports. If employees use a chatbot privately for work tasks without management knowing, it does not appear in any official questionnaire. And because the reference period is January 2025, the figure reflects adoption at the beginning of 2025, not the situation today.
Eurostat: Germany in the EU comparison
Eurostat publishes the same survey for all EU member states, with two decimal places and from the reference year 2021. The German values are therefore identical to those of Destatis, just less heavily rounded: 25.97% instead of 26%. Eurostat describes the coverage as all activities except agriculture, mining and the financial sector, more precisely sections C to J, L to N and group 95.1, in each case companies with 10 or more employees and self-employed persons. According to Eurostat (Statistics Explained), 157,000 of the 1.53 million companies in the EU were surveyed in 2025.
Time series 2021 to 2025
In Germany the share rose from 10.56% (2021) via 11.55% (2023) and 19.75% (2024) to 25.97% in 2025. The EU average grew from 7.65% to 19.95% over the same period. The gap between Germany and the EU was 6.02 percentage points in 2025. Most recently the EU grew slightly faster: up 6.47 percentage points on 2024, Germany up 6.22.
- Germany
- EU-27
Source: Eurostat, dataset isoc_eb_ai, as of 15.06.2026
Companies with 10 or more employees excluding agriculture, mining and the financial sector. 2022 not surveyed. 2025 including the new category image, video and audio generation.
Data as table
| Year | Germany | EU-27 |
|---|---|---|
| 2021 | 10.56% | 7.65% |
| 2023 | 11.55% | 8.06% |
| 2024 | 19.75% | 13.48% |
| 2025 | 25.97% | 19.95% |
Source: Eurostat, dataset isoc_eb_ai, indicator "at least one AI technology", as of 15.06.2026. The line connects 2021 directly with 2023, there is no value for 2022.
Germany in the EU ranking 2025
At 25.97%, Germany ranks 8th among the 27 EU member states. Denmark leads with 42.03%, followed by Finland and Sweden. Austria is one place ahead of Germany at 29.95%, while France (18.16%) and Italy (16.4%) are below the EU average. Companies in Romania use AI least often (5.21%). Eurostat notes a break in the time series for the Netherlands in 2025.
- Denmark42.03%
- Finland37.82%
- Sweden35.04%
- Belgium34.54%
- Luxembourg33.61%
- Netherlands33.21%
- Austria29.95%
- Germany25.97%
- Estonia23.40%
- Slovenia21.61%
- Malta21.51%
- Lithuania21.30%
- Spain20.27%
- EU-2719.95%
- Ireland19.64%
- France18.16%
- Slovakia18.00%
- Czechia17.60%
- Italy16.40%
- Croatia15.19%
- Latvia12.21%
- Portugal11.54%
- Hungary10.37%
- Cyprus9.27%
- Greece8.93%
- Bulgaria8.55%
- Poland8.36%
- Romania5.21%
Source: Eurostat, dataset isoc_eb_ai, as of 15.06.2026
Share of companies with 10 or more employees using at least one AI technology. Netherlands with a break in the time series in 2025.
Data as table
| Country | Values in % |
|---|---|
| Denmark | 42.03% |
| Finland | 37.82% |
| Sweden | 35.04% |
| Belgium | 34.54% |
| Luxembourg | 33.61% |
| Netherlands | 33.21% |
| Austria | 29.95% |
| Germany | 25.97% |
| Estonia | 23.40% |
| Slovenia | 21.61% |
| Malta | 21.51% |
| Lithuania | 21.30% |
| Spain | 20.27% |
| EU-27 | 19.95% |
| Ireland | 19.64% |
| France | 18.16% |
| Slovakia | 18.00% |
| Czechia | 17.60% |
| Italy | 16.40% |
| Croatia | 15.19% |
| Latvia | 12.21% |
| Portugal | 11.54% |
| Hungary | 10.37% |
| Cyprus | 9.27% |
| Greece | 8.93% |
| Bulgaria | 8.55% |
| Poland | 8.36% |
| Romania | 5.21% |
Source: Eurostat, dataset isoc_eb_ai, as of 15.06.2026. The order at the top and bottom is also confirmed by the Eurostat news item of 11.12.2025. Germany's rank is my own count among the 27 EU member states.
What companies use AI for
Eurostat asks AI users what they use AI for. In Germany, marketing or sales leads at 35.05%, as in the EU average (34.7%). Germany is clearly above the EU value for accounting, controlling and finance (27.43% against 23.11%) and clearly below it for research and innovation (12.81% against 19.05%). Worth noting: only 68.98% of German AI users name any of the seven purposes surveyed at all, compared with 73.39% in the EU average.
- Germany
- EU-27
- Marketing or salesGermany: 35.05%EU-27: 34.70%
- Administration or managementGermany: 27.64%EU-27: 31.04%
- Accounting and financeGermany: 27.43%EU-27: 23.11%
- Production processesGermany: 21.12%EU-27: 20.76%
- ICT securityGermany: 20.63%EU-27: 19.77%
- Research and innovationGermany: 12.81%EU-27: 19.05%
- LogisticsGermany: 4.57%EU-27: 6.08%
Source: Eurostat, dataset isoc_eb_ai, as of 15.06.2026
Share of companies that use AI. Multiple answers possible.
Data as table
| Purpose | Germany | EU-27 |
|---|---|---|
| Marketing or sales | 35.05% | 34.70% |
| Administration or management | 27.64% | 31.04% |
| Accounting and finance | 27.43% | 23.11% |
| Production processes | 21.12% | 20.76% |
| ICT security | 20.63% | 19.77% |
| Research and innovation | 12.81% | 19.05% |
| Logistics | 4.57% | 6.08% |
Source: Eurostat, dataset isoc_eb_ai, as of 15.06.2026. The Destatis press release for 2024 also puts marketing or sales first (33%), followed by production or service processes (25%).
AI adoption by sector
The spread between sectors is larger than between size classes. In information and communication, three in four companies used AI in 2025 (75.38%), in transport only 12.22%. In almost all sectors Germany is above the EU average.
| Sector (NACE Rev. 2 section) | Germany 2024 | Germany 2025 | EU-27 2025 |
|---|---|---|---|
| All sectors covered | 19.75% | 25.97% | 19.95% |
| Manufacturing (C) | 16.11% | 24.38% | 17.27% |
| Energy and water supply, waste management (D, E) | 17.55% | 22.97% | 18.16% |
| Construction (F) | 10.01% | 14.28% | 10.79% |
| Trade (G) | 16.29% | 25.46% | 18.62% |
| of which wholesale (46) | 21.67% | 32.36% | 23.42% |
| of which retail (47) | 14.46% | 23.42% | 15.46% |
| Transport and storage (H) | 11.48% | 12.22% | 11.15% |
| Accommodation and food services (I) | 8.26% | 14.63% | 11.98% |
| Information and communication (J) | 60.52% | 75.38% | 62.52% |
| Real estate (L) | 20.98% | 29.06% | 24.76% |
| Professional, scientific and technical activities (M) | 42.29% | 48.3% | 40.43% |
| of which advertising and market research, other professional activities, veterinary (73 to 75) | 49.84% | 57.24% | 48.02% |
| Administrative and support services (N) | 18.91% | 22.08% | 19.86% |
Table scrolls sideways
Source: Eurostat, dataset isoc_eb_ain2, companies with 10 or more employees, as of 15.06.2026.
Association and institute surveys: what they measure
Alongside official statistics there are surveys by associations and research institutes. They are often more recent and ask about things Destatis does not cover, such as budgets or AI agents. In exchange they work with smaller samples and their own definitions. I only include values I have checked in the institute's original publication. The surveys do not belong in a shared time series with Destatis or with each other.
Bitkom: 57% of companies with 20 or more employees
The German digital association Bitkom has Bitkom Research survey companies with 20 or more employees by phone every year. For the 2026 study, 603 companies were interviewed in calendar weeks 28 to 33 (July to mid-August 2026), and the target respondent was the person responsible for AI or digital technologies. Bitkom weights the data on the basis of the turnover tax statistics of the Federal Statistical Office and states a statistical margin of error of plus/minus 4 percent for the total sample. According to the results slide, the question was (my translation): "To what extent does your company use AI, or plan or discuss its use?" No list of technologies is given there.
- use AI
- 57%
- 2025: 36%, 2024: 20%
- plan or discuss using it
- 38%
- base: all 603 companies
- AI is not an issue
- 4%
- 2025: 17%, 2024: 41%
- of AI users do not exhaust its potential by their own assessment
- 87%
- 28% rather little, 59% not at all
Source: Bitkom, press release "Erstmals nutzt die Mehrheit der Unternehmen KI" (German), 14.09.2026
Among the 343 AI-using companies, AI is most often used in customer contact (72%, 2025: 88%) and in marketing and communication (54%, 2025: 57%). Internal knowledge management (22% after 11%) and sales (14% after 5%) have grown strongly. AI agents are used by 11% of companies that use, plan or discuss AI, and another 29% are planning them. Companies without AI mainly cite a lack of technical know-how (85%), legal hurdles (66%) and costs that are hard to calculate (63%). 70% of all companies now train their staff in using AI.
Source: Bitkom press release of 14.09.2026 and results presentation (both German). The previous year's values for the areas of use come from the same presentation, and the study report 2025 describes the previous survey of 604 companies.
ifo Business Survey: 54.5% in May 2026
The ifo Institute has asked the participants of its business survey about AI every year since 2023. In May 2026, 54.5% of the companies surveyed used AI in their business processes, after 40.9% in May 2025. Another 16% plan to use it and 21.6% are discussing it. By sector, industry leads at 58.7%, ahead of services (56.2%) and trade (around 45%). In main construction, the share rose from 7.1% to 39.8% within three years. Large companies use AI at a rate of 67.2%, small ones at 51.2% and medium-sized ones at 47.2%. How ifo defines the size classes is not stated on the page.
Important for interpretation: the question does not specify which applications count as AI. It only asks whether AI is already in use, planned, under discussion or not an issue. The authors describe this in their analysis in ifo Schnelldienst 8/2024 (German), in which they also give the earlier values (13.3% in June 2023, 27% in June 2024) and report around 6,000 responses in June 2024. For 2026, ifo does not state a sample size. The ifo survey also includes micro businesses: in 2025, according to ifo, 56% of large companies used AI, 38% of small and medium-sized companies and 31% of micro businesses. Among sectors, advertising and market research stood out in 2025 at 84.3%.
Source: ifo, 05.06.2026 and ifo, 16.06.2025 (both German).
KfW SME Panel: 20% of SMEs (2022 to 2024)
KfW Research measures AI use in the KfW SME Panel (KfW-Mittelstandspanel). According to KfW, it covers all private companies in all sectors with annual turnover of no more than EUR 500 million, by this definition 3.87 million companies in Germany. Unlike Destatis, micro businesses are explicitly included. In the 23rd wave in spring 2025, companies were asked whether they had used AI in the years 2022 to 2024, after an introductory definition with examples. A total of just over 6,800 company responses went into the study, and the values are extrapolated to the number of companies.
Result: 20% of SMEs used AI, according to KfW just under 780,000 companies. In the period 2016 to 2018 it was 4%. By size: 19% for fewer than 5 employees, 20% for 5 to under 10, 29% for 10 to under 50 and 36% for 50 or more employees. Knowledge-based services lead at 28%, construction is last at 8%. A follow-up analysis from July 2026 shows which technologies these are: natural language generation is used by 14% of SMEs, text recognition by 10%, data analytics by 3%.
Source: KfW Research, Fokus Volkswirtschaft no. 533 of 11.02.2026, no. 554 of 29.07.2026 and the description of the KfW SME Panel (all German).
IW Future Panel: 37% in industry and industry-related services
In wave 49 of its IW Future Panel (IW-Zukunftspanel), the German Economic Institute surveyed 1,038 companies from industry and industry-related services between 16.10. and 05.12.2024. Respondents were managing directors, board members or heads of strategy departments. 37.0% of the companies surveyed used AI. By size: 35.6% up to 49 employees, 56.1% for 50 to 249 and 66.3% for 250 or more employees. Business-related services led at 55.4%, with construction (21.9%) and wholesale and logistics (24.1%) at the bottom. The IW points out that the overall value cannot be calculated from the size classes, because only 970 of the 1,038 companies could be assigned to a size.
Source: IW-Report "Wettbewerbsfaktor Künstliche Intelligenz" of 04.07.2025 (German).
All surveys compared
| Source | Population | Sample | Year | Key value |
|---|---|---|---|---|
| Destatis, ICT survey | Companies with 10 or more employees, sections C to J, L to N, group 95.1 | 20,000 net, voluntary, online | 2025 (January) | 26% use at least one of eight AI technologies |
| Eurostat, isoc_eb_ai | as Destatis, all EU member states | 157,000 EU-wide | 2025 | Germany 25.97%, EU-27 19.95% |
| Bitkom Research | Companies with 20 or more employees, whole economy | 603, by phone | 2026 (weeks 28 to 33) | 57% use AI |
| ifo Business Survey | Participants of the ifo Business Survey from industry, services, trade and main construction, incl. micro businesses | not stated | 2026 (May) | 54.5% use AI in business processes |
| KfW SME Panel | Private companies with up to EUR 500 million turnover, incl. micro businesses | just over 6,800 responses | use 2022 to 2024 | 20% used AI |
| IW Future Panel | Industry and industry-related services | 1,038 | 2024 (October to December) | 37.0% use AI |
Table scrolls sideways
Why the figures are so far apart
Between 20% and 57% lies almost a factor of three. That is not a measurement error but the result of five differences that often get lost when the figures are cited.
What counts as AI. Destatis and Eurostat ask about eight specific technologies. KfW provides a definition with examples, ifo and Bitkom simply ask whether AI is used. With an open question, a company is likely to say "yes" more readily if individual employees use a chatbot. On the official list, it would first have to recognise a matching technology.
Which companies are asked. The size threshold shifts the value considerably, because large companies use AI much more often in every source. Bitkom starts at 20 employees, Destatis at 10, while KfW and ifo include micro businesses. Then there is the sector coverage: the IW only asks industry and industry-related services, Destatis leaves out health care and the financial sector.
When the question was asked. The official figure describes January 2025, Bitkom summer 2026. In a phase in which the rate rose by six percentage points from 2024 to 2025 according to Destatis, a year and a half is a long time. KfW even asks retrospectively about the years 2022 to 2024.
Who answers. Bitkom specifically addresses the person responsible for AI or digitalisation, the IW the management or strategy department. At Destatis, the company fills in an online questionnaire without a specific function being prescribed. Those who already deal with AI probably know the applications in their company better.
How results are extrapolated. Destatis and KfW extrapolate to the number of companies, so every small business counts as much as a corporation. Bitkom weights on the basis of the turnover tax statistics. For ifo and IW, no weighting is described in the publications I checked. Precisely because small companies form the majority, the weighting directly affects the overall rate.
The direction is still the same in every source: adoption is growing fast, large companies are ahead, and those who use AI are mostly still at the beginning by their own account.
What this means for small and medium-sized companies
If you run a company with 10 to 49 employees, the right benchmark is the official one: 23% of your size class used at least one AI technology at the beginning of 2025. The Bitkom figure of 57% describes companies with 20 or more employees in summer 2026 and is not suitable for creating pressure along the lines of "everyone else is already further ahead".
More revealing than the rate are the obstacles. Lack of knowledge comes first at Destatis as well as at Bitkom, with costs only following after. In my experience, getting started is therefore less a question of budget than of which process AI actually pays off for. If you still want to work out the running costs, you will find the prices of the major models in my comparison of AI API prices.
Legal uncertainty is the second-largest brake and grew in 2025. Part of it can be cleared up with manageable effort, for example the question of when AI-generated content has to be labelled. I have written a separate article on labelling AI advertising.
And finally the point the Bitkom study shows most clearly: using is not the same as benefiting. 87% of AI-using companies say they do not exhaust its potential. A "we use AI" rate therefore says little about whether a process has become faster, cheaper or better as a result.
AI in marketing: what the numbers show
Marketing is the area where AI is most widespread in companies. According to Eurostat, 35.05% of German AI users used AI for marketing or sales in 2025, which is 9.1% of all companies. In 2021 it was only 1.26% of all companies. In retail, 44.35% of AI users apply AI to marketing or sales, in the advertising and market research group 49.43%. Using a different method, Bitkom arrives at 54% of AI users who use AI in marketing and communication.
In practice this means that creating copy, images and ad variants with AI is already routine for many competitors. How this plays out in Google Ads, from automatically generated assets to AI-driven bidding strategies, is something I describe in my article on AI automation in Google Ads. And because AI assistants are becoming an advertising environment themselves, it is worth taking a look at advertising in AI assistants.
Source for the Eurostat values: dataset isoc_eb_ai and for the sectors isoc_eb_ain2, both as of 15.06.2026.
What is deliberately missing here
- Surveys by software vendors and consultancies whose method is not disclosed. They often deliver striking rates but cannot be put into context.
- ZEW: The ZEW study for the Federal Ministry for Economic Affairs from 2024 gives 11% for 2021 and 12% for 2023, based among other things on official EU statistics (ZEW press release of 28.08.2024). These are essentially the same values as above. I did not find a more recent overall rate from ZEW.
- More recent IW surveys, for example among HR managers, I could not check in the original source. They are therefore left out.
- The employee perspective. How many employees use AI at work is a different question with different sources.
- Advertising and search market figures. You will find those in my Google Ads statistics.
- My own AI use. How much AI I use in my own business and what it costs is shown transparently on the page AI in numbers.
Download: all values as CSV
All values on this page, including the complete tables, the previous years' values and my own calculations (percentage point differences, rank), are available as a CSV with population, source, link and retrieval date for each row:
How to cite these figures correctly
With AI statistics, most errors arise not from copying but from leaving things out. These details make a figure verifiable:
- Source with survey: "Destatis, ICT survey 2025" or "Bitkom Research 2026" instead of "according to a study".
- Population: "companies with 10 or more employees" or "with 20 or more employees". Without this, the figure is not comparable.
- Year or survey period: reference year 2025 for Destatis, May 2026 for ifo, 2022 to 2024 for KfW.
- Base for partial values: percent of all companies or percent of AI users? For purposes, technologies and reasons, it is almost always a subgroup.
- Link to the original publication, not to a home page.
Two examples of how this can look in a text:
Example 1: In 2025, 26% of companies in Germany with 10 or more employees used at least one AI technology (Destatis, survey on ICT usage in enterprises 2025).
Example 2: In summer 2026, 57% of companies with 20 or more employees said they use AI (Bitkom Research, telephone survey of 603 companies, July to August 2026).
Typical errors I come across: the Bitkom rate is cited as "57% of German companies", although smaller businesses are missing. The marketing share of 35.05% is read as a share of all companies, although it refers to AI users. And values from different surveys are linked into a time series, such as 20% (Destatis 2024) and 57% (Bitkom 2026), as if adoption had almost tripled in two years.
If you cite this page as a compilation, please name the primary source as well. Below the article you will find a ready-made citation with a link.
Methodology
- Selection rule
- Included are official statistics (Destatis, Eurostat) and surveys by associations and research institutes whose population, sample and period are stated in the original publication. Every value has been checked against the publisher's press release, report or database. Surveys by software vendors and figures from secondary articles are excluded.
- As of
- All sources retrieved on 08.10.2026. Destatis reference year 2025 (tables of 24.11.2025), Eurostat databases isoc_eb_ai and isoc_eb_ain2 as of 15.06.2026, Bitkom September 2026, ifo June 2026, KfW February and July 2026, IW July 2025.
- Updates
- Quarterly. Based on the previous rhythm, the Destatis figures for 2026 can be expected at the end of November 2026, and Eurostat has announced the update of its overview article for December 2026.
- Own calculations
- Percentage point differences and Germany's rank in the EU are calculated from the Eurostat values and labelled as own calculations in the CSV. Eurostat values are shown with two decimal places, Destatis values rounded as published.
- Coverage
- In its quality report, Destatis names sections C to J, L to M and group 95.1, while Eurostat names sections C to J, L to N and 95.1 for all countries and also publishes values for section N for Germany. I follow the Eurostat description.
- Limitations
- The official values for 2025 are based on the EU enterprise definition, earlier years on legal units. Image, video and audio generation was added as a technology in 2025. The Eurostat overview article with data as of December 2025 differs slightly from the database in places because of later revisions. Survey values are not comparable as a time series with each other or with official statistics.
Conclusion
The official statistics say: a good quarter of companies with 10 or more employees used AI at the beginning of 2025, more than half of the large ones, and Germany is in the upper third in the EU comparison. The 2026 surveys suggest that adoption has continued to rise since then, but they measure with different yardsticks. Anyone citing a figure should therefore always say whom it describes.
For your own company, the rate of others is only a reference point anyway. What matters is whether you find a process in which AI measurably saves time or money, and whether knowledge and legal questions are settled. That is exactly where my AI consulting comes in: as an AI audit for €700, in which I review your processes for sensible use cases, or as an implementation sprint from €3,500. If the process is already clear and just needs to run reliably, I build it as an AI automation.




