AI writes the shortlist. Who is on it?

Results

Four findings for the IT industry

Where the company is named, its average position is 2.2; in 42.0% of those cases it takes first place. Questions about a concrete use case name the company less often (17%) than the market overview of its category (28%).

The five systems

No AI system is the exception

The same 800 questions, five systems, all with web search. The differences between systems are small — the difference between segments is large.

Share of answers naming the company asked about. Gemini answered only 17 times for security and cloud; that value is not reliable (—).

Comparison

Of all industries, the one that builds the technology

To compare with the KI-Blindtest Mittelstand 2026, the IT figures are counted the same way: ChatGPT and Google only, per question. IT does worse — almost entirely because of system integrators and IT service providers.

Only partly comparable: in the Mittelstand study (June/July 2026) ChatGPT answered without web search and Google AI Overviews were measured; here all systems search the web and Google AI Mode was measured. The direction is robust; read the gap as an order of magnitude only.

Who gets recommended

Usually a mid-sized firm, not a corporation

Who the AI names instead of the company: share of competitors named, by size, based on 5,814 mentions. A corporation has more than 1,000 employees or belongs to a group.

Sources

What the AI relies on

Share of all 21,508 source citations from 3,499 answers with sources (6,446 different websites). Rounded.

Seven in ten sources are vendor websites — but only in one answer in ten is it the website of the company asked about. In 29% of answers with sources, the website of a named competitor serves as evidence, often its guide or best-of list. Comparison portals appear in at least one answer in three.

From the measurement

Two cases — anonymised

Snapshots from 5 November 2026. Company and competitor names are omitted for fairness.

To take away

Charts to download

The four key charts as PNG (3,200 px wide) and as PDF for print — labelled in German. Free to use with the source line shown in each chart.

• Mentions by segment

  PNG ↓ PDF ↓

• Comparison with the Mittelstand

  PNG ↓ PDF ↓

• Corporation or mid-sized firm

  PNG ↓ PDF ↓

• What the AI relies on

  PNG ↓ PDF ↓

Transparency

Methodology

1. 1

   Companies

   100 companies with 20 to 1,000 employees: 40 software vendors, 40 system integrators and IT service providers, 20 IT security and cloud providers. 91 are based in Germany, 6 in Switzerland, 3 in Austria.

2. 2

   Selection

   From association, label and partner directories (for example "Software Made in Germany", the German IT SME association BITMi, the BSI list of qualified incident-response providers, TeleTrusT, eco, system-integrator partner networks) — deliberately not from comparison portals or AI answers. Excluded: subsidiaries of corporations, customers and prospects of CodaAI. Not a random sample.

3. 3

   Questions

   Eight questions per company without the company name, phrased the way buyers ask: two on the market overview, four on the use case, two comparing competitors. Derived from each company's website and market and checked before measuring. 800 questions in total, in German.

4. 4

   Systems

   ChatGPT, Claude, Gemini, Perplexity and Google AI Mode, all with web search, in German, model versions as of November 2026. One answer per question and system. Measured on 5 and 6 November 2026.

5. 5

   Evaluation

   We recorded whether the company is named (including common short forms), in which position, which competitors from its peer set are named, which websites serve as sources and whether its own website is among them. Of 4,000 answers, 3,744 could be evaluated; the rest failed because one system intermittently did not respond.

6. 6

   Sources and company size

   The 420 most frequent source websites were classified individually, the rest extrapolated from a random sample of 200 citations. A corporation is a company with more than 1,000 employees or one that belongs to a group, based on public information.

Honesty before effect

Limitations

An analysis is only as credible as the limits it names itself.

• Not a random sample: the results hold for these 100 companies, not for "the IT industry" as a whole.
• A snapshot: AI answers vary from day to day and model updates can shift mentions. There is one answer per question and system.
• Only competitors from each company's pre-defined peer set are counted — the number of answers naming competitors is therefore more likely too low than too high. A mention is not always a recommendation.
• The shares of source types are rounded and may differ by a few percentage points.
• Gemini answered only 17 times for IT security and cloud; that single value is not reliable and is not shown.
• The comparison with the KI-Blindtest Mittelstand 2026 is only partly possible: there, ChatGPT answered without web search, Google AI Overviews were measured instead of AI Mode, and the models were older. The direction is robust; read the gap as an order of magnitude only.
• The study measures the German-speaking market in German. AI systems work the same way elsewhere, but we have not measured other markets — so we do not claim them.

How to cite

How to cite this analysis

The aggregated figures may be quoted freely with attribution — in articles, talks, books and presentations. Please always state the base (share of answers or number of companies) and note that this is not a random sample.

Key statements to quote

Each statement stands on its own and has its own link (symbol #).

1. 1

   In 79.5% of the 3,744 AI answers evaluated, the IT or software company asked about is not named; it is named in 20.5% (766 answers).

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 1

2. 2

   Software vendors are named in 32.6% of answers, IT security and cloud providers in 25.1%, system integrators and IT service providers in 5.9%.

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 2

3. 3

   20 of the 100 companies are not named in a single answer — among system integrators and IT service providers, 16 of 40.

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 3

4. 4

   In 46.4% of answers (1,737) the company is missing and the AI names at least one competitor instead.

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 4

5. 5

   59% of the competitors named are mid-sized firms, not corporations (for questions about system integrators and IT service providers, 72%).

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 5

6. 6

   Only in 9.7% of answers (365) is the company's own website used as a source; 26 of the 100 companies are never cited with their own website.

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 6

7. 7

   Around 70% of all source citations lead to vendor websites, around 13% to comparison portals and directories.

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 7

8. 8

   No AI system is the exception: the mention rate per system lies between 18% and 23% of answers.

   Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/# Link to statement 8

Suggested citation

Parrizas, O. (2026). AI Blind Test B2B Tech 2026 [KI-Blindtest B2B-Tech 2026]: Analysis of 3,744 AI answers on 100 software vendors, system integrators and IT service providers (Version 1.0). CodaAI. https://www.codaai.ai/en/study-b2b-tech/

Parrizas, O. (2026) AI Blind Test B2B Tech 2026: Analysis of 3,744 AI answers on 100 software vendors, system integrators and IT service providers. Version 1.0, data as of 5–6 November 2026. Gütersloh: CodaAI (AMP Beratung). Available at: https://www.codaai.ai/en/study-b2b-tech/

Source: CodaAI, AI Blind Test B2B Tech 2026 (n = 100 companies, 3,744 AI answers, 5–6 November 2026), https://www.codaai.ai/en/study-b2b-tech/

@techreport{parrizas2026aiblindtestb2btech,
  author      = {Parrizas, Oliver},
  title       = {AI Blind Test B2B Tech 2026: Analysis of 3,744 AI answers on 100 software vendors, system integrators and IT service providers},
  institution = {CodaAI (AMP Beratung)},
  address     = {G{\"u}tersloh},
  year        = {2026},
  month       = nov,
  note        = {Version 1.0, data as of 5–6 November 2026},
  url         = {https://www.codaai.ai/en/study-b2b-tech/}
}

Data and versions

• Key figures as CSV ↓

  all aggregated values with unit — for checking and your own charts

• Text version (Markdown) →

  the whole analysis as plain text, also for AI assistants

Questions on the method: hi@codaai.ai. Publisher details for reference lists: Facts about CodaAI.

FAQ