How visible are B2B companies in AI search? Data from 103 websites
By The Venoh team · Last updated September 25, 2026
In September 2026 we analysed the public websites of 103 B2B companies and asked an AI answer engine with live web search the questions their buyers ask. Most of these companies never appeared in the answers. This page publishes the aggregate numbers, how they were measured, and their limits.
Finding 1: most B2B companies are absent from AI answers about their own category
For each company we wrote a small set of questions a buyer in its category would ask, without naming the company, and put each to an AI engine with live web search. We then checked whether the company's own domain was among the sources the answer cited or drew on.
| Measure | Result |
|---|---|
| Companies tested | 101 |
| Buyer questions asked | 311 |
| Answers that cited the company's own site | 17 (5.5%) |
| Companies cited in at least one answer | 12 of 101 (11.9%) |
| Companies never cited | 89 of 101 (88.1%) |
| Answers that cited other websites | 160 (51.4%) |
In 160 of the 311 answers (51.4%), the engine cited other websites instead. In most of the rest it answered without citing any source. Either way, the company whose category was being asked about was usually not part of the answer.
Finding 2: the basic machine-readable signals are often missing
| Signal | Sites with it | Share |
|---|---|---|
| An llms.txt file at the site root | 36 of 101 | 35.6% |
| An XML sitemap we could find | 79 of 94 | 84.0% |
| FAQ-style content anywhere we could confirm | 42 of 101 | 41.6% |
| Any schema.org structured data on the pages read | 52 of 103 | 50.5% |
| Organization (or Corporation) structured data | 45 of 103 | 43.7% |
| FAQPage structured data | 19 of 103 | 18.4% |
Denominators differ because a check only counts when it gave a definite answer. For example, a site whose server timed out on the sitemap request is left out of the sitemap row rather than counted as having none.
None of these signals guarantees a citation, and this data does not show that any of them causes one. They are the parts of AI visibility a company controls directly, which is why they are the first things the how to get cited by AI checklist covers.
What this means for a B2B team
- Assume you are not in the answer until you have checked. In this sample, nearly nine in ten companies were never cited for their own category's questions.
- Find out who is being cited instead. When an AI answer cites sources, you can see them. They show which sites shape how your category is described.
- Fix what you control first. Clear answer-first pages, FAQ content that matches real buyer questions, and structured data that matches the visible page are all within a site team's reach. See what GEO is for the wider picture.
How the data was collected
- Sample: 103 B2B company websites that Venoh researched in September 2026: software and technology companies (32 of them from Y Combinator's company directory) and B2B companies in engineering, manufacturing and services, many based in India. It is not a random sample of all companies.
- Site checks: for each company, the homepage plus up to four pages its own navigation links to, read as plain HTML, plus separate requests for /llms.txt, the sitemap and common FAQ addresses. The exact method is on the methodology page.
- Citation test: usually three buyer questions per company, written without naming the company, each asked once to Claude with its web search tool on the date of the research. A company counted as cited when its own domain was among the sources of the answer.
- Counting: every figure on this page was tallied by script from the stored research records. Nothing was estimated or extrapolated.
Limits
- One engine. The citation test used Claude with web search only. Other engines, such as ChatGPT, Gemini or Perplexity, may cite different sources; this data says nothing about them.
- A small number of questions per company, each asked once on one day. AI answers change with wording and over time.
- The sample is companies Venoh chose to research, so it is not representative of all businesses.
- Site checks read a few pages per site, not whole sites, and do not run JavaScript.
You may quote or reuse these figures with a link to this page. To see the same test run on your own company with more questions, see the diagnostic.
Frequently asked questions
How often do AI search engines cite B2B companies' own websites?
In Venoh's September 2026 data, across 311 buyer questions put to Claude with web search, the company's own website was cited in 17 answers (5.5%), and only 12 of 101 companies were cited at least once.
How many B2B websites have an llms.txt file?
In Venoh's September 2026 sample, 36 of 101 B2B company websites (35.6%) served an llms.txt file at the site root. The sample is not random, so the share across all businesses may differ.
How many B2B websites use FAQPage structured data?
19 of 103 sites (18.4%) declared FAQPage structured data on the pages Venoh read, and 51 of 103 (49.5%) declared no schema.org structured data at all on those pages.
Can I use this data?
Yes. You may quote or reuse the figures with a link to this page, together with its limits: one AI engine, a few questions per company, and a non-random sample of 103 B2B websites researched in September 2026.
Want this checked on your own pages?
The Venoh diagnostic reviews your real pages and labels every finding by how it was determined. Send us your website address and a line about what you sell.