How to get your business recommended by Claude

What the September 2026 AIGROW Index shows about how Claude names businesses, what Anthropic says about how it reads the web, and the steps that follow from both.

Published by Guglielmo Paesano, founder of AIGROW · data under CC BY 4.0

35.7%Claude, local tradesof answers named at least one business
Answers
15,141
Names per answer
5.4
Repeated when asked again
38.8%
Named by no other assistant
69.8%
How the Index was made

Claude recommends businesses from what it already knows and, with search on, from what it finds on the web. In the September 2026 AIGROW Index, asked with no web search, it named at least one local business in 35.7% of 15,141 answers, and no other assistant named 69.8% of the businesses it gave. Check each assistant separately.

What the Index shows about Claude

1,400 pages of the September 2026 edition, each question asked three times, with no web search.

  1. Asked for a local business, Claude named at least one in 35.7% of its answers.

    9,737 of 15,141 answers about local trades named no business. A failed call is not counted as an answer: 0 of 16,797 calls failed. About software, it named no one in 0.0% of 1,656 answers.

  2. When it named someone, it gave 5.4 names per answer.

    Local trades, counting only the answers that named someone. Among the four assistants, from 2.5 for ChatGPT to 5.4 for Claude. About software, Claude gave 5.1.

  3. Asked the same question again, it repeated 38.8% of its names.

    Each question was asked three times. For each one, the names that came back in at least two runs, as a share of every name those runs gave. Among the four assistants, from 38.8% for Claude to 61.1% for Grok.

  4. 69.8% of the businesses Claude named, no other assistant named on the same page.

    11,347 of 16,251 different businesses, counted on the pages where another assistant also named someone. Being named by Claude says little about whether the others name you.

How Claude finds businesses

With no web search, Claude can name only businesses it learned about in training, which is what the figures above measure. With search on, it reads the web. What Anthropic says about that:

Claude-SearchBot
Indexes pages to improve Claude's search results. Anthropic says blocking it may reduce a site's visibility and accuracy in user search results.
Claude-User
Fetches a page when a Claude user asks a question that needs it.
ClaudeBot
Collects pages to train Anthropic's models, under a rule of its own.

Sources: Anthropic, how it crawls the web.

To see whether your site lets these crawlers in, use the free AI crawler check.

What to do, in this order

The first steps take an afternoon. Getting written about is the slow part, and the one that decides answers given without a search.

Ask Claude what your customers askPut the question the way a customer would, with the city: a dentist in Leeds, an accountant in Lyon. Then ask again in a new chat, because the answer changes: in the Index, Claude repeated 38.8% of its names when asked the same question again.freea new chat each time
Let Claude-SearchBot and Claude-User inOpen robots.txt and check that nothing blocks either of them, by name or through a rule for every bot. ClaudeBot, the training crawler, has its own rule and can stay blocked.robots.txttwo crawlers
Write out who you are and whereName, city, area, services and prices in the text of the page, not only in an image or a PDF. One page for each main service, and the same details, word for word, on your Google profile and in trade directories.on your sitetext, not images
Get written about where assistants readAsked with no web search, an assistant can name only businesses that were in the text it learned from. Reviews, local press and trade directories are where other people put your name next to your trade and city, and what they write reaches that knowledge only when a new model is trained.off your siteit takes months
Check the other assistants tooBeing named by one assistant says little about the others. The free scan asks ChatGPT, Claude, Gemini and Grok the same questions and shows which of them name someone else instead of you.free scanfour assistants

The four assistants side by side

The same kinds of questions, asked with no web search. Names per answer counts only the answers that named someone.

AssistantPagesNamed no one, localNames per answerRepeated when asked againNamed by no other assistant
ChatGPT1,4029.5%2.553.0%54.7%
Claude1,40064.3%5.438.8%69.8%
Gemini2,4000.0%5.158.5%74.3%
Grok10476.5%3.361.1%29.5%

What this does not show

No web search
Every answer came from what the model already holds, asked with no web search, no location and no chat history. Claude in its own app, with search on, can answer differently.
Through its maker’s tool
The first pages of this edition went through a reseller we no longer use, whose answers named someone far more often. This guide counts only the answers Claude gave through its maker’s own command-line tool, on 1,400 pages. The report gives each route its own row.
One month
These are the figures of the September 2026 edition. A new model can change any of them, and the next edition will show whether it did.
The maker's words
What Anthropic says about how it reads the web is quoted from its own documentation, as read in September 2026. Check it again before you act on it.

Does Claude name your business?

The free scan asks ChatGPT, Claude, Gemini and Grok your own customers' questions and shows who they name instead of you.

Edition September 2026 · data under CC BY 4.0