How AI assistants pick which local business to name
An AI assistant names the businesses it has seen described the same way in several independent places. What decides is agreement across sources, not the quality of your own website. A shop with one immaculate homepage and no second mention is invisible to the model, while a shop listed identically in six places gets named.
What the model actually reads
A language model answering "who is a good roofer in Bonn" does not browse the way a person does. It retrieves a handful of pages, reads them, and builds one answer out of what they agree on. Three properties decide whether your business is in that answer.
The first is presence in more than one place. A single source is a claim. Two sources that say the same thing are a fact, as far as the retrieval step is concerned. The second is a finished sentence to quote. Pages that answer the question in their first paragraph are cheap to quote; pages that build up to an answer over eight paragraphs are not. The third is machine readable detail: address, opening hours, trade, in a form that does not have to be guessed from a photograph of your shopfront.
Why directories get quoted and company sites do not
A company website says one thing: we are good. Every company website says it, so the sentence carries no information. A directory page says something a model can use: here are seven roofers in this town, in this order, with these addresses. That is a comparison, and a comparison is what the question asked for.
This is also why the head of a list matters so much more than a place further down. The answer paragraph of a member.show directory names the seven businesses at the top by name. Everything below is data, reachable but not quoted. When an assistant compresses that page into three sentences, the seven survive and the rest does not.
| What consumers said in 2026 | Share |
|---|---|
| Use ChatGPT or other AI tools for local recommendations | 45 % (6 % in 2025) |
| Read AI generated review summaries | 82 % |
| Willing to decide on the summary alone | 23 % |
| Trust AI platforms for business recommendations | 40 % |
| Trust AI as much as traditional reviews | 42 % |
| Still use Google as a review source | 71 % |
The number nobody expects
Traffic from AI platforms is still small. Across 101,574 websites measured over sixteen months, AI referrals accounted for 0.32 percent of all website traffic in 2026, up from 0.24 percent in 2025 and 0.02 percent in 2024. Anyone who tells you AI already replaced search is selling something.
The interesting figure is the second one. Visitors arriving from an AI platform spent 67.7 percent more time on the site, an average of 9 minutes 19 seconds against 5 minutes 33 seconds from organic search. They arrive having already read a recommendation, which is a different state of mind from having clicked the fourth blue link.
What to fix, in order
Start with the thing that is cheapest and most often broken: make your name, address and telephone number identical everywhere they appear. Not similar. Identical, down to the abbreviation of the street.
Then get onto two or three sources that a model is likely to retrieve, which means places that publish comparisons rather than self descriptions. Then, and only then, rewrite your own homepage so its first paragraph answers the question a customer would ask, instead of welcoming them to your website.
Questions and answers
Does my business need its own website for AI to name it?
How long before an assistant picks up a new listing?
Can I pay to be named in an AI answer?
Which assistant matters most?
What is the investment for a place at the head of a directory?
Sources
- BrightLocal, Local Consumer Review Survey 2026 1,002 consumers surveyed in the United States
- SE Ranking, AI traffic research study, 18 June 2026 101,574 websites, January 2025 to April 2026
- Google Search Central, Local business structured data Required and recommended properties, and the explicit note that rich results are never guaranteed
