How does ChatGPT find local businesses like yours?

When customers ask ChatGPT for a local plumber, accountant or carpenter — who does the AI choose? Here’s how local visibility in AI searches works, and how your SME gets included in the result. With specific data from SOCi, BrightLocal and the Relark.dk case study.

How generative search works for local businesses: when a customer asks ChatGPT or Perplexity for, say, "a good plumber near me", the AI pulls web sources, weighs local businesses' signals — mentions, directories, structured data — and recommends the ones it can best verify. If your site isn't readable to AI crawlers, you're not in the running.

ChatGPT draws on many more sources than Google Maps when recommending a local business. No one outside OpenAI knows the exact number — we previously stated ‘17 data sources’ here; we were unable to substantiate that figure, and it has been removed. The key figure: SOCi’s Local Visibility Index (Jan. 2026, US data) found that ChatGPT recommends only 1.2 per cent of the locations in its listings, compared with 35.9 per cent in Google’s classic 3-pack. There are simply very few slots in an AI response.

Why does ChatGPT choose a different local business to yours?

Local AI visibility is your business’s visibility in responses from ChatGPT, Perplexity, Gemini and Microsoft Copilot when users search for local services in their area. In 2026, local AI visibility differs significantly from traditional local SEO. Whereas Google Maps alone used to determine whether you were found locally, AI models now draw on Trustpilot reviews, Krak listings, LinkedIn profiles, Wikipedia articles, news articles, the structure of your own website — and yes, Google Business Profile too, though often as a secondary source. At Geoa, we measure your visibility across all these sources in a single, comprehensive GEO-score score.

The proportion of AI responses drawn from each individual source is not published — we previously had figures here (64% Trustpilot, 22% Google Business Profile) attributed to SOCi; these figures do not appear in SOCi’s report and have been removed. The point remains: your competitor may be ranked lower on Google Maps and yet still be more visible in the AI, because the model analyses more sources than just the map.

How does the AI actually make its choices?

Each AI platform has its own method for finding local businesses:

• ChatGPT (OpenAI Search): Uses the Bing index as a basis, giving significant weight to Trustpilot and Wikipedia data.
• Perplexity: Runs live web searches, scanning Krak, Eniro, 118.dk, Trustpilot and your website in real time.
• Gemini (Google AIO): Draws primarily from the Google index + Google Business Profile.
• Claude (via Brave Search): Prioritises E-E-A-T signals — LinkedIn, CVR, and founding dates.

3 specific reasons why AI overlooks your local business

1. Your data is inconsistent across platforms

If your CVR number, address, telephone number or opening hours do not match across Krak, Eniro, 118.dk, Trustpilot, LinkedIn and your own website, the model will have conflicting information about who you are. We are not aware of any published figures on the cost in terms of citations — but the consequence is logical enough, and inconsistent data will cost you customers in any case.

2. Your website is designed for people — not for AI

This is the most tangible of the three issues, and we have measured it ourselves: in our scan of 465 Danish websites on 21 July 2026, 5 per cent had virtually no readable text without JavaScript — they are effectively invisible to AI — and 38 per cent lacked structured data. See the full survey.

3. You have no reviews — or too few

There are actually figures available here, but they should be treated with a degree of caution: they come from Trustpilot’s own report (May 2026, approx. 800,000 AI responses analysed) — in other words, a provider assessing the value of its own product. Bearing this in mind: businesses listed on Trustpilot were mentioned in 53.5 per cent of relevant AI responses; the remaining 75.3 per cent require 80 or more reviews and regular responses to them, not just a profile. Review and trust sites together account for 14% of all citations — not Trustpilot alone.

How can you get started today?

1. Check your data across platforms. Krak, Eniro, 118.dk, Trustpilot, LinkedIn — must be identical.
2. Create or update your Trustpilot profile. Ask 10–15 customers for reviews.
3. Optimise your website for AI. JSON-LD schema, BLUF on the homepage, clear pricing.