The essentials
- The situation: a growing share of local searches go through an AI assistant ("a good plumber near Nantes") that recommends three or four businesses instead of a list.
- The sources: Google Business Profile, reviews, trade directories, local press, the company's site and town pages; assistants cross-check these sources.
- The main lever: consistency of information across every source, ahead of review volume and site content.
- Measurement: put ten typical local questions to three assistants, every quarter, and count the citations.
Local searches are gradually migrating to AI assistants: instead of typing "plumber Nantes" and scanning the local pack, a growing share of users ask "which reliable plumber near Nantes for a leak on a Sunday". The answer is no longer a list but a recommendation of three or four businesses, often with a justification attached. Local GEO means the work that gets you in there. It rests on the same foundations as local SEO but differs in the sources drawn on and in how the information has to be phrased. The general framework is set out in classic SEO or GEO.
Where assistants draw their local recommendations from
| Source | What it brings | Weight |
|---|---|---|
| Google Business Profile | Existence, category, address, opening hours, reviews, photos | Decisive, particularly for Gemini and AI Mode |
| Customer reviews (Google, directories, trade platforms) | Reputation, actual specialisms, strengths cited | High: assistants often reuse the reasons people gave for being satisfied |
| Directories and sector platforms | Confirmation of existence and activity | Medium, but necessary for consistency |
| The company's site | Detailed services, service area, prices, answers to questions | High for specific questions |
| Local press and news | Prominence, longevity, notable events | Low but distinguishing |
| Local forums and social media | Spontaneous recommendations | Variable, hard to influence honestly |
The six levers of local GEO
- Consistent information: name, address and phone number strictly identical everywhere. An old address on one directory is enough to create a doubt the assistant resolves by citing a competitor. It is the rule set out in NAP consistency.
- A complete Google Business Profile: a precise primary category, services listed one by one, up-to-date hours, recent photos, the questions section filled in; see optimising your Google Business Profile.
- Numerous, recent, detailed reviews: assistants readily cite recurring reasons for satisfaction. A review describing a specific job is worth more than ten five-star reviews with no text; the method is in collecting quality Google reviews.
- A site that answers local questions: a service area detailed town by town, lead times, indicative prices, emergencies covered, specialisms. These are exactly the criteria the questions put to assistants contain.
- Structured data: LocalBusiness markup with address, hours, area served and services, described in schema.org markup for local.
- External mentions: sector directories, trade federations, partners, local press, which confirm the business to the models.
What changes compared with classic local SEO
Three differences matter. A question put to an assistant is longer and more specific than a typed query: it contains criteria (urgency, day, budget, specialism) the site has to answer explicitly. Then, the assistant doesn't rank ten results but recommends three: there is no second page, and presence is binary. Finally, the justification counts as much as the presence: being cited as "specialising in emergency leaks, open on Sundays, very highly rated for speed" beats a neutral mention, and that phrasing comes from the reviews and the site.
Measuring your local presence in AI
The protocol takes fifteen minutes a quarter. Write ten questions the way a customer would ask them, mixing service, place and constraint ("who comes out in an emergency at the weekend in Nantes for a leak"). Put them to ChatGPT, Gemini and Perplexity, three times each to smooth out the variability. For each question, note whether the business is cited, in what position and with what justification. That reading gives you the share of citations, the source used and the arguments the models kept, which points directly at what to fix. The general method is in tracking visibility on ChatGPT and Perplexity.
The frequent mistakes
- Neglecting the listing in favour of the site: for local, the listing weighs more in assistants' answers.
- A vague service area: "the Nantes region" doesn't answer a question about a specific town.
- No indicative prices: questions put to assistants often concern price; a range, even a broad one, lets you be selected.
- Old reviews: models favour recent sources; a steady flow counts more than a high total that has stopped moving.
- Contradictory information: different hours between the site and the listing, an old number on a directory.
How GreenRed helps
Rather than juggling several tools, GreenRed's Google Business Profile module brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
What is local GEO?
The work that gets a local business recommended by AI assistants when a user looks for a service near them. It rests on the Google Business Profile, reviews, directories, the site and structured data, which the models cross-check.
What is the most important lever in local GEO?
Consistency of information across every source: name, address, phone, hours and services identical everywhere. A contradiction between a directory and the listing is enough to create a doubt the assistant resolves by citing a better-documented competitor.
Do reviews count for AI recommendations?
A great deal, and their content as much as the rating. Assistants reuse recurring reasons for satisfaction to justify their recommendations. A steady flow of detailed reviews describing specific jobs weighs more than a high but ageing total.
How do I know whether AI recommends my business?
By writing ten questions the way a customer would ask them, mixing service, place and constraint, then putting each of them three times to ChatGPT, Gemini and Perplexity, once a quarter. Note the citations obtained and the justifications given to the businesses selected.