AI search visibility (GEO)

AI assistants like ChatGPT, Perplexity, Claude, Gemini, and Google's AI Overviews are already the first place a meaningful share of users go for local recommendations. "Best dentist in Brighton". "Where can I get my watch repaired in central London?". Whether you're mentioned in those answers, and how, is decided by a different set of signals than traditional SEO. This is GEO: Generative Engine Optimisation.

Why GEO is different from SEO

Traditional SEO ranks pages. AI assistants don't show pages. They generate answers and optionally cite sources. The mechanics:

1. 1

   User asks a natural-language question

   Often longer and more contextual than a typed Google query. "I need a dentist in Brighton open on Saturdays who takes nervous patients." A model can parse all three constraints at once.

2. 2

   The model runs underlying searches

   One or more searches against web indexes (Bing, Google, Brave, internal indexes), Maps APIs, news feeds, partner data sources. The model interprets the user's question into one or more search queries.

3. 3

   The model synthesises an answer

   It picks which sources to cite, which businesses to name, and how to phrase it. The selection is partly about what's most relevant, partly about what's most reliably summarisable.

4. 4

   The user sees a paragraph or two

   Sometimes with citation links, sometimes without. The user typically does not click through to the SERP. They've already got their answer.

Two things follow from this. First, "ranking #1" doesn't exist for these queries. You're either in the answer or you're not. Second, the model decides what to mention based on what it can read about you across the web, plus its underlying-search results. Your job is to make sure both layers say something favourable.

Traditional SEO is about being found. GEO is about being cited. The mechanics differ, but the foundations overlap heavily.

SEO vs GEO: the key differences

What AI assistants actually use to pick businesses

Patterns we've observed, replicated, and stress-tested across the major assistants:

The query patterns: how to test where you appear

Don't just check "best [your category] in [your city]". Real users phrase queries in a hundred different ways. Test across these patterns:

• Recommend a [category] in [city]
• Where should I go for [specific service] near [location]?
• I need a [category] who can [specific need]
• What's a good [category] in [neighbourhood]?
• Compare [competitor] to alternatives
• Best [category] for [type of customer] in [city]
• Cheap / affordable [category] near [area]
• Highly-rated [category] in [city]
• [Category] with [specific attribute] in [city]
• Open now [category] near me [city]

Run these across ChatGPT, Perplexity, Gemini, and Claude. Note where you appear, what's said about you, and which competitors are mentioned that you didn't expect. The competitive map in AI search is sometimes very different from the Google competitive map. Our AI Visibility Tracker automates this so you can scan dozens of queries weekly without hand-running each one.

What to actually do

1. 1

   Make sure your traditional SEO is strong

   AI search uses web search as a foundation. Without organic visibility for your target queries, GEO is uphill. See the complete guide to local SEO for the foundations.

2. 2

   Build review depth

   100+ reviews with specific service mentions is the realistic floor for being quoted. See how to get more Google reviews. Encourage specific reviews (mentioning the service, not just generic praise) because those are what AI assistants quote.

3. 3

   Earn mentions in third-party 'best of' content

   Pitch local press and round-up writers. Sponsor local content where it makes sense. Each earned mention is a citation an AI model can pull from. Aim for 3 to 5 high-quality mentions per year, in publications a model might trust.

4. 4

   Add comprehensive schema markup

   LocalBusiness, Service, FAQ, Review, Product where applicable. The clearer your data is, the more accurately you'll be summarised. Models prefer structured sources over free-text ones because they're easier to extract from reliably.

5. 5

   Write clear, direct service pages

   Pages structured around real questions with clear answers ("How long does a [service] take?", "What does it cost?", "Do you offer evening appointments?") map directly onto how AI models summarise. Pages full of generic marketing copy do not.

6. 6

   Track and iterate

   AI assistants change their answers surprisingly often. Re-test monthly. Compare to where you rank in traditional search to see which signal is dominant for your sector. Some sectors (legal, medical) lean heavily on authoritative sources. Others (restaurants, beauty) lean on review sentiment.

What not to do

Sector-by-sector heuristics

We've seen consistent patterns across sectors for what AI assistants weight most heavily. Useful for prioritising effort.

Where this is going

AI search isn't replacing Google overnight, but it's eating click-through to websites at the top of the funnel. The businesses that win in three years are the ones that get cited in answers AND rank in traditional search. The ones that don't get cited slowly disappear, even if their organic rankings hold, because the user never sees the SERP. They just got an answer.

Where to go next