SEO & GEO
How to get recommended by ChatGPT for "best X in [city]" questions
Copy for AI
You get recommended by ChatGPT locally when the web confirms your specialisation and city.
Ask ChatGPT “which accountant in Ghent is good for a small limited company?” and you don’t get ten links. You get a short answer with two or three names.
Whoever is in that answer is on the shortlist. Whoever isn’t doesn’t exist for that buyer.
The verdict: how to get recommended by ChatGPT isn’t a trick. It’s a question of why the model names your peer in the same city, and not you.
This guide covers local and practice-area recommendations: “best X in city Y”. For ChatGPT visibility in general, read our guide to ChatGPT SEO.
Does a competitor already get named while you don’t? Start with your competitor gets mentioned by ChatGPT and you don’t.
Test it now: check with the free AI visibility checker whether you’re already named today.
How does ChatGPT put together a local recommendation?
It searches the web live, guesses where you are and summarises the sources it finds.
For a question with a place name or clear local intent, ChatGPT usually doesn’t rely on its training memory alone.
What OpenAI says about local search
OpenAI’s help page on searching with ChatGPT describes three steps:
- it rewrites your question into one or more targeted queries
- it sends those to external search partners
- it uses an approximate location from your IP address for local results
It also shows the sources the answer relies on. Our explainer on how ChatGPT uses a search index goes deeper.
Here’s what such a shortlist looks like in our own test:
Our own ChatGPT test, 17 October 2026. Three names, each with a label: that’s the whole shortlist a buyer sees.
A local question for an accountant or dentist produces the same format. A few names, a label each, and a source link.
Two consequences for your firm
That search step has two consequences for a firm or practice.
- You must be findable in the results ChatGPT requests. A firm that doesn’t appear for “accountant Ghent freelancers” rarely makes the selection.
- You must be easy to summarise. The model picks sources that quickly say who you are, where you are and what you specialise in.
Vague pages such as “we help you with all your needs” give it no usable sentence.
The same logic as local search
The model weighs roughly the same signals as a local search engine.
In its explanation of local ranking, Google names three main factors: relevance, distance and prominence. It adds that more reviews and positive ratings can help your local ranking.
You see the same logic in AI answers:
Does your offer match the question? Are you nearby? Do other sources confirm you’re known?
Three yeses get you named.
Which sources decide whether you get named?
Five types of sources come back every time in the answers we analyse per city and practice area.
How they fit together is also covered in local SEO for AI. Here’s the overview:
| Source | What the model takes from it | What you control |
|---|---|---|
| Your own specialisation pages | Practice area, target client, city, approach | One clear page per service and region |
| Google Business Profile and reviews | Category, address, rating, recent experiences | Right category, complete details, replies to reviews |
| Professional registers and bodies | Confirmation that you exist and are accredited | A correct listing in your profession’s register |
| Local directories, press and comparison sites | Prominence in the region | Mentions with the same name and specialisation |
| Consistent name, address and phone | Certainty that all those sources are about the same firm | Identical details everywhere |
A sources panel shows which of those the model actually opened. In this test, it was 10 pages:
Perplexity’s sources panel for our own test prompt, 17 October 2026: 10 web pages behind one answer.
ChatGPT’s Sources button works the same way. Our breakdown of which sources ChatGPT cites shows the patterns by source type.
Two points deserve extra attention.
Specialisation per city
A firm with one general “Services” page almost always loses to a firm with a specific page.
For example, “Bookkeeping for start-ups in Ghent”, explaining who it’s for and how it works.
The model looks for a sentence that answers the question. Give it that sentence. Our guide to local landing pages shows how to build one without thin copies.
Consistency
Say your firm sits in the register under an old name, in a directory with a previous address and on your site with a new phone number.
Then the model doubts whether they’re the same entity. Doubt means it would rather name a competitor it has no doubts about.
Google’s own guidelines for representing your business ask the same: use your real-world name and represent the business “as it’s consistently represented and recognized in the real world”.
How to straighten that out is explained in entity consistency and NAP consistency.
Quickest win: optimising your Google Business Profile.
How do you measure your AI visibility per city and practice area?
With a fixed set of buyer questions per practice area and city, each asked several times.
AI answers fluctuate. A lot. The same question can name three different firms today than it did yesterday, so a single test says little.
The five-step measurement
This is how we approach it in our own analyses:
- Write the buyer questions. Draft ten to twenty questions the way a real client asks them.
- Repeat each question several times. In different sessions and on different days.
- Record who gets named in each answer. You, your competitors, and the sources shown.
- Calculate your share. In what share of answers are you named?
- Look at the sources. A competitor’s sources usually point to exactly what you’re missing.
Good buyer questions sound like a client, for example:
- “good dentist in Antwerp who treats anxious patients”
- “employment lawyer in Leuven for employers”
- “best accountant in Ghent for a sole trader”
Vary the wording, the city and the specialisation. How to find those questions is covered in prompt research for ChatGPT.
What the share tells you
Say you ask 15 questions, three times each: 45 answers. If you’re named in 9 of them, your share is 20%.
Then count the questions where a competitor is named and you aren’t. That gap is your to-do list.
The pattern we often see
The picture is often sobering.
In many cities and practice areas, the same two or three firms come back time after time. Firms that do invest in a website or reviews are hardly named at all.
The difference: it rarely lies in the quality of the work. It lies in how easily the model can confirm their specialisation and location.
Check the clicks too
Being named is step one. Then check whether those answers send visitors.
In GA4, ChatGPT shows up as its own source:
GA4 Traffic acquisition in our own property, last 90 days, October 2026: 58 sessions from chatgpt.com.
That’s a small number, but it shows the channel is real. Our guide to tracking ChatGPT traffic in GA4 shows the setup.
To go deeper, follow the steps of an AI visibility audit. Or let the GEO checker test whether AI crawlers can read your site.
Measure first. Then fix.
What do you fix first?
Start with what helps the model recognise you with certainty, and only then add new content.
This order works in practice:
1. Align your details
Name, address, phone and specialisations identical everywhere:
- your own site
- your Google Business Profile
- your profession’s register
- the main local citations and directories
2. Get your business profile in order
Set the right primary category, your services, opening hours and photos.
Ask satisfied clients for a review and reply to every review, including the critical ones. Our guide on how to ask for Google reviews has wording that works.
3. One page per specialisation and region
Make it concrete: the question as a subheading and a direct answer below it.
Watch out: no thin copies per town. Write real content per client type.
4. Proof of expertise
Show a recognisable author or partner, qualifications and accreditations, publications or talks.
A face helps. The model prefers to name a firm with one.
5. Mentions outside your site
Local press, trade journals, professional bodies, partner pages.
A mention without a link counts too. So ask for it.
Then, and only then, more content
Only after that foundation does it pay to write more articles.
Rule of thumb: a knowledge base without a clear entity behind it produces few recommendations.
What about professional rules for lawyers, dentists and accountants?
For regulated professions, the question isn’t only whether you get named, but also how you communicate.
Lawyers, dentists, notaries and accountants have their own rules on advertising and publicity.
What’s often off-limits
These are often not allowed, or risky, in these professions:
- superlatives such as “the best dentist in Antwerp”
- comparisons with colleagues
- promises about results
Why that’s good news
What AI models reward fits those rules well.
Factual, informative content about your specialisation, your way of working and your accreditations is exactly what a model can summarise.
You earn it with facts: a recommendation comes from verifiable information, not advertising language.
Per profession
We worked this out per profession:
- Lawyers: GEO for law firms, with the professional conduct constraints.
- Accountants and bookkeepers: SEO for accountants and GEO for accountants.
- Dentists: SEO services for dentists, with the rules for health-care advertising.
When should you have your local AI visibility measured?
When a handful of self-run questions no longer gives you a clear picture.
You can run a first test yourself with five or six questions. That’s enough to see whether you’re named at all.
When to bring in help
You want outside help when you need:
- a reliable picture per city and practice area
- repeated measurements over months
- your share against the competitors named instead of you
- a concrete action plan per gap
That’s work for a GEO agency. For the map-pack side of the same signals, see our local SEO service.
Want to understand first why buyers already choose via ChatGPT? Read your B2B buyer already has a shortlist and clients say they found you via ChatGPT.
The sooner you measure, the sooner you see a trend.
Frequently asked questions
Does ChatGPT use my location for a local question?
Yes, usually an approximate one. According to OpenAI, ChatGPT can use an approximate location based on the IP address to show local results, and a more precise location if the user shares it. If the person names a city in the question, that city naturally carries the most weight.
Does a Google Business Profile help for ChatGPT too?
Indirectly, yes. A complete profile with the right category, correct details and recent reviews strengthens your local visibility in the search results AI assistants draw on, and it confirms your name, address and specialisation. It’s rarely the only source, but it’s one of the fastest to get in order.
How often should I measure my AI visibility?
Monthly with the same fixed set of questions is enough for most firms. What matters is that you ask the same questions every time, repeat them several times and keep the results, so you see a trend rather than a snapshot.
Can I pay to get recommended by ChatGPT?
You can’t buy an organic recommendation in an answer. What you can do is strengthen the sources the model relies on: your specialisation pages, your business profile, your listings in registers and directories, and your reviews. For advertising inside ChatGPT, read ChatGPT ads.
Where does your firm stand today?
Most firms that aren’t named lack proof, not quality.
Start with the five fixes above, then measure again in a month.
Want an outside view per city and practice area? Get in touch and we’ll go through who ChatGPT names instead of you, and why.
Free website scan
Enter your website and get an automatic scan within minutes, with concrete technical and SEO improvements. No sales pitch.
We only use your details for your scan. No spam, unsubscribe anytime.