The query nobody used to win
For a long time the most contested phrase in local marketing was the one that put a category and a city next to each other. Best plumber in Leeds. Best wedding photographer in Cork. Best branding agency in Belfast. Those queries decided who got the call.
Google won that battle. Maps, reviews, schema, citations on Yelp and Yell, Google Business Profile filled out properly. Whole agencies built themselves on getting clients into the local pack.
That model assumed the buyer was scrolling. Scanning a list. Comparing reviews. Clicking through. The buyer did the filtering.
Buyers don't scroll any more. They ask. And the answer comes back as a sentence, not a list.
What's actually happening
The thing I keep noticing, in my own behaviour and in clients I talk to, is that the question hasn't changed. The interface has.
People still want the best [thing] in [their city]. That bit is permanent. What's gone is the patience to evaluate ten options. If the model says one name, that name wins. If it says three, the buyer might check those three. Anyone outside that shortlist is invisible to them for that decision.
I tested this with my own studio a few months ago. I asked four different models who the best branding agencies in Belfast were. The answers overlapped on maybe two names across all four. The rest of the lists barely agreed with each other. Most of the firms I'd have expected to dominate weren't in any of them. Some that were, I'd never heard of.
That's the new shape of the local market. Not "we're number three on Google." More like "we either are or we aren't on the list, and the list is short."
Why local SEO doesn't translate
The instinct, when something new comes along, is to assume the old playbook will mostly work with a few tweaks. It won't.
Local SEO was built around signals Google's algorithm could verify cheaply. Distance from the searcher. Review count. Category match. A clean Business Profile. Citations across local directories. Backlinks from the local press. All of these still matter for Google itself, and Google still drives a lot of traffic. So don't bin the playbook.
But AI platforms are pulling from a different set of sources, and they care about different things.
They want corroboration. Multiple independent mentions across sources they trust, saying broadly the same thing about who you are.
They want substance. A two-line listing on Yell isn't enough. They want sentences that describe what you do, who you do it for, and why someone might choose you.
They want signal from places real people gather. Reddit threads. Quora answers. Industry-specific forums. YouTube discussions. Editorial writeups.
A perfect Google Business Profile gets you almost none of that. You can be the best-optimised local business in the country and still be a black hole as far as the model is concerned.
The intersection that wins
Here's what I keep coming back to. The brands that show up at the intersection of local and sectoral are the ones the models pick.
Local on its own isn't enough. The Belfast coffee shop that's only known for being in Belfast loses to the Belfast coffee shop that's known for sourcing single-origin Ethiopian beans from a specific cooperative.
Sectoral on its own isn't enough either. A specialist branding studio with no clear geographic identity loses to a specialist branding studio whose Belfast presence is documented across the local press, design awards, and community events.
It's the overlap. A clear local presence (publications, reviews, community, named in local roundups) plus a clear sectoral story (positioning, citations from people in the field, a defensible point of view) is the combination that gets you into the answer.
This makes sense if you think about how the model is reasoning. It's looking for high-confidence matches. Two specific filters narrowing down to one or two names is exactly the structure these queries take. The brands that fit both filters, with evidence the model can see, win.
What I'd do if I were starting from scratch
A few things, in roughly the order I'd tackle them.
Run the queries on yourself first. Pick five variations of the "best X in [your city]" query a buyer might use. Run each through ChatGPT, Claude, Gemini, and Perplexity. Write down what comes back. Now you have a baseline. Without one you're guessing.
Get into local press, properly. Not a vanity profile. Real mentions in publications the model will trust. For Belfast that's the Belfast Telegraph, Sync NI, the Irish News. For other cities, the equivalent. If you can be the named expert in a roundup or a feature piece, that's worth more than a hundred LinkedIn posts.
Build a story the model can extract in a paragraph. If someone reads your homepage, your About page, and your services page, can they describe what you do, where you do it, and who you do it for in three sentences? If they can't, neither can the model.
Show up where local buyers ask questions. Subreddits for your city, Facebook groups, Quora answers about local services. Do it as a participant, not a marketer. Over time, the model reads those threads as ground truth about who's worth recommending.
Earn community proof. Speak at local events. Sponsor things that matter in your sector. Get involved in industry roundups in your category. The model can see this footprint when other people write about it.
Run schema and structured data on your site. LocalBusiness, Organization, FAQ. The technical layer that tells AI platforms exactly what kind of business you are and where you are. It's a one-day job and it's free.
What I'm not promising
I'm not going to tell you this is settled. The platforms are all moving. Citation behaviour shifts month to month. What works for Gemini today may not be what works for ChatGPT in six months.
But the direction is clear. Local discovery is going through a re-platforming, and the brands that figure out the new shape early are the ones who'll be the named answer in their cities for the next decade.
The category-in-city query was always the most valuable phrase in local marketing. It still is. The fight just moved to a different field.