Local Search Is Splitting: Maps, Google, and AI Answers

You rank well in your city and the phone is quieter than it should be. The reason is not your ranking. Local intent now resolves across three different surfaces with three different ranking systems, and the one where shortlists get formed, AI generated answers, does not read your location page the way Google does.

Here is the direct answer to the question most marketing leads are asking: you are invisible in ChatGPT or Gemini because those engines mostly repeat what other sources say about you, and your local program was built to make your own pages rank. Those are different jobs. Fix citability, not rankings, and do it per surface.

Position 4.6, zero clicks

Look at a query like "calabasas seo company." Average position 4.6. Eleven impressions. No clicks. The instinct is to go chase position two. That is the wrong read.

A page one position with no clicks usually means the answer was consumed before the click. The searcher got what they needed from the map pack, from an AI generated summary, or from a third party page that listed five options including you, and they never needed your site. Ranking was never the product. Being chosen was.

This matters more now because the audience moved. Buyers increasingly ask an AI tool for a local recommendation, then verify on Google, a map, or a review site before they call. AI did not replace a surface. It added one. One journey, three surfaces, and you can lose on any of them.

Local is three surfaces now

Stop buying local search as a single line item. Budget it as three.

Map pack. Proximity, category, review volume and velocity, profile completeness, and behavioral signals on the listing itself. Your Google Business Profile is the asset. This surface rewards operational hygiene more than content.

Classic organic. Location pages, service pages, internal linking, page experience, links. This is the surface most agencies still optimize because it is the one with the familiar dashboard.

AI answers. A model assembles a shortlist from sources it trusts and names three to five businesses. There is no position, no impression share, no blue link to count. You are either in the answer or you are not. This surface rewards corroboration: what other credible sources say about you, in text, with your service and your city named together.

In our BeFound AI work, we see the overlap between what ranks first on Google and what AI engines actually cite collapse from roughly 70% to as low as 17%. Those are two different audiences on two different signal sets. Treating them as one budget line is how brands end up dominant on a surface their buyers already left.

Why the engines disagree with each other

The harder truth: the AI layer is not one surface either. The engines disagree about where local truth comes from.

Your Google Business Profile is not irrelevant to AI. It is the most structured, most widely syndicated record of your business that exists, and engines resolving a local entity have to reconcile against it.

But the engines do not weight the same sources. The same local query returns different businesses and different cited sources depending on the engine. A Steady Demand analysis of 14,472 citations across 50 US metros, reported by Search Engine Land, found nearly 60% of Gemini's local citations went to businesses' own websites, while ChatGPT leaned far more on Reddit and directories. Category shifts it again. A restaurant shortlist and a B2B services shortlist are not assembled from the same places.

Thin review volume also keeps you out of the answer, because a business with almost no third party text about it gives a model nothing to assemble from.

Four engines. Four effective ranking systems. One of them mostly reads your own site, another mostly reads what strangers wrote about you.

Run this audit before you spend another dollar

You cannot optimize a surface you have not measured. This takes an afternoon and a spreadsheet.

1. Write ten local intent prompts. Phrase them the way a buyer talks, not the way a keyword tool writes. "Best [service] company in [city]." "Who should I hire for [problem] near [neighborhood]." "Compare [service] providers in [metro]." "Is [your brand] any good." Include two prompts naming a competitor, because that is how shortlists actually get built.
2. Run all ten in ChatGPT, Perplexity, Gemini and Claude. Forty runs. Use a clean session with no history or personalization.
3. Log four columns per run: which businesses were named, in what order, which sources were cited, and whether your brand appeared at all.
4. Mark the gaps. A gap is any prompt where you are absent but a direct competitor is present. Next to each gap, write the source the engine used instead. That source is your target, not the engine.
5. Score the pattern, not the instance. If you are absent across all four engines, the problem is corroboration. If you are present in Gemini but absent in ChatGPT, the problem is third party presence. If you are present everywhere but ranked last, the problem is review text and specificity.

Run it quarterly. The answers move.

Fix citability first

Three fixes, in this order. Do not reorder them.

Entity consistency. Name, address, phone, service categories and service area identical everywhere they appear, starting with your Google Business Profile. Models resolve entities before they recommend them. Conflicting records mean you are two half-trusted businesses instead of one trusted one.

Structured service and location data. Schema for organization, service, service area and FAQ. Structured comparison content. Clear, parseable statements of what you do, where, and for whom. This is the one lever entirely under your control. It is also where most location pages fail: they are written to rank, full of soft copy, with the actual facts buried.

Earned citations that name the service and the geography together. Press, directories, roundups, review text, community threads. The phrasing matters. A mention of your brand with no service and no city gives a model nothing to match a local query against. Ask customers for reviews that say what you did and where. Review text is source material for your shortlist position.

Content and paid social are the fuel here, not the finish line. Their job on this surface is to produce third party mentions and quotable, structured source material, which is a different brief from producing clicks.

Where to start

Order of operations: audit the four engines first, fix entity consistency second, structure the site third, build earned mentions fourth, and keep the map pack healthy throughout. Anything else is optimizing a surface before you know whether you are on it.

We built BeFound AI for exactly this problem, auditing AI citation visibility across ChatGPT, Perplexity, Gemini and Claude, then structuring content and earning the third party citations that get a brand named in the answer. If you rank locally and still never show up in the shortlist, that is the gap to close. Email hello@payanimedia.com and we will map it.

Frequently asked questions

We rank in the top three locally. Why does ChatGPT never mention us?

Because ranking and citation run on different signals. Google rewards your own pages; AI engines largely assemble shortlists from what other sources say about you, including your Google Business Profile, directories, review text and community threads. In our BeFound AI work we see the overlap between what ranks first on Google and what AI engines cite fall from roughly 70% to as low as 17%.

Is my Google Business Profile still worth maintaining for AI search?

Yes, and it is arguably more important now. BrightLocal's analysis of 1.9 million citations across 116,670 domains found Google Business Profile accounted for 28.63% of all citations and appeared for 94% of businesses, making it the single most cited local source.

Do our customers actually use AI for local searches?

BrightLocal's Local Consumer Review Survey found consumers using AI tools for local business recommendations rose from 6% to 45% in a year, with 64% of adults aged 30 to 44 asking an AI tool for a business recommendation in the past year. Use is lowest among over-60s. Most people who used AI still verified on Google or another channel, so it adds a surface rather than replacing one.

Why do different AI engines give completely different local answers?

They trust different sources. A Steady Demand analysis of 14,472 citations across 50 US metros, reported by Search Engine Land, found nearly 60% of Gemini's local citations went to businesses' own websites, while ChatGPT leaned far more on Reddit and directories. That is why the audit has to run across ChatGPT, Perplexity, Gemini and Claude, not just one.

Do review counts affect whether AI names us?

Directionally, yes. A vendor study from 8ight.AI, reported by PPC Land, found ChatGPT named only 3.4% of tested local firms with under 25 reviews. Because it is vendor-run, treat it as a signal rather than a hard threshold, but build review volume and make sure review text names the service and the city.