Short answer. By being consistently described across the handful of sources an AI assembles local answers from: a complete Google Business Profile, reviews across several platforms, directory listings, structured data on the website, and third-party mentions in local press and forums. The stakes changed because an AI local answer typically names only a couple of businesses rather than ten — a shorter shortlist than the old map pack, where being outside it means being unseen, not merely lower on a page.
Key takeaways
- AI local answers name a small shortlist of businesses, not a ranked list of ten, so missing the shortlist means being invisible rather than merely lower down.
- Only 38% of Google AI Overview citations now come from pages that also rank in Google's top ten, down from 76% roughly seven months earlier, so ranking well no longer guarantees inclusion.
- AI answers are commonly assembled from five source types: the Google Business Profile, reviews across multiple platforms, directories, the business's own website, and local press or community mentions.
- Businesses recommended by AI assistants tend to carry review averages around 4.3 stars spread across several platforms rather than concentrated on one.
- Inconsistent basic information — a mismatched phone number, an old address, conflicting hours — is one of the most common reasons a business is left out of an AI answer.
What actually changed for local search?
The result set collapsed, and the behavioural shift was fast.
| Measure | Finding |
|---|---|
| AI Overview citations that also rank in Google's top ten | 38%, down from 76% about seven months earlier |
| Citations from pages ranked 11–100 | Roughly a third of all AI Overview citations |
| Number of businesses a typical AI local answer names | Usually one to three |
A map pack is Google's block of local business listings shown above the organic results for a "near me" search; an AI answer compresses that further, to a handful of names or fewer. Ranking well still helps, but it no longer guarantees inclusion — a page ranked well outside the top ten can still be the one an AI system cites, and vice versa.
Where do AI local answers come from?
From five source types, only one of which most businesses actively manage.
A Google Business Profile is the free, owner-managed listing that controls how a business appears in Google Search, Maps, and Google's own AI answers — and it remains the backbone of most local AI answers, since Google's AI surfaces read Google's local data directly. Categories, hours, service areas, attributes, photos and Q&A all feed it.
Reviews across platforms, not just Google, are the second pillar. Volume, rating, recency and breadth all matter, and breadth is what most businesses neglect: businesses appearing in AI recommendations reportedly average around 4.3 stars across multiple platforms rather than on one alone.
Directories and best-of lists come third. Yelp, Apple Maps, industry-specific directories and local roundup articles are frequently crawled and cited, and businesses listed on a wide spread of authoritative directories are reported as more likely to appear.
The business's own website is the fourth source — structured data, location and service-area pages, and content answering the questions customers actually ask before buying.
Community and local press are the fifth. Reddit threads, local news sites and community blogs are commonly pulled into best-of and recommendation answers, much as Reddit and forum activity now shapes broader AI brand mentions — and this is the source type a business has the least direct control over.
Most published figures in this area come from marketing vendors rather than peer-reviewed research, and methodologies differ from one analysis to the next. The direction — AI-assisted local search growing, and citation sources widening beyond top rankings — is consistent across analyses even where the exact percentages are not; treat specific numbers as indicative rather than precise.
Why are most local businesses invisible?
Because inconsistency makes a machine choose a safer answer, and most businesses are inconsistent without knowing it.
| Failure | Why it costs you the mention |
|---|---|
| Conflicting basic data | A different phone number on an old listing, a former address on a review site, mismatched hours. When sources disagree, the low-risk move is to name a business whose data agrees everywhere |
| Reviews concentrated in one place | A strong Google rating and nothing elsewhere reads as a thin evidence base, not a strong one |
| No structured data | Without machine-readable markup stating name, address, hours, service area and services, a system has to infer the details from page copy — and inference is where errors enter |
| No third-party mentions | A business appearing only on its own website has no corroboration, and corroboration from independent sources is what these systems weight |
LocalBusiness schema is structured markup that states a business's name, address, hours, and service area in a format a machine parses directly rather than guesses at, and it is the fastest way to close the "no structured data" gap; structured data and entity identity work the same way for brands generally, not only for local listings.
A single inconsistent phone number can be enough to be skipped.
There is also a language dimension that gets overlooked wherever customers do not all search in one language. Reviews, listings and content exist in whichever language they were written in, so a business well described in one language can be invisible to a customer asking in another. In multilingual cities that is not an edge case — it is a large share of the market, and it is one of the seven reasons AI is not citing a brand even when the underlying business is a good fit for the query.
What should a local business do?
Fix the facts first, then build corroboration, then measure. In that order, because the first is cheap and blocks everything else.
- Complete the Google Business Profile properly. Correct primary category, all relevant secondary categories, precise service areas, current hours including holidays, real photos with descriptive filenames, answered questions.
- Make name, address and phone identical everywhere. Audit every listing you can find, including ones you did not create. Fix the contradictions before anything else.
- Spread reviews across platforms. Ask on Google most of the time, but direct a share of requests to Yelp, Facebook or your industry's main directory. Respond to reviews, including negative ones — response rate is itself a signal.
- Add machine-readable location data, following the same technical checklist that helps any website get cited by AI engines: address, geo coordinates, hours, service area and services listed explicitly.
- Publish the questions customers ask before buying. Pricing ranges, what to expect, how to compare providers. These are the prompts people put to an assistant.
- Pursue local press and community presence. A mention in a local publication, or a genuine non-astroturfed presence in community forums, does more than another page on your own site.
- Then measure. Ask the assistants the exact near-me and best-in-city questions your customers use. Record whether you are named, how you are described, which competitors appear, and which sources are cited. Those cited sources are your target list, and measuring AI visibility without fooling yourself is worth reading before building the monthly report. Repeat monthly.
Businesses that outsource this work tend to look for the same criteria that matter across AEO and GEO providers generally: a stated methodology, verifiable proof points, and a monitoring cadence rather than a one-off audit. Lifewood's own AEO programme applies this same discipline to multilingual local listings.