Someone asks an assistant for a business like yours and gets two or three names back. Here is where those names come from, what influences the choice, and what nobody can honestly promise you.
The 2026 analysis behind these figures covered more than 350,000 business locations across 2,751 brands, almost all of them multi-location names with marketing teams behind them. A single location starts further back than these numbers suggest.
For twenty years, local search worked one way. A customer typed something into Google, got a page of results, and made up their own mind. Being sixth still meant being seen. There was room to be mediocre and survive.
That page is being replaced by an answer. Google now puts an AI summary above its own results, and millions of people skip the search box entirely and ask an assistant instead. Either way, the output is not ten options. It is two or three names, and then it stops.
Critically, there is no notification. Nobody tells you a customer asked, got three names, and called one of them. No ranking drop, no alert. You simply never learn you were in the running. The only way to find out is to go and look.
Result #6 still gets seen
There is no #4. It stops.
Assistants are not consulting a secret ranking table. They assemble an answer from whatever they can find about the businesses in your area, and for a local business that comes down to three sources.
Your Google Business Profile is the structured record an assistant reads first: categories, services, hours, attributes, photos, questions answered. It is the closest thing to a machine-readable description of your business that exists, and it is the one an assistant trusts most because it is verified.
Volume, recency, rating, and above all the actual words customers use. This is the largest body of natural language about your business anywhere on the internet, written by people with no reason to flatter you. It is also the only one of the three that moves every week.
Everything else the web says: your own site, directories, local press, industry platforms. Assistants answer questions, so they favour businesses whose presence answers questions. Consistency matters as much as volume here, because contradictory details reduce confidence.
This is the part most local businesses get wrong. Winning the map pack and being named by an assistant are related but genuinely different outcomes. In a 2026 analysis of the same set of businesses, fewer than half the map pack winners appeared in AI recommendations at all.
Same businesses, measured both ways (source).
Part of the reason is proximity. Google Maps weights how close you are to the searcher very heavily, which is why a mediocre business on the right corner still ranks. An assistant asked for “the best plumber in Columbus” is not standing anywhere. It has no location to be near, so it leans harder on the things that describe quality and trust: your rating, how many people wrote about you, how recently, what they said, and whether your details agree across the web.
Which is quietly good news for a genuinely good business on a bad corner.
Open ChatGPT, paste the line opposite, and put in your own trade and town. Run a variation or two: cheapest, best reviewed, open late.
Pay attention to two things. Whether you are named at all, and who is named instead. That second list is more useful than your own result, because it tells you exactly who the assistant currently considers the credible option in your market.
If you would rather have it done properly across all four assistants with your live Google data alongside it, our free report does that in about two minutes.
Best [your trade] in [your town]? Give me your top three.
Illustrative layout, not a real result.
Nobody can guarantee an assistant will name you. Answers shift with phrasing and change without notice. A guarantee means the seller either does not understand the systems or hopes you do not.
Filtering customers so only happy ones reach Google violates Google’s policies and sits badly with the FTC’s rules on review suppression. Offer everyone a private channel first, then let them all review.
A 5.0 from eleven reviews is weaker than a 4.6 from three hundred. Volume and recency carry real weight, and a perfect score on thin evidence reads as thin evidence.
Different hours on your site than your profile, an old phone number in a directory. Each inconsistency chips away at how confidently a model will describe you.
None of these are exotic. They are the ordinary ways a good business ends up invisible.
Plainly: nobody outside these companies knows exactly how the models weigh their inputs, because none of them publish it, and it changes. Anyone selling certainty here does not have it.
What is established is which inputs exist. Assistants read Google Business Profiles, review content, and the wider web, and businesses that are strong on all three turn up in these answers far more often. That is not a theory about the algorithm. It is an observation about the outcome.
So the sensible move is not to chase a formula. It is to be unmistakably the best-documented option in your area, and to check regularly whether it is working.
It depends on your town and your trade, which is why the honest answer is to measure it rather than assume either way. What is not in dispute is the direction of travel: Google now places an AI answer above its own results, and the work that earns an AI recommendation takes months to compound. Starting after it decides your bookings is starting too late.
No. AI answers vary by how the question is phrased, change without warning, and differ between users. Any agency guaranteeing placement is either misunderstanding the systems or misleading you. What can be promised is measurement and improvement of the inputs those systems demonstrably draw on.
They overlap heavily but treating them as identical is the common mistake. Research across more than 350,000 business locations found that fewer than half the businesses winning Google's local map pack also appear in AI recommendations. Ranking well is a strong signal, not a free pass.
Your review signal. Volume, recency and the specific language customers use. It is the largest body of text about your business that an assistant can read and quote, it is the only lever that moves week to week, and most local businesses are barely working it.
Profile fixes can register within weeks. Review signal builds over 60 to 90 days depending on customer volume. AI recommendations are the slowest of the three, because assistants take time to pick up and trust changes. Anyone quoting you a fortnight is guessing.
Free report, about two minutes, no card. We pull your live Google data, check who is outranking you nearby, ask the assistants about your market, and email you the lot.
Get my free report