AI search
How to show up when customers ask ChatGPT for a recommendation
Customers are asking ChatGPT and Gemini for recommendations. What actually changed, what the data really says, and the free work that helps.

Somewhere in the last year, “you need to optimize for AI” became the thing every marketer says to a local business owner. The pitch is that customers now ask ChatGPT for a plumber the way they used to Google one, so if the robot has never heard of you, you're invisible. Some of that is true. A lot of it is being sold to you by the same people who used to sell you “local SEO,” with a new coat of paint and a higher retainer.
This piece tries to do the calmer, more useful thing: separate what has actually changed from what is being hyped, using real numbers with their sources labelled — and then tell you the short list of things that genuinely help, almost all of which are free.
The shift is real. Here's how real.
Start with the honest version of the scary stat. In BrightLocal's 2026 consumer survey, 45% of US consumers said they now use AI tools — ChatGPT, Google's AI Mode, Gemini — to find or choose a local business, up from just 6% a year earlier. That made AI the third most-used discovery channel, behind only Google and Facebook, ahead of Yelp and Tripadvisor.
That is a genuinely fast shift, and worth taking seriously. But label the source: this is a vendor survey — BrightLocal sells local-search tools — of 1,002 US adults, and “have you ever used AI to look for a business” is a low bar that says nothing about how often people do it or whether they act on the answer. It measures curiosity as much as behaviour.
The more rigorous number is about what happens on the results page itself. In July 2025 the Pew Research Center — not a marketing firm, using the real browsing data of 900 US adults across nearly 69,000 Google searches, not a poll — found that when Google showed an AI summary at the top, people clicked a normal search result only 8% of the time, versus 15% when there was no summary. They clicked a link inside the AI answer itself just 1% of the time.
What AI actually reads when it answers a local question
A recommendation engine has to get its facts from somewhere, and it's not magic — each assistant leans on a specific set of sources. Knowing which ones is most of the battle, because it tells you where your effort should go.
Two patterns hold up across the research. First, an assistant's answer is only as good as the place data it's grounded in. Second, when it needs to describe what you actually do, it reads your own website. In BrightLocal's teardown of ChatGPT's local search results, business websites were the single most-cited source at 58%, ahead of third-party mentions (27%) and directories (15%). Label it, though: that's a small vendor study — 800 manual searches, ChatGPT's web-search feature only, captured in late 2024 — so treat 58% as a direction, not a law of nature. The direction is the useful part: your own pages are what the model quotes.
| Assistant | Main place data | How accurate (SOCi) |
|---|---|---|
| Google AI (AI Overviews, AI Mode) & Gemini | Google Maps / your Google Business Profile | ~100% — it's grounded directly in Maps |
| ChatGPT | Foursquare's Places API (a real partnership), plus the open web | ~68% |
| Perplexity | Open web + directories | ~68% |
Accuracy figures are from SOCi's 2026 Local Visibility Index, a vendor study of multi-location brands — a useful relative signal (Gemini is clearly more accurate because it reads Google Maps), not a precise grade for your specific business.
Two things on that table are worth spelling out. Google's AI surfaces and Gemini pull straight from Google Maps, which is why keeping your Google Business Profile accurate is the highest-leverage thing you can do — it feeds the old local search and the new AI one at the same time.
For ChatGPT, the Foursquare partnership is real: OpenAI announced Foursquare as a places-data partner for ChatGPT search in December 2024, and Foursquare's Places API covers 100M-plus locations. You'll see it repeated that “60–70% of ChatGPT's local results come from Foursquare.” We could not trace that figure to any primary source — it appears only in SEO blog posts citing unnamed “independent analyses.” The partnership is documented; the percentage is not. Claiming a free Foursquare listing is a reasonable, five-minute hedge, but don't let anyone bill you for it as if it's a proven lever.
The reality check: being picked is hard, and still tiny
Here is the part the “AI is eating local search” posts skip. Getting recommended by an AI assistant is genuinely difficult, and even when it happens, the traffic it sends is currently very small. Both of those cut against panicking.
The stat you'll see everywhere is that only 1.2% of businesses get recommended by AI, from SOCi. Read the fine print before you internalize it: that number is ChatGPT only, and the study looked exclusively at multi-location brands — 350,000 locations across 2,751 chains, the Starbucks-and-regional-franchise end of the market, not the independent shop. The same study found those brands were recommended by Gemini 11% of the time and Perplexity 7.4%, and appeared in Google's local 3-pack 35.9% of the time. So the honest read is “AI recommendation is much harder to earn than a Google local listing, especially on ChatGPT” — not a literal 1-in-83 lottery for your corner cafe.
And the traffic, today, is a rounding error. Sterling Sky — a well-regarded local-search agency that publishes its methodology — looked at one of their largest multi-location clients and found ChatGPT sent about 2% of the traffic that Google did in 2026 (up from 0.1% the year before, so growing fast, but off a tiny base). That was still only about a fifth of what Bing alone referred. AI referral traffic is climbing; it is nowhere near replacing search.
Finally, the intent split, which is the most reassuring number here. When Whitespark analysed 540 real queries, they found that for local-intent searches — the “barber near me,” “emergency plumber” type where someone is actually about to hire — the familiar local map pack showed up 93% of the time and AI Overviews just 15%. AI dominates the informational questions (“how do I unclog a drain”); the old surface still owns the moments that turn into customers.
What actually helps (and it's mostly free)
The good news buried in all of this: the things that make an AI assistant more likely to know and recommend you are the exact same things that already help you in normal search. There is no separate “AI channel” to build. Do these, in order.
1. Claim and clean up your Google Business Profile
This is the single highest-return task, because Google Maps is what grounds Google's AI answers and Gemini — the assistants that were most accurate in the SOCi data precisely because they read it. Get your name, category, hours, address and services right and keep them right. Our full walkthrough is in how to show up on Google.
2. Get real reviews, the allowed way
Assistants weigh reputation heavily — a business with steady, genuine reviews reads as trustworthy to a model the same way it does to a person. Ask every customer, offer nothing in exchange, and don't filter out the unhappy ones (that's against Google's policy and, since 2024, US law). The how-to, with the rules, is in how to get more Google reviews.
3. Have a clear website with a page per service
This is the piece that's specific to the AI question. Your Google profile can say “plumber.” It can't answer “do they do tankless water heater installs on a Sunday?” Your website can — and since your own site is the source these models quote most, a page that plainly describes each service, in each area you serve, is the raw material an assistant reads to recommend you for that specific job. One page listing eight services is weaker than eight clear pages. What belongs on each is in what to put on a small business website.
4. Claim the free listings, once
Spend an afternoon claiming the place-data sources these assistants read, then forget about them: a free Foursquare listing (what ChatGPT leans on), Apple Business Connect (free, and now open to businesses with no storefront), and Bing Places. Accurate and consistent beats “on 100 directories.”
The one thing to skip: an “AI optimization” retainer
A whole industry has appeared overnight selling “AI visibility,” “GEO” (generative engine optimization), and monthly retainers to get you recommended by ChatGPT. Be skeptical for a simple reason: almost every alarming statistic you've read on this topic — including some in this very article — was published by a company that sells the fix. That doesn't make the numbers all wrong, but it should make you slow down before paying a recurring fee to solve a problem the same vendor defined for you.
There is no secret sauce a retainer buys. The models read your Google profile, your reviews, your website, and a handful of public directories. Those are things you can get right once, for free, and they help you everywhere — not just in an AI answer that currently sends 2% of the traffic Google does. If someone can't show you a measurable result and a method, they're selling anxiety.
The bottom line
- Yes, the behaviour is shifting — 45% of consumers have tried AI for local recommendations, and AI answers are keeping people from clicking through. That's worth understanding.
- No, it's not an emergency. Being recommended is hard, the traffic is still tiny, and for “hire someone now” searches the old map pack still shows up 93% of the time.
- The work is the same work. Accurate Google profile, real reviews, a clear website with a page per service, a few free listings. All of it helps old search and new AI at once.
- Don't buy the retainer. The people scaring you about AI search are usually the ones selling the cure.
Sources
- 1.Pew Research Center — Google users are less likely to click links when an AI summary appears (July 2025, real browsing data, n=900)
- 2.BrightLocal — Half of consumers are asking AI for business recommendations (2026 consumer survey, n=1,002)
- 3.BrightLocal — Uncovering ChatGPT search sources (business websites 58%; 800 searches, Nov 2024, ChatGPT Search only)
- 4.SOCi — 2026 Local Visibility Index: few brands are chosen (multi-location brands; ChatGPT 1.2%, Gemini 11%, Perplexity 7.4%)
- 5.OpenAI — Introducing ChatGPT search (names Foursquare among place-data partners, Dec 2024)
- 6.Sterling Sky — The State of Local SEO in 2026 (ChatGPT ≈2% of Google's traffic for a large multi-location client)
- 7.Whitespark — The prevalence of AI Overviews in local search (540 queries; local pack 93% vs AI Overviews 15% on local-intent)
- 8.Apple — Business Connect expands to businesses without a storefront (free listing, Oct 2024)
Skip the build entirely
Describe your business to Sonara and get a real, hosted website back in minutes — bookings, enquiries and a database already wired in.
Start building