# Local businesses: how AI picks who to recommend and how much reviews weigh

> When AI Overviews, Gemini and ChatGPT answer local searches, where they take the names from, how much rating, volume and review text count, and what to do this week.

URL: https://geosnap.ai/en/blog/attivita-locali-come-ai-sceglie-chi-consigliare-quanto-pesano-recensioni
Language: English
Version IT: https://geosnap.ai/blog/attivita-locali-come-ai-sceglie-chi-consigliare-quanto-pesano-recensioni
Author: Rinald Sefa, CMO Geosnap · Category: GEO strategy
Publisher: Geosnap (Maind Group S.r.l.), https://geosnap.ai

## In short

In local searches AI steps in mainly when the question carries a specific need: across 53,900 Google results pages analysed by Synup, an AI Overview appears on 4% of "near me" searches and 75% of "best X for…" searches, and 60% of the cited sources are the business's own website or Google profile. A high rating and many reviews are the entry ticket (an average of 4.7 stars and over 230 reviews in the map); the edge comes from what reviews say, because Gemini and ChatGPT read them to decide which questions to recommend you for. Complete your profile, ask for descriptive, steady reviews, keep your data consistent everywhere and check what AI engines answer about your name.

- Synup (September 2026, 53,900 US SERPs): the map appears in 91-97% of local searches; an AI Overview in 4% of "near me" searches, 27% of "best X" and 75% of those with a specific need, where it shows business cards half of the time.
- 60% of citations in local AI Overviews point to the business's own site or Google profile; google.com is the most cited domain, followed by Yelp, Reddit, Facebook and Tripadvisor.
- Since July 2026 ChatGPT gets reviews, photos and business data from Yelp (a non-exclusive deal); where Yelp is marginal, your website and the review platforms it finds on the web are what count.
- Whitespark 2026: review signals are growing in importance; high rating is the sixth map factor, number of text reviews ninth, recency eleventh; businesses in the map average 4.67-4.79 stars and 233-261 reviews.
- Descriptive reviews (service, area, problem solved) are what AI uses to generate sections like "People talk most about"; AI-generated text and photos on the profile and keywords in the business name are negative factors instead.

Written by the Geosnap AI agent, reviewed and approved by Rinald Sefa.

Translated from the Italian original. [Read the original](https://geosnap.ai/blog/attivita-locali-come-ai-sceglie-chi-consigliare-quanto-pesano-recensioni)

If you run a restaurant, a practice, a shop or a service business, the question is no longer just "do I show up in Google's map?". More and more people ask Google AI Mode, AI Overviews, Gemini or ChatGPT "who is the best dentist near me" and get two or three names, not ten links. Here are the latest figures on when AI steps into local searches, where it takes the names it recommends from, how much reviews really weigh and what to do this week.

## How often does AI answer local searches?

It depends on how the question is phrased: a bare search shows almost only the map, a question with a specific need almost always gets an AI answer. The [Synup study](https://synup.com/research/local-serp-study) from September 2026 analysed 53,900 Google results pages across 250 US cities and 49 business categories, rendering each page in a browser (a plain scraper saw AI Overviews on 1.1% of pages, the browser on 75.4%).

| Search type | Map (local pack) | AI Overview | AI Overview with business cards |
| --- | --- | --- | --- |
| "plumber near me" | 97.4% | 4.0% | 0.9% |
| "plumber Manchester" | 91.4% | 8.6% | 3.8% |
| "best plumber in Manchester" | 92.7% | 27.0% | 15.9% |
| "best plumber in Manchester for urgent leaks" | 91.7% | 75.4% | 48.0% |

Two things to read in this table. First: the map is almost always there, so the Google profile remains the foundation. Second: as soon as the person adds a need, AI takes over and in half of the cases shows business cards right inside the answer, 2-3 per answer on average. Frequency varies a lot by sector: on "best X" searches an AI Overview appears 80% of the time for home builders, 69% for banks and accounting firms, 49% for lawyers, but only 1.2% for restaurants and 0.4% for cafés. The data is American; outside the US, AI Overviews and AI Mode are live in many countries, but the sector split needs checking on your own searches.

## Where does AI take the names from when it recommends a business?

Mostly from what the business publishes about itself and from what others say about it on a handful of platforms. In the Synup study, 60% of the citations in local AI Overviews point to the business's own footprint: its website or its Google profile. The most cited domain overall is google.com, meaning Business Profiles, with 13,777 citations; then Yelp (8,184 citations, and in 97% of cases the page was also on organic page one), Reddit, Facebook, YouTube, Instagram and Tripadvisor. In practice Google's AI fishes where Google already has the data: profile, reviews, photos and opening hours.

ChatGPT is a different picture, because it has no Google Maps data. In July 2026 [Axios](https://axios.com/2026/07/23/yelp-reviews-chatgpt-geo-partnership) reported the deal under which Yelp licenses reviews, photos and business information to OpenAI: ChatGPT shows them in local answers with Yelp branding and links, and a "Request a Quote" button is coming. [MarTech](https://martech.org/yelp-brings-reviews-and-local-leads-to-chatgpt/) adds that the deal is not exclusive. In markets where Yelp is marginal, what counts for ChatGPT is your website and the platforms it finds when it searches the web: Google reviews, Tripadvisor, industry portals, local press.

The principle holds for every engine: [Whitespark's Local Search Ranking Factors survey](https://whitespark.ca/local-search-ranking-factors/) finds that three of the five most important factors for AI search visibility are citations, meaning consistent mentions of the business on other sites, and sums it up as "in AI SEO, mentions (citations) are the new link". The same theme from the brand side is in [why AI cites Reddit, forums and reviews](https://geosnap.ai/en/blog/reddit-forum-recensioni-perche-ai-li-cita).

## How much do reviews really weigh?

A lot for getting ranked, and more and more for how AI describes you. The Whitespark 2026 survey (47 experts scoring 187 factors, published in November 2025) names as its main change "an increased importance of review signals and behavioural signals". For the map, the primary profile category still ranks first, then proximity and keywords in the business name; high ratings (4-5 stars) sit sixth, the number of reviews with text ninth, review recency eleventh, a steady flow over time fourteenth. Keywords in review text and positive sentiment appear further down, around fortieth place; owner responses much lower.

Synup's numbers show what it takes to be in the map today: businesses in the local pack average a rating between 4.67 and 4.79 and between 233 and 261 reviews; on "best X" searches two out of three have at least 4.8 stars. A high score and volume are the entry ticket, not an advantage.

The advantage comes from what reviews say, because AI reads them. Gemini in Google Maps answers questions about businesses using the profile's reviews and photos, and Google states that Ask Maps draws on over 300 million places and reviews from more than 500 million contributors (figures reported by [GatherUp](https://gatherup.com/blog/google-just-changed-the-rules-of-local-discovery-is-your-business-ready/); Ask Maps is currently available in the US and India, as [Arc Intermedia](https://www.arcintermedia.com/shoptalk/google-maps-adds-gemini-what-it-means-for-local-seo/) notes). In some tests, [WebProNews](https://www.webpronews.com/geminis-local-lens-googles-ai-reveals-how-it-sees-your-business) saw Gemini generate sections such as "People talk most about" or "Tips from reviewers" out of reviews. A review that says "great" adds nothing; one that says "they replaced the boiler in a day and left the bathroom clean" teaches the AI which questions to recommend you for.

A useful detail: [Whitespark](https://whitespark.ca/blog/22-local-seo-developments-you-need-to-know-about-from-q3-2026/) analysed 765 businesses and only 25% of profile categories get Google's "review questions" (which services did you get, how much did you pay), which help structure feedback. If your category does not have them, you have to guide customers yourself.

## What to do this week

1. **Complete your Google profile**: precise primary category, additional categories, predefined services, attributes, real opening hours. Whitespark quotes Google's guidance for generative AI features, which draw on the profile and Merchant Center feeds: "Keep your local business details complete".
2. **Ask for descriptive, steady reviews**: after each job, ask customers to mention the service, the area and the problem solved. Four reviews a month for a year beat fifty in one week.
3. **Reply to all of them**, especially the negative ones, with specifics: AI reads the replies and the overall sentiment too. How to fix a negative perception is covered in [when AI criticises your brand](https://geosnap.ai/en/blog/quando-ai-critica-il-tuo-brand-dati-sentiment-risposte-ai-come-correggerlo).
4. **Keep name, address, phone and hours consistent** across your site, Google profile, Tripadvisor, industry portals and social accounts: these are the citations AI uses as evidence.
5. **On your site, one page per service** with city, indicative prices and FAQs: it is the top organic factor in the Whitespark survey and it is what ChatGPT reads when it looks you up.
6. **Check what AI says about you**: ask ChatGPT, Gemini and AI Mode "best [category] in [city]" and "[business name] reviews". With Geosnap you can track these questions over time and see which sources get cited when you appear and when you are missing.

One figure keeps things grounded: in a Yext survey of about 4,000 consumers quoted by Whitespark, only 5% use AI as a direct path to purchase; the rest check on Google, on the website or on the map what the AI said. Profile and site must confirm the answer, not contradict it.

## What to avoid

Yesterday's shortcuts cost more today. Stuffing keywords into the business name correlates with more map positions, but it breaks Google's guidelines and Whitespark openly advises against it. AI-generated text and photos on the profile are among the new negative factors in the 2026 survey, with suspension risk. Hours that differ from real ones to "always be open" damage experience and trust. And bought or batch-written reviews are the fastest way to lose both the map and the AI answer, because inconsistent sentiment is the first thing a model notices.

## Sources

- [What's Ranking in Local Search: 2026](https://synup.com/research/local-serp-study) (Synup Research, 2026-09-18)
- [Whitespark's Official 2026 Local Search Ranking Factors Report](https://whitespark.ca/local-search-ranking-factors/) (Whitespark, 2025-11-06)
- [22 Local SEO Developments You Need To Know About From Q3 2026](https://whitespark.ca/blog/22-local-seo-developments-you-need-to-know-about-from-q3-2026/) (Whitespark, 2026-10-08)
- [Exclusive: Yelp partners with ChatGPT to surface reviews](https://axios.com/2026/07/23/yelp-reviews-chatgpt-geo-partnership) (Axios, 2026-07-23)
- [Yelp brings reviews and local leads to ChatGPT](https://martech.org/yelp-brings-reviews-and-local-leads-to-chatgpt/) (MarTech, 2026-07-28)
- [Google Just Changed the Rules of Local Discovery. Is Your Business Ready?](https://gatherup.com/blog/google-just-changed-the-rules-of-local-discovery-is-your-business-ready/) (GatherUp, 2026-03-26)
- [Google Maps Adds Gemini-Powered “Ask Maps”: What It Means for Local SEO and Reputation Management](https://www.arcintermedia.com/shoptalk/google-maps-adds-gemini-what-it-means-for-local-seo/) (Arc Intermedia, 2026-03-17)
- [Gemini’s Local Lens: Google’s AI Reveals How It Sees Your Business](https://www.webpronews.com/geminis-local-lens-googles-ai-reveals-how-it-sees-your-business) (WebProNews)

## Frequently asked questions

### Do AI Overviews show up for local searches outside the US?

Yes, AI Overviews and AI Mode are live in many countries, but the frequency figures by search type and sector quoted here come from Synup's US study. The general rule holds: the more specific the question ("best X in [city] for [need]"), the more likely an AI answer. Check your main searches from a phone in your own city.

### How many reviews do I need to appear in AI answers?

There is no stated threshold. In the Synup study, businesses in the map average between 233 and 261 reviews and a rating between 4.67 and 4.79, and Google's AI draws mostly from those businesses. More than the count, what matters is a steady flow over time and the content: reviews that name services, area and outcomes.

### Does ChatGPT read my business's Google reviews?

It has no direct access to Google Maps data. Since July 2026 OpenAI has a licensing deal with Yelp for reviews, photos and business data; beyond that, ChatGPT searches the web and reads the public pages it finds, including your website and indexed review portals. That is why you need one page per service with clear information.

### Should I put keywords in my business name on Google?

No. Whitespark's study of 1.4 million profiles shows it correlates with more map positions, but it breaks Google's guidelines and can get the profile suspended. Use your real name and put keywords in categories, predefined services and your website pages.

### How do I check what AI says about my business?

Ask ChatGPT, Gemini and Google AI Mode "best [category] in [city]", "[category] in [city] for [need]" and "[business name] reviews", and note whether you appear, with which description and which sources are cited. Repeat monthly: answers change often. Geosnap automates this check on ChatGPT, Gemini and AI Mode and shows the cited sources.