Local businesses: the sources AI assistants cite to recommend you

"Who's the best osteopath near me?" People increasingly ask ChatGPT or Google AI Mode this question, not just the classic search bar. In the US, 45% of consumers used an AI tool to find a local business in the past 12 months, up from 6% a year earlier, according to BrightLocal's Local Consumer Review Survey 2026.

The catch: AI keeps very few names. SOCi's 2026 Local Visibility Index shows that ChatGPT recommends 1.2% of the locations studied, while Google's local pack shows 35.9%. Ranking well on Google does not guarantee a mention in a generated answer.

To make that short list, you need to know where AI looks for information: website, listings, reviews, local press. Here is what the studies show, engine by engine, and what to check to get recommended on Google and in AI answers alike.

Key takeaways

• An AI assistant doesn't "know" your business: it runs searches and assembles what your website, listings, reviews and the media say about you.
• According to Yext, 86% of AI citations come from sources a brand manages or can influence, led by websites (44%) and directory listings (42%).
• On "best X in Y" queries, editorial mentions (guides, press, curated lists) take a much bigger share, which makes them the lever that sets you apart.
• On Google, AI Overviews and AI Mode rest on the same foundations as SEO: indexed pages, an up-to-date Business Profile, structured data that matches the content.

How an AI assistant picks a local business

When you ask ChatGPT for a local business, it doesn't draw on a frozen memory. It runs a web search and can use the user's location to find local results. OpenAI's documentation on ChatGPT search states that, depending on the restaurant, it may show availability from OpenTable, Resy or Yelp. It also sets a technical condition: the site must allow OAI-SearchBot to crawl it. Each answer therefore depends on the question asked: the method to choose which prompts to track in GEO applies to local queries too.

Google works on the same principle, with its own data. AI Overviews and AI Mode use a "query fan-out" technique: the system runs several related searches across subtopics and data sources, then assembles the answer and its links.

The direct consequence: your visibility in an AI answer depends on what the sources consulted say about you, and on whether they agree. An address that differs between your website and your listing, or outdated opening hours on a directory, lower your chances of being picked: AI looks for corroboration before it recommends. Much of that corroboration comes from third-party media that AI readily cites, which becomes decisive on comparative queries.

Takeaway: AI rebuilds each answer from real-time searches. Whether you get recommended depends on how consistent all the sources about you are, not on a single channel.

The 4 families of sources AI cites

The Yext study on the sources of AI citations, published in October 2025, is the largest available on the topic: 6.8 million citations from more than 1.6 million answers generated by Gemini, OpenAI and Perplexity. It sorts sources by how much control the brand has over them. This table shows its breakdown.

The 86% figure for brand-managed sources is good news: most of the work is in your hands. But it describes an average across all query types. In a local market where your competitors also have a website and an up-to-date listing, these two families become a prerequisite, not an advantage.

Industry gaps matter too. In healthcare, listings and specialised directories account for 52.6% of citations. In food service, reviews and social content rise to 13.3%, the highest share of all the industries studied.

In the AI citation analyses Getfluence runs for its clients, the right question is not "which outlet has the biggest audience?" but "which outlets does AI already cite on the questions I'm targeting?". On a local query, a regional news title, a city guide or a specialised trade outlet often come back from one answer to the next.

Takeaway: websites and listings make up most citations. They are essential, but every competitor has access to them: the difference is often made elsewhere.

ChatGPT, Gemini, Perplexity, AI Mode: different sources

Each engine has its preferences, which explains why a business can be recommended by one and ignored by another. According to Yext, Gemini favours business websites, OpenAI leans heavily on third-party listings and Perplexity spreads across more sources, with more review platforms and local pages.

For ChatGPT, BrightLocal's study of its local sources adds another angle. Across 800 searches in 20 industries and 20 US cities, business websites made up 58% of the sources shown, editorial mentions 27% and directories 15%. BrightLocal also notes that ChatGPT relies mostly on Bing's index. A complete Bing Places listing is therefore anything but optional.

For Google AI Mode, the logic stays within the Google ecosystem: the official documentation explicitly recommends checking that your Business Profile and Merchant Center information is up to date.

These data have one limit: they cover the US market. The dominant directories change from country to country, but the mechanics (website, listings, reviews, mentions) stay the same.

Takeaway: don't work on a single engine. Google Business Profile for Gemini and AI Mode, Bing Places and third-party directories for ChatGPT, a broad presence for Perplexity.

Customer reviews: a filter more than a source

Reviews account for only 8% of citations in the Yext study. Yet they play a decisive role upstream. SOCi's 2026 Local Visibility Index, covering more than 350,000 locations of 2,751 multi-location brands, shows that locations recommended by ChatGPT average a 4.3-star rating. The report describes customer sentiment as one of the strongest filters for AI visibility.

In other words, AI doesn't necessarily cite your reviews, but it reads them to decide whether you deserve a recommendation. An average rating that is too low rules you out before sources even come into play.

In practice: a steady flow of recent reviews, systematic replies (including to negative reviews) and reviews that describe the service precisely. A review that mentions "e-bike repair within 24 hours" gives AI usable information that "great shop" doesn't contain.

Takeaway: reviews mainly act as a trust threshold. Without a solid rating, other efforts have little effect on AI recommendations.

Local press and curated lists: the lever for "best X in Y" queries

The share of media and editorial mentions varies widely by type of question. In the Yext study, press, forums and other sources account for only 6% of citations across all queries. But BrightLocal finds that on ChatGPT's informational searches ("best Italian restaurant in…", "which plumber to choose in…"), editorial mentions rise to 35% of sources, against 43% for business websites. These mentions include Wikipedia and titles such as Thrillist, Eater or Forbes.

These comparative queries are exactly where recommendations are decided. When a user asks for "the best", AI looks for an outside opinion to choose between candidates. Your website says you are the best; an article in the local press, a neighbourhood guide or a themed selection confirms it. That outside corroboration is what separates two businesses with equivalent listings.

These mentions also support Google rankings. An article in a local or specialised outlet, with useful content and a contextual link to your website, strengthens your authority in classic results, which themselves feed AI Overviews. Links and mentions work together: one strengthens your ranking, the other your credibility in the eyes of AI. Publishing where AI already looks for information works better than aiming for a prestigious outlet it doesn't consult for this kind of question. That is the principle behind Getfluence's AI-First™ approach: start from the sources AI already cites, then publish useful, contextual content there.

Takeaway: on "best X in Y" queries, AI looks for outside validation. Presence in local press and guides is the lever local businesses use least.

On Google: what feeds AI Overviews and AI Mode for local

Google is clear on one point: there is no special optimisation for appearing in its AI features. According to Google Search Central's documentation on AI features, a page must be indexed and eligible to be shown in Google Search with a snippet to appear as a supporting link in AI Overviews or AI Mode. SEO best practices therefore remain the foundation.

Three recommendations from this documentation directly concern local businesses:

• An up-to-date Business Profile: opening hours, categories, services, photos. Google explicitly mentions this point.
• Structured data that matches the visible content: LocalBusiness markup that states different hours or a different address from the page creates an inconsistency.
• Accessible text content: key information (services, service area, indicative prices) must appear as text on the page, not only in an image.

One page per location, with an address, a service area and precisely described services, serves both local SEO and the multiple searches AI Mode runs.

Takeaway: for Google, local AI visibility is the natural continuation of local SEO. A well-indexed website and an up-to-date listing remain the entry point.

Action plan: check the sources that talk about you

Before optimising, you need to know what AI sees. This method takes five steps:

• Ask the right questions. Ask ChatGPT, Perplexity and Google AI Mode 5 to 10 questions your customers would really ask ("best florist in Manchester", "florist open on Sunday in Manchester").
• Record the sources shown. For each answer, note the cited links and sort them into the 4 families: website, listings, reviews, press and guides.
• Fix inconsistencies. Name, address, phone number, opening hours: align your website, Google Business Profile, Bing Places and the directories of your industry.
• Check technical access. Allow OAI-SearchBot and Googlebot in your robots.txt file as well as with your host or CDN, and make sure your location pages are indexed.
• Close the editorial gap. Count how often each outlet or guide comes up in your records. Those that cite your competitors on several questions, and never you, are your priority: this is the sorting Getfluence applies before recommending an outlet to a client.

Repeat this check every month. AI answers vary from one session to the next: what counts is the trend across several checks, not a single result. BrightLocal also reports that 88% of AI users check the cited sources: those sources must convince a human as well.

Takeaway: a monthly record of cited sources shows exactly where you are missing. It turns a vague question ("why doesn't AI cite me?") into a list of actions.

AI assistants don't recommend businesses at random: they cite those whose website, listings, reviews and editorial mentions tell the same story. An up-to-date website and listings are the minimum; customer ratings do the sorting; local press and guides decide between candidates on the queries that matter. The next step is simple: run your first 5 questions in ChatGPT and AI Mode, and see which sources cite your competitors instead of you.

Your questions about the sources behind local AI recommendations

Where does ChatGPT get information about local businesses?

ChatGPT runs a web search and can use the user's location. According to BrightLocal's study, it relies mostly on Bing's index and cites business websites first, then editorial mentions and directories. For restaurants, OpenAI says it may show availability from OpenTable, Resy or Yelp. According to Yext, OpenAI leans heavily on third-party directory listings.

Does Google Business Profile play a role in AI answers?

Yes for Google: its documentation on AI features recommends keeping Business Profile information up to date. AI Overviews and AI Mode run several searches across different data sources to build an answer. The listing isn't enough on its own: the web page must also be indexed and eligible for a snippet in Google Search.

Do customer reviews influence AI recommendations?

They mainly act as a filter. According to SOCi's 2026 Local Visibility Index, locations recommended by ChatGPT average 4.3 stars. Reviews account for only 8% of citations in the Yext study, but 13.3% in food service. A low rating rules a business out before sources even come into play.

How can I find out which sources an AI uses to talk about my business?

Ask ChatGPT, Perplexity and Google AI Mode the questions your customers would ask, such as "best florist in Manchester", from several browsers with no history. Record the sources shown under each answer and sort them: your website, your listings, review platforms, press and guides. Repeat the exercise every month to track changes.

Sources

• Local Consumer Review Survey 2026: AI & Trust, BrightLocal, March 2026
• 2026 Local Visibility Index, SOCi, January 2026
• AI Doesn't Rank, It Cites. And 86% of Its Sources Are Brand-Managed, Yext, October 2025
• Uncovering ChatGPT Search Sources, BrightLocal, data collected in November 2024 (page updated in November 2025)
• AI features and your website, Google Search Central, accessed in 2026
• ChatGPT search, OpenAI Help Center, accessed in 2026