ChatGPT can name 93.59% of local businesses in a given market, according to a July 2026 study of 10,000 US businesses by Insites. It still refuses to recommend 40.88% of them when a shopper asks an unbranded question like "best plumber in Austin." Perplexity performs worse: 68.79% of the same businesses never surfaced in its recommendations at all. The gap between being known and being recommended is where most local and multi-location brands lose AI visibility, and the signals that close that gap are not the ones most SEO teams spent the last decade optimizing for.
What ChatGPT and Perplexity actually know about your business
AI engines can usually name a business, but naming it correctly is a separate problem. In the Insites study, only 55.58% of ChatGPT's answers and 48.05% of Perplexity's answers fully matched the business's own Google Business Profile listing, and a wrong phone number was the single most common error. The same study found ChatGPT describes 79.8% of businesses positively or very positively even when its facts are wrong, so a warm tone is no guarantee the details underneath hold up. The failure mode gets stranger at the edges of what these systems know. A BBC reporter tested this by posting a false claim about himself on his own website, and within a day several AI tools, including ChatGPT and Google AI Overviews, repeated it as fact, according to reporting by Nextdoor's Kelsey Grady. That kind of confident error is a plain hallucination, and it's also the exact scenario a correction workflow exists to catch before it spreads across every engine that indexed it.
Reviews outweigh everything else, including your website
Review volume is the single strongest predictor of AI recommendation in the Insites dataset. Businesses that both ChatGPT and Perplexity surfaced averaged 133.4 Google reviews, compared with 10.7 for businesses neither platform ever mentioned, a gap of more than twelve times that dwarfs every other signal measured. Local Pack presence told a similar story: 74.6% of the most-recommended businesses held a Local Pack position, against 21.1% of the least-recommended. Classic search visibility still mattered too. Businesses surfaced by both AI platforms ranked for 71.1% of their target keywords in traditional search, versus 13.5% for businesses surfaced by neither, so AI recommendation and organic ranking reinforce each other rather than replacing one another. The pattern fits what VizibleAI has tracked more broadly as LLM Brand Visibility pulls weight away from backlinks and toward evidence a shopper, or an AI model, can actually verify. Most teams that already track brand mentions across ChatGPT, Gemini, and Perplexity are set up to extend that same monitoring down to the location level without building a new process from scratch.
Corroboration, not optimization: AI engines trust the businesses that show up the same way across many independent sources, not the ones with the cleanest website code.
Where ChatGPT and Google's AI actually pull local data from
Google Business Profile supplies roughly 28.5% of all citations behind Google's AI Overviews and AI Mode, according to an August 2026 BrightLocal analysis of nearly 2 million citations across 1,355 business locations. Yelp runs a distant second, appearing in about 80% of ChatGPT's local answers, making it the single most-used source for that platform specifically. Facebook showed up in citations for about 93% of the businesses studied, and TripAdvisor and OpenTable ranked highly for the restaurants that made up roughly a fifth of the sample. Individual business websites still account for about 93% of all domains cited, so a brand's own site isn't irrelevant, it is simply one voice among many; an earlier Insites analysis of the same research put ChatGPT's website-citation rate at 72% of awareness and reputation queries specifically. The pattern echoes what VizibleAI has found looking at citations that never touch a brand's own domain: AI engines corroborate an identity across directories before they lean on a homepage.
The technical SEO signals that don't move the needle
Core Web Vitals, structured data, llms.txt, and most classic technical SEO checks showed no meaningful correlation with AI recommendation rates in the Insites data, a reversal of a decade of advice built around Google's organic ranking algorithm. Freshness was the strongest single factor behind positive sentiment, and content depth correlated with accurate answers, which lines up with VizibleAI's own reporting on why fresher content earns more AI citations. Because these systems read raw page code rather than execute JavaScript, leftover or outdated copy buried in a page's markup can still shape what a model says about a business well after the visible page has been updated. None of this makes technical SEO worthless, and a fast, well-marked-up site remains better than a slow, messy one. It means AI engines weigh what a business's listings say about it everywhere over how cleanly any single page happens to be marked up.
Why bigger, older businesses keep winning the recommendation game
Larger, longer-established businesses hold a structural advantage in this system, because accumulating hundreds of reviews and dozens of consistent directory listings takes years a new location hasn't had yet. None of that reflects which business does better work. It mostly reflects how long a location has had to build up reviews, consistent directory data, and search visibility that corroborate each other. That dynamic mirrors a pattern VizibleAI has tracked across categories well beyond local search, where well-known brands keep winning AI recommendations by default, as covered in our look at incumbent bias in AI brand recommendations. For an agency managing AI Share of Voice across a dozen multi-location clients, the fix has less to do with any single page and more to do with systematically building review volume and directory consistency across every location, while keeping each Local Pack listing active and claimed. That workflow is covered in our guide to managing AI visibility across many client brands at once.
The reliability problem AI inherits from star ratings
AI recommendations are only as reliable as the reviews behind them, and star ratings carry a well-documented inflation problem that predates any chatbot. Nextdoor's Kelsey Grady reports that FTC economist Devesh Raval found most businesses average four stars or better, including ones with real, documented complaints, with almost no rating falling below two stars. The FTC finalized a rule banning fake and purchased reviews in 2024, with enforcement warning letters beginning in December 2025 and penalties starting above $53,000 per violation. None of that has slowed adoption: the same reporting puts AI assistant use for local recommendations at 45% of Americans, up from just 6% a year earlier, citing a 2026 BrightLocal survey. The one reassuring number is that 82% of people still check a website, social platform, or review site themselves before trusting an AI summary outright.
Frequently Asked Questions
Do AI engines use my Google Business Profile to generate answers about my business?
Yes. Google Business Profile supplies around 28.5% of citations behind Google's AI Overviews and AI Mode, more than any other source, according to BrightLocal's August 2026 analysis of local AI citations. Keeping it accurate and fully claimed is one of the most effective fixes available to a local business.
Why does ChatGPT know about my business but never recommend it?
Knowledge and recommendation are separate steps. ChatGPT recognized 93.59% of businesses in a 2026 Insites study but never recommended 40.88% of them in unbranded searches, usually because they lacked enough reviews or directory consistency to be corroborated across sources.
Do Google reviews really matter more than website SEO for AI visibility?
In the available data, yes, by a wide margin. Businesses recommended by both ChatGPT and Perplexity averaged 133.4 Google reviews versus 10.7 for businesses neither platform mentioned, while Core Web Vitals and structured data showed little measurable correlation with recommendation rates.
Does having a popular website still help local AI visibility?
It helps, but less than reviews and directory presence do. Individual business websites make up about 93% of all domains cited by AI engines, so a site remains one voice among many rather than the primary source these engines rely on.
Can fake or inflated reviews fool AI recommendation engines?
To some degree, yes, which is why the FTC's 2024 rule against fake and purchased reviews matters for AI visibility, not just consumer trust. Most people don't take AI recommendations at face value either: 82% still verify a business elsewhere before deciding.
What's the fastest way to improve AI visibility for a multi-location brand?
Standardize and grow review volume location by location, and make sure every Google Business Profile and directory listing agrees with the others. Agencies managing several locations typically track this as part of a broader AI visibility program rather than one location at a time.