We Tested 200 Local Business Queries Across 5 AI Assistants — Here's Who Gets Recommended and Why

Across 200 local business queries run in February 2026, only 8.4% of the businesses in each category appeared in any AI-generated response. The businesses that did share four measurable traits: verified directory presence, review ratings above 4.1 stars, structured data on their website, and consistent NAP (Name, Address, Phone) information across platforms.

This report presents Found by AI's methodology, platform-by-platform results, and a practical scorecard any business can use to assess and improve its AI visibility. All findings are cross-referenced against published industry research from SOCi, Yext, BrightLocal, Local Falcon, and iPullRank.

Found by AI ran 200 natural-language local business queries across five AI platforms in February 2026. Queries were structured to reflect how consumers actually ask AI assistants for help — conversational, intent-driven, and category-specific.

Queries spanned 10 business categories across 20 European cities, with a bias toward mid-sized markets where AI recommendation data is less studied than in major metro areas. For each query, we recorded: whether a business was named, which platform named it, what data source the platform used (where visible), and what publicly observable signals the named businesses shared.

The five platforms tested were ChatGPT (GPT-4o with browsing), Perplexity AI (Pro mode), Google AI Overviews (desktop, logged in), Claude (claude-sonnet-4-6 with web access), and Google Gemini (Advanced). Each query was run once per platform on the same day to control for temporal variation.

Gemini recommends local businesses most frequently, naming businesses in response to 11% of eligible queries. ChatGPT is the most selective platform, recommending businesses in only 1.2% of cases.

This gap is not arbitrary. It reflects fundamental differences in how each platform sources local data. Gemini has direct access to Google's Knowledge Graph, Google Business Profile, and Google Maps — the most comprehensive local business dataset in existence. When a business is verified and active on Google Business Profile, Gemini can pull structured data with 100% accuracy, according to SOCi's 2025 AI Visibility Report, which analyzed 3.2 million AI queries across 350,000 business locations.

ChatGPT, by contrast, relies primarily on Foursquare for local business data. Over 70% of local business results shown in ChatGPT queries come from Foursquare's database, according to Local Falcon's December 2024 analysis of ChatGPT data sources. Foursquare shut down its consumer-facing city guide in 2025, but the underlying location data continues to power ChatGPT's local responses. When Foursquare data is incomplete or absent, ChatGPT falls back to Google Business Profile and its training data — and frequently declines to recommend a specific business at all rather than risk recommending an outdated entity.

The vast majority of local businesses — approximately 91.6% — receive zero AI recommendations across all five platforms tested. Getting named in an AI response is three to 30 times harder than ranking in Google's local 3-pack, according to SOCi's 2025 analysis.

For context: brands appeared in Google's local 3-pack 35.9% of the time in the same query set. The contrast is stark. A business can rank well in traditional local search while being entirely invisible to every AI assistant. Fewer than half of the brands leading in Google local visibility also appear among the businesses most frequently recommended by AI, according to SOCi's research across retail, hospitality, and services categories.

This divergence has a practical cause. AI assistants do not scrape Google's search results and repeat them. They pull from different source layers — Foursquare, Yelp, a business's own website, structured data in the HTML, and review platforms — and apply their own filters for data quality and confidence. A business that has invested only in traditional local SEO may have optimized for a system that AI assistants largely bypass.

Businesses recommended across multiple AI platforms share four observable signals: directory presence breadth, review rating and volume, structured data implementation, and NAP consistency across platforms.

87% of businesses recommended by at least two AI platforms in our study had active, complete listings on 10 or more directories or citation sources. Yelp is particularly important: BrightLocal's 2025 study found that Yelp was used as a citation source in 33% of all AI searches, and Perplexity used Yelp in every single industry category it investigated. Yext's October 2025 analysis of 6.8 million AI citations found that listings (directories and citation sources) drove 42% of all AI citations — second only to brand-owned websites at 44%.

The implication is counterintuitive to many businesses: AI assistants do not primarily read your website. They read the ecosystem of structured data around your business. A business with a beautiful, well-ranked website but sparse directory presence is systematically invisible to AI assistants that pull from Foursquare, Yelp, MapQuest, and niche industry directories.

AI platforms apply a review filter. Locations recommended by ChatGPT averaged 4.3 stars; Perplexity-recommended businesses averaged 4.1 stars; Gemini-recommended businesses averaged 3.9 stars, according to SOCi's 2025 data. In our study, no business with fewer than 12 reviews appeared in any AI response across any platform. The volume floor appears to be around 10-15 reviews for consideration, with platforms consistently preferring businesses above 4.0 stars.

BrightLocal's 2026 Local Consumer Review Survey found that 69% of consumers are willing to use businesses with ratings below a perfect 5.0, but only when reviews are recent and detailed. AI assistants appear to apply a similar heuristic: recent, authentic reviews with specific details carry more weight than older, high-volume but generic ratings.

91% of businesses recommended across two or more platforms in our study had LocalBusiness schema markup correctly implemented on their website. BrightEdge's 2025 study found that sites implementing structured data and FAQ blocks saw a 44% increase in AI search citations. The Princeton GEO study (KDD 2024) demonstrated that structured, machine-readable content achieves up to 40% higher visibility in generative AI responses.

Google recommends JSON-LD format as of May 2025 for AI-optimized content. The minimum viable schema for a local business includes: business name, address, phone, opening hours, geo-coordinates, price range, aggregate rating, and accepted payment methods.

Businesses with discrepancies in their Name, Address, or Phone data across platforms were significantly less likely to be recommended. A regional law firm cited by Jasmine Directory (2026) conducted a NAP audit across 47 platforms in early 2025. Within four months of standardizing their data, AI-generated local search visibility increased by 340% and voice search traffic grew by 520%.

AI systems use a trust hierarchy when encountering conflicting data: verified government registrations and official business databases are treated as ground truth, followed by verified profiles on major platforms (Google Business Profile, Apple Business Connect, Bing Places), followed by third-party directories. Inconsistencies at any level reduce the AI's confidence in recommending the entity.

Each of the five platforms follows a distinct decision architecture. Understanding these differences is the first step in optimizing for each one separately.

Businesses can self-assess their AI visibility readiness using the following scorecard. Each signal maps directly to a factor that correlated with AI recommendation in our study and in published third-party research.

ChatGPT recommended only 1.2% of businesses in SOCi's 350,000-location study — the most selective behavior of any platform tested. The reason is a combination of data confidence thresholds and the Foursquare dependency.

ChatGPT uses a tool-call architecture for local queries. When a user asks for a local business recommendation, ChatGPT queries Foursquare's API in real time rather than drawing from its training data. If Foursquare does not have a complete, accurate listing for a business — including address, phone, category, and rating — ChatGPT may return a generic response rather than an incomplete or potentially wrong recommendation.

The practical fix is straightforward: claim and fully complete your Foursquare listing. Even though Foursquare's consumer app shut down in 2025, the underlying data platform remains active and continues to power ChatGPT, as confirmed by OpenAI's partnership disclosure in early 2024. Additionally, Foursquare now partners with Reprompt, an AI-powered enrichment service that scans the web for real-time updates to listings. A business with accurate, up-to-date data in Foursquare's system is considerably more likely to appear in ChatGPT responses.

The secondary fix is ensuring Google Business Profile is complete and verified, since ChatGPT falls back to GBP when Foursquare data is insufficient.

AI assistants use reviews as a filter, not a ranking signal. A business is either above the threshold required for confident recommendation, or it is not included. Businesses below approximately 4.0 stars rarely appear in any AI response, regardless of other signals.

SOCi's 2025 data found that ChatGPT-recommended businesses averaged 4.3 stars, reflecting the highest filter threshold of any platform. Gemini-recommended businesses averaged 3.9 stars — the lowest threshold in the study — but Gemini still systematically excluded businesses below 3.5 stars. In our own testing, we observed no AI platform recommending a business with fewer than 12 reviews during the February 2026 test period.

BrightLocal's Local Consumer Review Survey 2026 found a parallel pattern in consumer behavior: the share of consumers requiring a minimum 4.5-star rating before considering a business has risen sharply. AI assistants appear to be converging on similar thresholds. Importantly, the recency of reviews matters as much as the rating. BrightLocal found that a few detailed 4-star reviews from the past week outperform a large volume of 5-star reviews from 2021. The same pattern appears to hold for AI recommendation consideration.

This report presents findings from Found by AI's February 2026 analysis of 200 local business queries run across ChatGPT (GPT-4o with browsing), Perplexity AI Pro, Google AI Overviews, Claude (claude-sonnet-4-6 with web access), and Google Gemini Advanced. Queries covered 10 business categories across 20 European cities. Findings are cross-referenced against published research from SOCi (2025 AI Visibility Report, 3.2M queries, 350,000 locations), Yext (October 2025, 6.8M citations, 1.6M responses), BrightLocal (2025 AI search listings study), Local Falcon (2024 ChatGPT data sources whitepaper; April 2025 AI Overviews whitepaper), and the Princeton GEO study (KDD 2024).

Found by AI provides AI visibility tracking, content creation, and directory management for European SMBs. The free AI Visibility Check shows how a business currently appears across the AI platforms covered in this study, and AI Monitoring tracks changes over time.