A study of 28.5 million AI-generated answers reveals which platform drives local AI discovery — and why 91–97% of citations come from searches where buyers don’t yet know your name.
The study: Foundation Inc & AirOps analyzed 28.5 million AI-generated local business answers. Yelp earned 512,680 citations in Q4 2025, capturing 72.5% of Google AI Mode local citations and 62.1% in Perplexity.
The query breakdown: Between 91% and 97% of those citations came from unbranded category searches — queries like “best plumber in Austin” — not brand-name searches.
The confirmation: Google’s May 2026 Core Update rewarded canonical reference brands +10–24% while penalizing generic aggregators −13–29%. The same pattern: entity coherence and trust signals win.
Most local businesses think of AI search as an extension of Google rankings. Earn a higher position, appear more often in AI answers. The research does not support this view.
Two independent studies published in June 2026 tell a different story — one about which platform actually drives local AI discovery, and one about the type of search queries that trigger local AI recommendations. Both findings have direct implications for any local service business investing in search visibility.
Finding 01
The Foundation and AirOps analyzed 28.5 million AI-generated answers to local business queries. Their primary finding: Yelp earned 512,680 citations in Q4 2025 and captured 72.5% of local discovery citations in Google AI Mode and 62.1% in Perplexity.
Foundation Inc & AirOps · 28.5M AI-Generated Local Answers Analyzed
Of Google AI Mode local discovery citations run through Yelp —
making it the dominant platform for AI local recommendations
Yelp citation volume grew 19x between September and November 2025,
while the total AI citation market grew roughly 3x.
The growth curve reinforces why this matters: Yelp citation volume grew 19 times between September and November 2025, while the total AI citation market grew roughly 3 times. Yelp was growing citation share 6 times faster than the overall market.
For Google AI Mode specifically, Yelp is not just a strong citation source — it is effectively the default. When someone asks Google AI to recommend a local plumber, water damage company, or medical spa, the answer is overwhelmingly likely to draw from Yelp profile data.
Finding 02
The query-level breakdown of those 512,680 Yelp citations reveals something more important than the raw volume: between 91% and 97% of AI citations came from unbranded category searches — queries like “best plumber in Austin” or “water damage restoration near me” — not from people searching for a specific business by name.
Foundation Inc & AirOps · Query-Level Citation Breakdown
Of Yelp’s AI citations came from unbranded category searches —
queries from buyers who do not yet know which business they want
Brand-name queries accounted for only 3–9% of citations
and produced roughly half the citation density of category searches.
Brand-name queries — searching for a specific business by name — accounted for only 3–9% of citations and produced roughly half the citation density of category searches. The AI local recommendation moment is almost entirely a discovery event, not a validation event.
This reshapes where local businesses should focus their citation-building effort. Optimizing for brand-name AI visibility is a late-funnel tactic. The early-funnel opportunity — where AI shapes which businesses buyers even consider — runs primarily through unbranded category citations on platforms like Yelp, Angi, and BBB.
Finding 03
Google’s May 2026 Core Update completed on June 2. Independent analysis by Aleyda Solis using SISTRIX data found the update rewarded a specific type of source.
| Source Type | UK Visibility Change | US Visibility Change |
|---|---|---|
| Canonical reference brands — authoritative, entity-coherent, well-reviewed local sources | +24% | +10% |
| Reference aggregators and tools — generic content without genuine topical authority | −29% | −13% |
| Forum and Q&A surfaces — low-quality community content without authentic participation signals | −24% | −12% |
| Local-market retailers (UK) — locally-relevant ccTLD sources | +20% | n/a |
| US .com marketplaces in the UK index — internationally-focused aggregators serving the wrong market | −55% | n/a |
A “canonical reference brand” is a business that Google’s systems recognize as the authoritative, consistent, trustworthy source for its category in its location. Consistent business information across all platforms, strong review signals, genuine local citations, and clear entity coherence are the characteristics that define this category.
The Foundation/AirOps local AI citation study and Google’s May core update describe the same underlying mechanism from two different angles. AI systems — whether Google’s or ChatGPT’s — are applying a trust and authority filter to local recommendations. The businesses that pass that filter are the ones with consistent entity signals, strong third-party platform presence, and genuine review depth.
The businesses that fail it — thin profiles, inconsistent NAP, sparse reviews, generic content — are not being recommended, regardless of how they rank in traditional organic search.
Given Yelp’s 72.5% share of Google AI Mode local citations, a business with a complete, well-reviewed, accurate Google Business Profile but a thin or unclaimed Yelp profile has a major structural AI visibility gap. Both platforms feed the local AI recommendation layer. Neither can be treated as optional.
Claimed and verified. Primary category matches GBP. Hours current and accurate. At least 10 reviews with a recency signal (3+ in the last 90 days). Photos updated in the last 6 months.
Based on citation volume data across AI platforms, the priority order for local service businesses is: Yelp (72.5% of Google AI Mode local citations) → Angi / HomeAdvisor (home services) → BBB (credibility signal across all categories) → Apple Maps → Bing Places (feeds ChatGPT recommendations) → Facebook → Industry-specific directories.
With 91–97% of AI local citations triggered by unbranded category searches, the question to ask is not “does AI mention my business when someone searches for my name?” The question is: “does AI recommend my business when someone searches for my category in my city?”
Start with category + city prompts: “best [service] in [city],” “[service] near [neighborhood],” “who should I call for [emergency service] in [city].” These are the prompts that drive the 91–97% of citations that actually shape buyer decisions.
Google’s May core update explicitly penalized generic aggregators and rewarded authoritative, entity-coherent sources. The same pattern appears in the SISTRIX ChatGPT citation tracking: when OpenAI updated to GPT-5.5, generic travel and booking aggregators lost 50–60% of their citation volume while local editorial sources and authentic community platforms gained.
Across every AI platform, the direction is the same. Consistent, verifiable entity information across all platforms is the durable foundation. Shortcuts that manufacture the appearance of authority without the underlying signals are being detected and discounted at model-update frequency.
We test your visibility across ChatGPT, Perplexity, Gemini, and Google AI Mode for your primary service categories — and show exactly where the gaps are.
Independent, data-backed • No contracts • Results verifiable in your own analytics