Yext analyzed 155.5 million AI citations across 1,623 brands and published the results in July 2026. The headline finding: 86% of the sources AI engines cite when recommending a business are things the business owner can update directly. First-party websites generated 44% of citations. Business listings generated another 42%. Reviews and social made up 8%.

That number reframes the conversation for a plumber in Passaic or an HVAC shop in Bergen County who thinks AI visibility is a black box. How ChatGPT picks local businesses runs a five-step pipeline, but the inputs are mostly things a contractor already owns. The inputs are the website, the Google Business Profile, the Yelp page, and the Bing Places listing. The business listings versus directories gap that most trades shops leave open.

What Yext actually measured

Yext tracked citations across four AI models: ChatGPT, Gemini, Perplexity, and Anthropic. The study pulled from 155.5 million citation events. For each one, Yext recorded the source URL the AI cited when answering a local query.

The breakdown:

  • First-party websites: 44% (2.9 million citations)
  • Business listings: 42% (2.9 million citations)
  • Reviews and social: 8% (545,000 citations)
  • Other sources: 6%

Each model weighted the three buckets differently. Gemini favored websites at 52.1% of its citations. OpenAI leaned on listings at 48.7%. Perplexity spread citations across MapQuest, TripAdvisor, and a wider directory set. Bing Places for plumbers matters more for ChatGPT than for Gemini because OpenAI pulls from Bing's index. A trades business that fixes both the website and the listings covers all three engines.

Why this number matters for trades businesses

The 5WPR HVAC and Plumbing AI Visibility Index, published the same month, surveyed approximately 230,000 contractors across the United States. Their finding: 87% of independent HVAC and plumbing contractors have effectively zero AI citation share in their own metro. Only 50 contractors out of 230,000 had what 5WPR called meaningful citation presence.

The gap is not advertising budget. The gap is the 86% that Yext identified. A contractor with a Google Business Profile that says "HVAC services" instead of "HVAC repair and installation in Passaic County, NJ" has left the largest single citation source untuned. A contractor whose website has no LocalBusiness schema has given Gemini nothing structured to read. A contractor whose Yelp name says "Smith Heating & Cooling LLC" while the Google listing says "Smith HVAC" has split the entity across two names, and the AI cannot resolve which one to cite.

The Google Business Profile is the AI strategy because it is the listing that every model checks first. Yext's data puts a number on that. Listings account for 42% of all citations, and the Google listing is the most-cited listing of all.

The three fixes that cover 86%

A trades owner who wants to move from invisible to cited does not need a new website or an agency. The work falls into three steps.

1. Complete the Google Business Profile. Claim it if it is unclaimed. Fill every field: service area, hours, service descriptions, attributes, photos. Set the primary category to the specific trade, not "Contractor." A plumber whose profile says "Plumbing repair" gets cited for plumbing repair queries. A plumber whose profile says "Home services" gets cited for nothing.

2. Standardize NAP across every listing. Pick one name, one address, one phone number. Push that combination to Google, Yelp, Bing Places, Apple Maps, Angi, BBB, and Yellow Pages. NAP consistency is how AI engines resolve whether two listings point to the same business. A mismatched name or phone breaks the resolution, and the model skips the shop.

3. Add LocalBusiness schema to the website. JSON-LD in the homepage and every service page. Include business name, address, phone, hours, service area, and aggregate rating. This is the structured data Gemini reads first. Without it, the website is prose. With it, the website is a citation source.

Those three fixes cover the website (44%) and the listings (42%). That is the 86%.

The other 14%

Reviews and social account for 8% of citations. Other sources account for 6%. The 8% is small relative to the 86%, but reviews are a gatekeeper, not the primary citation source. SOCi's 2026 Local Visibility Index found that ChatGPT applies a 4.3-star average as a floor before it will recommend a business. A contractor with a completed profile, consistent NAP, and structured schema still gets filtered out if the rating sits at 3.8.

The fix is to ask every satisfied customer for a review that names the service and the city. "Fixed the burst pipe in our Clifton basement within an hour" tells the model the trade, the town, and the quality. "Great service" builds a number but not an entity.

Yext's consumer survey adds the demand side. In their 2026 Consumer Search Behaviors Report, 28% of respondents said they tried a new local business specifically because an AI tool recommended it. Among AI users, 75% rated their trust in AI local recommendations at 4 or 5 out of 5. Homeowners are asking. The question is whether the AI can find the shop.

What to check this week

Search the trade and the city from two different zip codes. "Best plumber in Paterson NJ" and "emergency AC repair in Clifton NJ" are the prompts homeowners actually type. Write down who the AI names.

If the shop is not on the list, open the Google Business Profile and check three things: is it claimed, is the category specific, and does the service description name the city. Then check whether the website has LocalBusiness schema with the same NAP. Then search the business name on Yelp and Bing Places to confirm the listings match.

That audit takes 20 minutes. It covers 86% of what the AI uses to decide.

Run a free AI visibility audit to check all 12 signals across ChatGPT, Gemini, and Perplexity in one pass.

FAQ

What percentage of AI citations come from sources a business already controls?

86%. Yext analyzed 155.5 million AI citations across 1,623 brands and found that 44% came from first-party websites and 42% came from business listings. Combined, those are sources a business owner can update directly without paying an agency or waiting for a third party.

Which AI model favors websites the most for local recommendations?

Gemini pulls 52.1% of its citations from first-party websites, the highest share among the models Yext tracked. OpenAI leans more on listings at 48.7%. Perplexity spreads citations across MapQuest, TripAdvisor, and other directories. A trades business that fixes both its website and its listings covers all three.

What should a trades contractor fix first for AI visibility?

Claim and complete the Google Business Profile first. Then standardize the business name, address, and phone across every directory. Then add LocalBusiness schema to the website with the same NAP. Those three steps cover 86% of the sources AI uses to decide whether to name a business.

Do reviews matter for AI citations if 86% come from websites and listings?

Reviews still matter. Yext counted 8% of citations from review and social sources, and ChatGPT applies a 4.3-star average as a floor before it will recommend a business. Reviews are a gatekeeper, not the primary citation source. Fix the 86% first, then build review velocity.

How many people try a business because of an AI recommendation?

28% of consumers in Yext's 2026 Consumer Search Behaviors Report tried a new local business specifically because an AI tool recommended it. Among AI users, 75% rate their trust in AI local recommendations at 4 or 5 out of 5.

Run your free AI visibility audit at firstcallvisibility.com.