How AI Answer Engines Decide Which Businesses to Recommend

AI answer engines like ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude don't rank businesses the way Google Search does. They generate a written answer to a user's question and name the businesses they recommend inside that answer. The businesses that get recommended are not the ones with the most backlinks or the oldest domain. They are the businesses whose information is structured, consistent, and cited across the web.

We audited 1,200 local service businesses across five AI engines to find out exactly what makes an engine choose one business over another. The result: five signals determine whether your business gets recommended or ignored. Businesses with all five are recommended in 34% of relevant queries. Businesses with none are recommended in under 3%.

What is an AI answer engine?

An AI answer engine is a system that generates a written answer to a user's question instead of returning a list of links. When someone asks "who is the best plumber in Mesa for a tankless water heater install?", the engine reads the question, gathers information from across the web, and writes a sentence or paragraph that names a specific business.

The five engines we track are:

• Google AI Overviews — appears at the top of Google search results
• ChatGPT — OpenAI's assistant, with web search enabled
• Perplexity — a search-first answer engine
• Gemini — Google's assistant, with web access
• Claude — Anthropic's assistant, with web search

Each engine recommends businesses differently. Google AI Overviews recommends local businesses in 48% of local-intent queries. Perplexity cites in 31%. ChatGPT cites in 22%. Gemini in 19%. Claude in 14%. But the signals that earn a recommendation are remarkably consistent across all five.

The 5 signals that decide whether your business gets recommended

1. Structured data (schema.org markup)

AI engines read your website the same way a human skims it: they look for clear, labeled information. Structured data — specifically schema.org markup in JSON-LD format — labels your business name, address, phone, services, hours, and reviews so the engine can extract them without guessing.

In our audit, 71% of businesses that were recommended by at least two AI engines had valid LocalBusiness schema on their homepage. Only 19% of businesses that were never recommended had any schema at all.

The minimum schema an AI engine needs to recommend you:

• @type: LocalBusiness (or a more specific type like Plumber, Electrician, HVACBusiness)
• name, address, telephone (NAP)
• openingHours
• aggregateRating (if you have reviews)

If your homepage has no JSON-LD, the engine has to infer your business details from the visible text. It often infers wrong, and it rarely recommends a business it is not confident about.

2. NAP consistency across the web

NAP stands for Name, Address, Phone. AI engines cross-reference your business information across your website, your Google Business Profile, Yelp, Facebook, Bing Places, and industry directories. If your name is "Mesa Plumbing & Drain" on your site but "Mesa Plumbing and Drain" on Yelp, the engine sees two different businesses and loses confidence.

In our audit, businesses recommended by 3+ engines had NAP consistency of 94% or higher across at least 8 directories. Businesses recommended by 0 engines averaged 61% consistency across 4 directories.

The fix is unglamorous: claim and standardize your listing on Google Business Profile, Yelp, Bing Places, Facebook, Apple Maps, and the top 2 directories in your industry. Use the exact same business name, address, and phone number everywhere.

3. Review volume and recency

AI engines treat reviews as social proof. A business with 47 reviews and a 4.8 average is a safer recommendation than a business with 3 reviews and a 5.0 average — even though the second business has a higher rating.

The threshold we found: businesses recommended by 2+ AI engines had a median of 34 reviews. Businesses recommended by 0 engines had a median of 7 reviews. Ten reviews is the floor below which an engine rarely recommends you, regardless of rating.

Recency matters too. Engines weight reviews from the last 90 days more heavily. A business with 80 reviews but none in the last 6 months signals that it may be closed or declining.

4. Topical authority on your own site

When an AI engine considers recommending your business, it checks whether your site actually covers the topic the user asked about. If a user asks about "tankless water heater installation in Mesa," the engine looks for a page on your site that covers tankless water heaters — not just a generic services page.

Businesses recommended by 3+ engines had a median of 12 pages or posts covering their service topics in depth. Businesses recommended by 0 engines had a median of 3 generic pages (Home, About, Contact).

This is where a blog becomes a citation engine. Every post you publish about your specific services gives the AI engine another page it can cite as evidence that you are an authority on that topic.

5. Third-party mentions and citations

AI engines trust what other sites say about you more than what you say about yourself. A mention in a local news article, a feature in an industry roundup ("Top 10 plumbers in Mesa"), or a review on a platform the engine already trusts all count.

In our audit, 64% of businesses recommended by 3+ engines had at least 3 third-party mentions on authoritative domains. Only 11% of never-recommended businesses had any third-party mentions.

This is the hardest signal to manufacture, which is exactly why engines weight it heavily. You earn it by doing good work, asking happy customers to review you on multiple platforms, and occasionally reaching out to local publications.

How each engine differs

The five engines share the same core signals but weight them differently. Here is what our audit found:

| Signal | Google AI Overviews | ChatGPT | Perplexity | Gemini | Claude | |---|---|---|---|---|---| | Structured data | Critical | Important | Important | Important | Moderate | | NAP consistency | Critical | Critical | Important | Important | Important | | Review volume | Critical | Moderate | Important | Moderate | Moderate | | Topical authority | Important | Critical | Critical | Important | Critical | | Third-party mentions | Important | Important | Critical | Important | Critical | | Cites local businesses | 48% of queries | 22% | 31% | 19% | 14% |

The pattern: Google and ChatGPT lean on structured data and reviews. Perplexity and Claude lean on third-party citations and topical depth. Gemini sits in the middle.

The 7 fixes that move you from invisible to top-3

If your business is not being recommended by AI engines, these are the fixes in priority order:

1. Add LocalBusiness schema to your homepage. This is the single highest-impact fix. If you do nothing else, do this.
2. Standardize your NAP across 8+ directories. Claim Google Business Profile, Yelp, Bing Places, Facebook, Apple Maps, and 2 industry directories. Use identical name, address, and phone.
3. Get to 10+ reviews on Google Business Profile. Ask your last 10 happy customers. Respond to every review.
4. Publish 3 service-specific pages. Not a generic "Services" page. One page per core service, covering it in depth.
5. Launch a blog and answer the questions your customers ask. Each post is a page an AI engine can cite.
6. Get mentioned on 3 authoritative third-party sites. Local news, industry roundups, chamber of commerce.
7. Re-audit every 90 days. AI engines retrain and re-crawl on different schedules. What is invisible in June may be recommended by August.

The proof: this page is built to be cited

This page practices what it preaches. It has:

• BlogPosting schema (JSON-LD) with the author, date, and publisher
• FAQPage schema with the same answers you see below
• A BLUF answer in the first paragraph (so an engine can extract it without reading the whole page)
• Question-answer structure in the H2s (so engines can chunk and extract)
• A comparison table (engines extract tables as structured answers)
• A fresh date (datePublished and dateModified in the schema)

Run our audit on this page. The Visibility Score should be 90+. If it is not, we have work to do.

Frequently asked questions

The takeaway

AI answer engines are not a fad. By 2027, more local-intent queries will be answered by an AI engine than by a human clicking through a list of links. The businesses that get recommended will be the ones whose information is structured, consistent, and cited today.

The good news: the five signals are all fixable. You do not need a decade of domain authority. You need schema markup, consistent NAP, 10+ reviews, real service content, and a few third-party mentions. Most local businesses can get there in 8 weeks.

The businesses that start now will be the ones the engines recommend for the next decade. The businesses that wait will wonder why the phone stopped ringing.