How AI Search Actually Decides Who to Recommend
ChatGPT, Perplexity, and Gemini recommend the local businesses they can find and verify across the public web: consistent name, address, and phone data, complete LocalBusiness schema, a steady flow of real reviews, and citations across directories. Nobody can guarantee placement in an AI answer, but every one of those signals is something you already control.
When someone asks ChatGPT "who's the best bathroom remodeler near me," the model is not flipping through a phone book. It is synthesizing the public web: search indexes, review platforms, directories, your website, your Google Business Profile, and anything else it can retrieve about businesses like yours. ChatGPT's search draws on Bing's index and live web results, Gemini leans on Google's, and Perplexity retrieves pages in real time and cites them. Different plumbing, same principle: an AI can only recommend what it can find and verify.
That produces two practical rules. First, the businesses that get recommended are the ones with a strong, consistent, machine-readable presence across many sources, not the ones that gamed a single ranking factor. Second, nobody can guarantee you a spot in an AI answer, because these systems are probabilistic and the same question asked twice can return different names. Anyone selling guaranteed ChatGPT placement is selling something that does not exist.
The Four Signals That Make You Recommendable
Strip away the noise and AI visibility for a local business comes down to four signals, and every one of them is something you already control:
- Consistent entity data: your name, address, phone, and service area stated identically everywhere, so models are confident you are one real business, not three similar ones
- Structured data: LocalBusiness schema that states who you are, what you do, and where, in the machine-readable language retrieval systems parse first
- Reviews: volume, recency, and substance. AI answers lean heavily on review text and ratings when deciding who deserves a recommendation
- Citations: mentions across directories, local press, and industry sites that corroborate everything else
Consistency Is a Ranking Factor Now
None of the four signals are exotic. They are the same ones that drive Map Pack rankings, which is why businesses that do local SEO well tend to surface in AI answers early. The difference is emphasis: language models reward clarity and consistency even more than Google does, because resolving entities is their entire job. If your site says "Becker Custom Homes," your Yelp page says "Becker Homes LLC," and your profile says "Becker Construction," a human shrugs it off. A model may treat them as three separate businesses and recommend none of them.
This is why the unglamorous work matters so much. Every directory that matches your website exactly is one more piece of evidence an AI can use to confirm you are real, established, and safe to name in an answer.
Schema and llms.txt: The Technical Layer
Schema markup is the closest thing AI search has to a cheat code, and it is a legitimate one. A complete LocalBusiness schema graph, with your real NAP, geo, license info, service areas, and review data, lets retrieval systems disambiguate you instantly. Pages without schema can still get cited, but they are harder to identify with confidence, and AI recommendation is fundamentally a confidence game. Our technical SEO and schema engineering work exists precisely because this layer is invisible when it is right and disqualifying when it is wrong.
You will also hear about llms.txt, a proposed standard file, like robots.txt but for AI, that tells language models what your site is and which pages matter. The honest status in 2026: no major AI provider has confirmed its crawlers read llms.txt, and there is no solid evidence it directly moves citations yet. It also takes about twenty minutes to create, costs nothing, and hurts nothing. Add one with a short business description and links to your key pages, but treat it as a lottery ticket, not a strategy.
What Can a Local Business Do This Month?
This is the part you can act on, in rough priority order:
- Make sure your site is indexed in Bing, not just Google. ChatGPT's search leans on Bing, and many local sites have never submitted there
- Claim and complete your Google Business Profile and Bing Places listing with identical information
- Audit your NAP across the top directories: Yelp, Facebook, Angi, the BBB, Apple Maps. Fix every mismatch
- Add or repair LocalBusiness schema on your homepage and contact page
- Ask your next ten happy customers for reviews, and respond to every review you already have
- Add an FAQ section to your site answering the questions customers actually ask, in plain language. AI systems reward direct question-and-answer content
- Publish one page that plainly states your services and service area, naming the cities you actually work in
Why This Compounds
None of that requires new software or a new website. It requires an afternoon of honest cleanup, and it compounds: every consistent citation and every fresh review makes the next AI answer slightly more likely to include you. The businesses that start now are building a lead that gets harder to close with every month of evidence they accumulate.
What Doesn't Work
A growing industry wants to sell you "AI SEO" shortcuts. Save your money on these:
- Paying for placement in ChatGPT answers. There is no paid placement in organic AI answers
- Prompt-injection gimmicks, like hidden text on your site instructing AI models to recommend you. Search engines and AI providers actively filter this
- Buying reviews to boost your AI reputation. Fake reviews are a suspension-level risk on Google and increasingly easy for platforms to detect
- Rewriting your whole site for "AI optimization." If your site already says clearly what you do, where, and for whom, you have most of what AI needs
- Expecting results in a week. Indexes, citations, and review velocity move in weeks and months
The Pattern Behind Every Shortcut
Every item on that list tries to fool the system instead of informing it, and that is a waste at best and a penalty at worst. AI search punishes the same shortcuts Google spent two decades learning to punish, only faster, because the models inherit the filters.
Find Out Where You Stand
The uncomfortable truth about AI visibility is that most owners have no idea whether ChatGPT recommends them today. The fix is measurement. Ask the AI tools the questions your customers ask ("best kitchen remodeler in Portland," "who reglazes bathtubs near Beaverton") and see who comes up. Then ask why: whose reviews, whose schema, and whose citations are feeding those answers.
Or skip the manual work. Our free ECARAT visibility audit measures your AI-search visibility across ChatGPT, Perplexity, and Gemini alongside your search presence, Google Business Profile, reviews, and structured data, benchmarked against five local competitors, with a prioritized fix list. It is the same assessment behind our AI visibility audit service, and it tells you in one report whether AI search is already sending your customers to someone else.