Thomas David Jacob digital agency blog
Back to Blog

How AI Assistants Decide Which Local Businesses to Recommend

Thomas David Jacob Team

When someone asks an AI assistant for “a good electrician in Lake Oswego” or “a wedding photographer near Portland,” it names a few businesses. How does it choose? The honest answer is that ChatGPT, Gemini, Perplexity, Copilot, and Claude do not publish a ranking formula, and anyone who claims to know it exactly is guessing.

What we can do is separate three things: what the companies have documented, what shows up consistently when you test it, and what is reasonable inference. This post keeps those three apart. For the bigger picture, start with AI search vs Google.

What Is Actually Documented

A few facts are on the record, and they are more useful than they sound.

  • Google says there is no special trick. Its documentation states there are no additional requirements to appear in AI Overviews or AI Mode, no special optimizations, and no special schema needed. A page needs to be indexed and eligible to show in Search with a snippet. That tells us AI features in Google are built on the same crawl, index, and quality systems as regular search.
  • ChatGPT search has its own crawler.OpenAI's OAI-SearchBot surfaces websites in ChatGPT's search features, and OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. GPTBot is a different crawler, used for training. You can allow one and block the other. Blocking GPTBot does not remove you from ChatGPT search.
  • Microsoft shows you your citations. Bing Webmaster Tools added an AI Performance report in public preview in February 2026, showing how often your content is cited in Copilot and other Microsoft AI answers. It only works if your site is verified there, and a site that blocks AI crawlers will not see data.

The pattern: the assistants answer from web content they can reach and read. If they can't reach you, you can't be named.

What You Can Reasonably Infer

Beyond the documented facts, an assistant is trying to give a safe, specific answer. That means it favors businesses where the evidence agrees with itself. These are inferences from how the systems work and from what we see when we test them, not guarantees.

1. Your basic facts are consistent everywhere

Business name, address, phone, hours, and services should match across your website, Google Business Profile, and major directories. When the sources disagree, a cautious system has less reason to trust any of them.

2. Your website says plainly what you do and where

A page titled “Services” with three vague sentences gives an assistant almost nothing to quote. A page that says “We install and repair tankless water heaters in Oregon City, West Linn, and Lake Oswego” gives it a sentence it can lift. Specific beats clever.

3. Other people say it too

Reviews, local directories, chamber and association listings, and press mentions all describe your business in someone else's words. An assistant weighing “who is good?” looks for that outside confirmation, the same way a person would ask a friend.

4. You answer real questions

Customers ask assistants full questions: how much notice does a photographer need, does a loan officer work with first-time buyers, is there parking at the restaurant. Pages that answer those questions directly, with a real person's expertise behind them, are easier to cite than pages that only list services.

5. Your site can be read at all

Text locked in images or PDFs, content that only loads after heavy scripts run, and robots.txt rules that block search crawlers all make you harder to read. This is the cheapest fix on the list and the one most often missed.

What We Do Not Know

We don't know exactly how much weight any assistant gives reviews versus website text versus third-party mentions. We don't know how often the models change their behavior. Answers also differ between ChatGPT, Gemini, Perplexity, Copilot, and Claude, and between the same assistant on different days. So we don't sell tricks. We fix the foundation, then test and re-test, and we tell clients what we see.

Things That Do Not Help

  • An llms.txt file or similar “AI file.” Google says you don't need new machine-readable files to appear in its AI features, and we don't treat any such file as a guarantee anywhere.
  • Special “AI schema.” Normal structured data that matches your visible page is fine. It is not a switch that makes assistants recommend you.
  • Stuffing your page with questions and keywords. It reads badly to people and to machines.
  • Fake or purchased reviews. The FTC finalized a rule in 2024 banning fake reviews and testimonials, and platforms detect them.

Where to Start

Check what the assistants currently say about you. We wrote a step-by-step guide: How to Check Whether ChatGPT, Gemini, and Perplexity Recommend Your Business. If you want the same check done thoroughly, with a prioritized fix list, that is what the audit in our AI Visibility & Authority service covers. Questions? Get in touch.


Sources: Google Search Central, AI features documentation; OpenAI crawler documentation (OAI-SearchBot and GPTBot); Bing Webmaster Tools AI Performance report announcement, February 2026; FTC fake reviews rule, 2024.

Related Reading

Ready to Grow Your Business Online?

The Thomas David Jacob team works with businesses across Oregon City, Portland, and the greater metro area. Let's talk about what we can do for yours.

Get a Free Consultation