Why ChatGPT Recommends Some Businesses and Ignores Others

Try this before you read the rest of this post. Open ChatGPT and ask it the question your customers would ask. Not your business name, the question. “Who does landscaping in Timmins?” “Best accountant for small construction companies near Barrie?”

If you got named, congratulations, you can stop reading and go enjoy your day. If a competitor got named and you did not, keep going, because I want to show you exactly why that happened. It is not random and it is not rigged. After 20-plus years of watching businesses fight over Google rankings, I find this new contest refreshingly honest. The businesses AI recommends are almost always the ones with the clearest, most consistent, most corroborated information on the public web.

Which is a thing you can fix.

How the recommendation actually gets made

When you ask ChatGPT for a recommendation, something pretty unmysterious happens behind the curtain.

For current, local questions, it runs a web search. That search leans heavily on Bing’s index. It pulls back a handful of pages that look relevant: business websites, directories, review sites, local articles, “best of” lists. Then the model reads what it fetched and writes an answer, usually naming the two or three businesses the sources describe most clearly and most positively.

Notice what did not happen. Nobody checked how pretty your website is. Nobody watched your promo video. The model read text, cross-referenced it, and repeated what the most trustworthy text agreed on.

So the entire game is: what does the text of the internet say about you, and does it agree with itself?

The traits recommended businesses share

I have run this test dozens of times for clients and prospects, and the businesses that come up over and over share five traits. None of them are glamorous.

1. Their website says what they do in plain words

The winners have a homepage that reads like an answer. “We are a family-run plumbing company serving Sault Ste. Marie and Algoma since 1998. We handle residential service calls, renos, and new builds.” A machine can lift that whole sentence into an answer.

The losers have “Excellence in every drop” over a stock photo. I am not exaggerating. There are service businesses whose sites never state their city in text a machine can read. The AI cannot recommend what it cannot parse. This is the single most common failure, and the cheapest to fix.

2. They exist beyond their own website

An AI is careful about vouching for a business on that business’s word alone. Anyone can claim anything on their own site. What tips the scales is corroboration: Google and Bing business profiles, reviews, a chamber of commerce listing, a local news mention, a supplier page that names them.

Each of those is a second witness. You do not need fifty. Even four or five consistent third-party sources put you miles ahead of a competitor who exists only on their own domain.

3. Their information agrees with itself

Same business name everywhere. Same phone number. Same service area. It sounds trivial, but engines treat contradiction as risk. If your site says “serving all of Northern Ontario” and your Google profile says one town, or half the directories list your old address, the model has to guess, and models that are unsure leave you out rather than risk a wrong answer.

The cleanup is tedious. It is also free, and it is one of the highest-return afternoons you can spend on your marketing.

4. Their reviews tell a story, not just a score

Models read review text, not just star counts. Twenty reviews that specifically mention “rebuilt our deck,” “showed up when they said,” “fair price” give the model concrete, quotable evidence about what you do and how you do it. A 4.9 with three reviews that say “great!!” gives it almost nothing.

You influence this by how you ask. When a happy customer offers to leave a review, ask them to mention the specific job. Most will.

5. Their site is technically readable

Crawlers behind AI search have to fetch and parse your pages. Sites that block bots in robots.txt, bury text in images, or take eight seconds to load get skipped. This is table stakes, and it is the same table stakes SEO has always had. If you have not checked in a while, my free speed test and AI search grader will show you where you stand in a couple of minutes.

What does not work

Because every new channel attracts snake oil, a quick list of things that will not get you recommended:

• Stuffing “AI” keywords into your pages. The model is not searching for the word ChatGPT on your site. It is searching for evidence you are good at what you do.
• Fake reviews. Review platforms are getting better at catching them, and a pile of vague five-star reviews from accounts with no history reads as noise at best.
• A hundred auto-generated location pages. Thin, duplicated content damages the trust you are trying to build with both Google and AI engines.

There is no trick, and I mean that as encouragement. The absence of a trick means a small honest business can beat a bigger sloppier one, which is not always true in paid channels.

Make it a habit

The businesses that will own AI recommendations in two years are the ones checking today. Monthly, five minutes:

1. Ask ChatGPT, Perplexity, and Google the three or four questions your customers actually ask.
2. Note who gets named. That list is your real competition.
3. Ask each engine “what do you know about [your business name]” and read for errors. Wrong info means inconsistent sources, and now you know what to go fix.

This post is part of my AI search series. The full pillar guide covers the complete playbook, and if you would rather point at the problem and have someone else fix it, my AI search optimization service exists for exactly that.