← Writing

How a small business gets recommended by ChatGPT

By Israel Traore · October 16, 2026

More of your customers now ask an AI assistant before they ask Google: "Who's a good immigration lawyer in New Jersey?" "What's the best software for a small garage?" The assistant answers with two or three names. If yours isn't one of them, you never had a chance.

Getting into that answer has a name: GEO, generative engine optimization. It's the AI-era cousin of SEO, and most of it is work a small business can do itself.

How an AI assistant decides who to recommend

An assistant like ChatGPT, Gemini, Claude or Perplexity knows about a business in two ways:

1. What it learned in training. A large snapshot of the web, taken months before the model was released. You can't edit it, and it changes slowly.
2. What it finds when you ask. For "who's good near me" questions, most assistants now search the web, read the top results and summarize what those sources agree on. ChatGPT's search leans on Bing, Gemini on Google, and Claude and Perplexity run their own searches.

The second one is where a small business can win, and fast. The assistant repeats what several trusted pages say in agreement. One page about you is a claim. Five consistent pages from different places are a fact.

The six things that move it

1. Be one consistent entity everywhere. Same business name, same one-sentence description, same town and phone on your website, Google Business Profile, Bing Places, Apple Business Connect, LinkedIn, Yelp and your industry directories. When sources disagree, the assistant hedges or leaves you out.
2. Claim Bing Places. Almost everyone forgets it, and it feeds ChatGPT's search.
3. Get talked about by others. Reviews from real customers, a local news mention, a podcast, a comparison article, an honest answer you wrote in a forum. Assistants trust third parties more than your own site.
4. Answer the exact question on your site. If people ask "who does X for Y in Z", have a page that says plainly "We do X for Y in Z", with an FAQ, real numbers and real examples. Structured data (schema.org markup) helps machines read it.
5. Let the AI crawlers in. Some hosting and security settings block AI bots by default. If GPTBot, ClaudeBot or PerplexityBot can't read your site, you don't exist to them. Check your firewall and robots.txt.
6. Measure every month. Ask each assistant the 10–15 questions your buyers would ask, with memory and personalization off, and write down who it names. That list is your scoreboard.

What it looks like in practice

When I started on this for my own company, ChatGPT and Gemini could already describe our product correctly when asked about it by name — our own website was doing its job. But asked "what's the best auto repair shop software in New Jersey?", neither named us. They listed the big national brands and leaned on one review site in particular.

That told us exactly where the gap was: not our website, but the absence of third-party sources tying us to the category. So the work became listings, honest comparison pages, reviews from real customers and local press — and a monthly re-check of the same questions.

What not to do

• Don't buy or fake reviews. It's against every platform's rules and, in the United States, the FTC's 2024 rule on fake reviews. It also gets profiles suspended.
• Don't astroturf forums. Posting fake "has anyone tried X?" threads gets you banned, and the backlash becomes part of what the AI reads about you.
• Don't trust anyone who guarantees a recommendation. Nobody controls the models. What you can promise is to be found more often, measurably.