How to get your business recommended by ChatGPT and Gemini.

How ChatGPT and Gemini choose which businesses to name.

When someone asks ChatGPT or Gemini what to buy, the assistant answers from two places. One is what the model learned in training. The other is a live web search it runs before it writes, on the questions where it decides fresh information will help. Google describes this for its own AI features as a "query fan-out": several related searches, run across subtopics, to build one answer (Google Search Central).

The second route is the one you can influence this month. An assistant that searches reads the pages its searches return and names the businesses those pages describe. A business with no page that matches the search has nothing for the assistant to read, so it has nothing to name.

We asked Gemini this exact question, "How do I get my business recommended by ChatGPT and Gemini?", on 27 September 2026. Before it answered, it ran five searches, among them "how to get business recommended by chatgpt", "gemini business recommendation criteria" and "strategies to get ai chatbots to recommend businesses". It then cited eight sources, from agency blogs to a Google support thread and a Reddit discussion. That one answer is a small sample, and it shows the mechanism well: a question in, searches out, and the pages those searches found shaping the reply.

Make sure the assistants can read your site.

Nothing else on this list works if a crawler cannot read your pages. Three checks come first.

• Allow the crawlers. OpenAI's documentation says that to appear in ChatGPT's search results, a site should allow OAI-SearchBot in its robots.txt (OpenAI). For Google's AI Overviews and AI Mode, a page has to be indexed and eligible to show in Google Search with a snippet (Google Search Central).
• Serve real HTML. A site that builds its pages in the browser can look empty to a crawler that does not run JavaScript. Frameworks that render on the server, or services that pre-render pages for crawlers, fix that.
• Put the title and description in the head. We found our own site sending its title after the body had started, so the snippet Google showed was built from footer links. Check the HTML your server returns, not the page your browser shows.

Answer the questions your buyers ask, one page per question.

An assistant searches for the question it was asked, often almost word for word. The searches Gemini ran for "best AI recommendation tracking tools for marketers" were phrases like "best ai recommendation tracking tools for business marketing". A page that answers that phrase is a page it can find.

So work from the questions, not from your product. For each buying question in your category, ask whether you have a page that answers it directly:

• a list of the options for "best X for Y" questions, with yours among them and the others described fairly;
• a comparison for "X vs Y" and "alternative to X" questions;
• a guide for "how do I" questions, with the steps in order;
• a definition for "what is" questions.

Put the answer in the first paragraph, name who it is for, and give the facts a buyer compares: prices, limits, what is included. Pages that state specifics give an assistant something to quote.

Get onto the pages the assistants already trust.

Your own site is one of many sources. When we tracked four questions about AI visibility tools over ten days, the pages Gemini cited in 10 or more of 42 answers were all lists of tools published by third parties, from Zapier's roundups to independent marketing blogs. An answer built from those pages can only name the businesses they mention.

So find the lists in your category and make sure they describe you:

• Collect the pages the assistants cite for your buying questions. How to find which websites AI search engines cite in your industry covers how.
• Where a list includes competitors and not you, contact the publisher with the facts they would need to add you.
• Keep your listings on review sites and directories accurate. When we asked Gemini where AI answers come from, it pointed to review aggregators such as G2 and Capterra for B2B software, and said some platforms favor YouTube and Reddit.

Keep the facts about you consistent.

An assistant that names you still has to describe you, and it takes the description from what it read. If your pricing page says one thing, a two-year-old review says another and a directory lists a plan you retired, the answer can repeat any of them.

• Keep one current page for prices and plans, and update it the day they change.
• Correct outdated listings on directories and review sites.
• Use the same name for your business everywhere, and say plainly what it does in the first sentence of your homepage.

For a local business, Google's own data carries extra weight. When we asked Gemini how to get recommended, it told businesses with a physical location to keep their Google Business Profile complete and accurate, and called it a significant signal for local recommendations. A complete profile with current hours, categories and reviews is part of the same work.

Measure the questions, then repeat.

A single check tells you little, because the same question gets a different answer the next day. Pick 10 to 20 buying questions, ask them on a schedule, and record who each answer named and which pages it cited. How to track whether AI recommends your business covers the method, and how to measure whether AI recommendations are working covers the numbers.

Then sort the questions into the ones you can act on. Where a competitor was named from a page the answer cited, write the page that answers the same question better. Where nobody in your category was named, a clear page is often the only candidate the assistant finds. Where the assistant answered without searching, no page published this month changes the answer, so spend your time on the others.

Where to start this week.

If you do five things this week, make them these:

1. Open your robots.txt and confirm it allows OAI-SearchBot and Googlebot. Then view the raw HTML of your home page and confirm the title and description sit in the head.
2. Write down ten buying questions and ask each in ChatGPT and Gemini. Note who was named and which pages were cited.
3. Pick the question where a competitor was named from a page you could match or beat, and write that page.
4. Contact the publisher of the most-cited list that leaves you out, with the facts they need to add you.
5. Put a date two weeks out to ask the same ten questions again.

What does not work.

Some tactics circulate that waste time:

• Stuffing pages with the assistant's name. Mentioning ChatGPT on your page does not make ChatGPT mention you.
• Hidden text and keyword lists. Assistants quote what a reader can see; a block of keywords gives them nothing to quote.
• Buying placement in the answer. The organic answer does not sell positions. Where an assistant sells ads, they appear as sponsored placements beside it.
• Publishing dozens of thin pages at once. One page that fully answers a question does more than ten that each answer it partly, and search engines treat mass-produced pages with suspicion.

The rest is the same loop, repeated: find the questions, publish the page that answers each one, get described accurately on the pages the assistants trust, and measure again.