How to get your business recommended by ChatGPT

Your Google Business Profile and your reviews. We put them first in every fix plan, because that profile is the record the assistants check to confirm you are a real business doing a named service in a named place, and the rest of the work is harder to credit while that record is thin.

Fill in every field: categories, service area, opening hours, phone number. Add the services you sell and recent photos of the work. Then set up a routine that asks for a review after every finished job. We treat recency as part of the job rather than a nice extra: a profile whose reviews stopped some time ago gives an assistant less to go on than one still collecting them.

The ones they cite in answers for your city, which you can read off the answers themselves. Perplexity and ChatGPT show their sources. In the answers we have stored for bookkeepers, the sites turning up include the local chamber of commerce, Bark, LinkedIn and industry software directories. Your industry will have its own set, so check rather than copy that list.

Run the questions a buyer would ask, open the citations, and list every site that appears. Claim your entry on each one, and keep the name, address, phone and service wording identical to your Google profile so there is nothing to contradict.

One for each question a buyer asks, because pages that answer one specific question are what assistants quote. With no page on a topic, an answer about that topic has nothing of yours to lean on, no matter how much of that work you do.

List the questions you get asked on a first call: what it costs, how long it takes, what happens if the filing is late, do you work with my software, do you cover my suburb. Give each one a page or a clearly headed section, phrase the heading the way the buyer phrases it, and answer in the first two sentences rather than circling for three paragraphs.

• One page per service, one per topic you want to be named for.
• Headings written as questions in the buyer's words, not internal service names.
• The direct answer first, detail after it.
• Named client results where you have them, with the specifics you are allowed to publish.

Yes, and the title tag is the piece to get right: it should contain your service and your city. Clever taglines read well to people who already know you and tell an assistant nothing it can use.

The same goes for the first paragraph on the page. A sentence naming the work, the kind of client and the area beats a paragraph about commitment to excellence, because the assistant is looking for something it can repeat in an answer.

Structured data is markup that tells a machine what it is reading instead of leaving it to guess. Add LocalBusiness markup with your name, address, phone, hours and service area, and FAQ markup on the question pages you just wrote.

Then publish an llms.txt file at your site root: a plain text file saying what your business does, where it operates, what each important page covers, and where to find it. It is a small file, and it puts those facts in one place an assistant can read.

Open your robots.txt and read it before you do anything else on this list, because a blocked crawler makes the rest of the work invisible. If a plugin or a security setting has disallowed the AI user agents, nothing you publish can reach them.

Then claim and verify a Bing Places listing. ChatGPT's web search draws on Bing's index, so your Bing listing feeds that assistant directly.

Two kinds of page do this work: best-in-city roundup articles, and public forum threads where a person asks for a recommendation and gets names back. Assistants lean on pages where somebody other than you vouches for you.

For roundups, find the ones that already rank for best in your city, read how they pick firms, and ask to be considered with the details they need. For forums, be present where your buyers ask, answer questions as yourself without pitching, and let clients mention you rather than doing it for them. Astroturfing is easy to spot and the thread will turn on you.

Count mentions across many questions and several runs. Answers vary run to run, so one good answer after a fix proves nothing you can build on, and a fix that worked can look like noise for a week.

Counted from our production database on 16 September 2026, we had 5,900+ local businesses tracked, 17,000+ AI recommendations analyzed and 16 US cities tested.

A free check counts your name across those 48 buyer questions in all four assistants, shows you who was named instead of you, and writes out the first fix. Starter at $29 a month is built for the fixing stage: weekly checks, the full fix plan, the questions you gained and lost each week, and a confirmation once each fix is live.