How to get recommended by ChatGPT: 10 steps from our data

The short version

Ten steps to get your business recommended by ChatGPT. Each one rests on something our daily tracking actually showed, not on theory.

1. Let ChatGPT read your website
2. Say who, what and where at the top of each page
3. Track buying questions, not your brand name
4. Take the questions nobody owns yet
5. Get onto the list sites ChatGPT reads
6. Fix the directories that matter in your country
7. Cover each city separately
8. Study the rival pages ChatGPT reads
9. Write for spoken questions
10. Measure rates over weeks, not single answers

I'll be straight about one thing up front. These steps come from what ChatGPT read and recommended across 1,104 answers for seven businesses. They are not yet a before-and-after experiment. We're applying them to our demo business now and will publish what moves. More on that below.

How we got these steps

The short answer: we tracked who ChatGPT named, and which sites it read, every morning for seven real businesses.

Our tool, LLM Visibility Lens, asks each business's buying questions in ChatGPT every day, logged out, and stores every answer and every source link. From 9 September to 8 October 2026 that came to 1,104 ChatGPT answers for:

• a Vancouver electrician
• a Massachusetts roofer
• a Florida dental training institute
• an online mobility store
• an online sod farm
• two others

Only the electrician, Kato Electrical, is named, because his account is our public demo. Some steps also draw on the same tool's checks in Gemini, Google AI Mode, AI Overviews and Perplexity.

Each step lists the effort involved and the evidence behind it. The method is on the how it works page.

I sell this tracker, so weigh that. The demo is public, so you can check the electrician's numbers yourself.

Want your customers' real buying questions first? The free question finder writes them from your website.

Step 1: Let ChatGPT read your website

Effort: low. Evidence: strong.

When ChatGPT cited a business's own website, it usually recommended that business. The trouble is it rarely read it.

• The electrician was named in 14% of all ChatGPT answers, but in 7 of the 9 answers that cited his own site.
• The mobility store was named in 5% of all answers, and in 10 of 20 that cited its site.
• The sod farm's site was cited in fewer than three answers. It was named once in 204.

What to do:

• Allow OpenAI's crawler. OpenAI's crawler documentation says sites that block OAI-SearchBot won't appear in ChatGPT search answers. Check robots.txt, and check your firewall or CDN too: a "block bots" switch can stop it without telling you.
• Get into Bing. Submit your sitemap in Bing Webmaster Tools. ChatGPT's search uses outside search providers as well as its own crawler.

More detail is in which websites ChatGPT trusts.

Step 2: Say who, what and where at the top of each page

Effort: low. Evidence: medium.

ChatGPT and Google's AI sometimes read your page and then recommend someone else. Usually that's because the page explains the topic but never says plainly who does the work, and where.

We saw it with Kato's emergency page. AI systems used it as a source for emergency questions and then named other electricians; in AI Overviews it was cited in 21 answers and he was named in 7.

What to do:

• In the first two or three lines of every service and area page, say the business name, the service and the places you cover.
• Then answer the customer's actual question: cost, timing, process, who to call.

The named vs cited section of our AI Overviews guide explains the gap in more detail.

Step 3: Track buying questions, not your brand name

Effort: low. Evidence: strong.

AI happily names a brand when you ask about the brand. It's a different story when a buyer asks who to buy from.

In a test of 60 answers about a well-known mobility brand, it was named in 88% of answers to questions containing its name and 0% of ready-to-buy questions.

What to do: list the questions customers ask before they know you, such as "best [service] in [city]", "who can [job] near me", "how much does [job] cost in [city]". Track those, not your name.

Step 4: Take the questions nobody owns yet

Effort: medium. Evidence: strong.

For many buying questions, ChatGPT names nobody at all. That's the cheapest visibility you'll find.

• For "how much does an electrical panel upgrade cost in Vancouver", 38 of 43 answers across AI systems named no business.
• Across all five AI systems, the only time they agreed with each other was when they agreed that nobody wins: 52 question-days in our same-day study.

What to do: find the questions where answers name no one, and write the page that answers each one outright, with real local prices, timing, permits and your name next to the answer. We first wrote about these "unclaimed" questions in We asked AI who to hire 80 times.

Step 5: Get onto the list sites ChatGPT reads

Effort: medium. Evidence: strong.

ChatGPT leans heavily on "best of" list sites. They help enormously when they include you, and not at all when they don't.

For the electrician, named in 14% of answers overall:

• publicleads.com: named in 11 of 11 answers that cited it
• threebestrated.ca: 19 of 30
• industryoversight.ca: 1 of 30

What to do:

• Look at the sources under ChatGPT's answers to your buying questions.
• Note the list sites that keep appearing, and check whether you're on them for your city.
• Before paying any of them, read how they choose businesses. Some are editorial, some take nominations, some sell placement.

Step 6: Fix the directories that matter in your country

Effort: medium. Evidence: medium.

Which directories ChatGPT reads depends on where you are and what you do. A generic list of "top 50 citation sites" wastes time.

• Canadian electrician: when ChatGPT cited HomeStars, he was named in 40% of those answers, against 14% overall.
• US businesses: Angi, HomeAdvisor, BBB and Yellow Pages appeared across several accounts.
• Yelp and Google Maps: each cited in fewer than three ChatGPT answers for every business we track, including the US ones. BrightLocal's study found Yelp far more often across the US, UK and Australia, so check your own sources rather than trusting either study blindly.

What to do: complete and keep consistent the two or three directories ChatGPT actually cites for your trade and country. Use the same name, address, phone and services everywhere, and mirror them in LocalBusiness schema on your site.

Step 7: Cover each city separately

Effort: high. Evidence: strong.

Being recommended in one city says nothing about the next one, even when it's 20 minutes away.

For the same electrician, in ChatGPT:

The difference came from which list sites and pages ChatGPT read for each city, as we showed in Google Business Profile and AI answers.

What to do: for each place you serve, check the questions separately. Make sure your site, your directory profiles and the local list sites all say you serve that place.

Step 8: Study the rival pages ChatGPT reads

Effort: medium. Evidence: strong.

Competitors' websites can push you out, or carry you along, depending on whether ChatGPT already sees you as one of the established names in that area.

• When ChatGPT read a rival electrician's site, Intel Power's, the demo electrician was named in 0 of 64 answers.
• In the roofer's market, answers citing three rival roofers' sites still named the roofer 82–85% of the time.

What to do: find the rival pages ChatGPT keeps citing, and write a more specific, more useful page for the same service and place. Being in the local "known set" is what turns a rival's page from a threat into a neutral.

Step 9: Write for spoken questions

Effort: medium. Evidence: medium.

People increasingly ask AI the way they talk: "Is there an electrician open right now in Port Coquitlam that can come out tonight?" Those questions carry conditions, and the answers name businesses that clearly state them.

In one week of tracking, ChatGPT named the electrician in 10 of 60 answers to typed questions, and 0 of 36 spoken ones.

What to do: put the conditions on your pages in plain words: hours, after-hours service, same-day availability, residential or commercial, and the jobs you actually do. See spoken questions.

Step 10: Measure rates over weeks, not single answers

Effort: low. Evidence: strong.

ChatGPT's answer to the same question changes from day to day, so one check proves nothing.

• The same check run four times in one day scored 13, 25, 18 and 20.
• Each AI system flipped the business between named and not named on 9–17% of consecutive days.
• On the same morning, five AI systems agreed only 1 time in 5. More in 5 AIs, same questions, 1,814 answers.

What to do:

• Ask the same buying questions on a fixed schedule for at least three to four weeks.
• Report how often you're named, per AI system, with the number of answers behind it.
• Give any change a few weeks before you judge it.

Our guide to what one free check can and can't do shows how many answers you need, and why there's no rank #1 in ChatGPT explains why positions mislead.

What not to bother with

Some popular tactics didn't show up anywhere in our data.

• Relying on your Google Business Profile for ChatGPT. In the Burnaby study, ChatGPT never linked to Google once. Keep the profile strong for Google's own AI. See Google AI Mode for local businesses.
• Chasing Yelp for ChatGPT, at least in the trades we track. It barely appeared.
• Screenshots of one answer. See step 10.
• "AI search volume" numbers. Nobody publishes what people ask ChatGPT, so treat any such figure as an estimate.
• Stuffing pages with "ChatGPT" keywords. ChatGPT reads for answers, not phrases.

Where we are starting from

Here's the honest bit. Before working through these steps on our own accounts, ChatGPT's weekly rates were flat:

The electrician's question set grew from 10 to 16 in late September, so his weeks before and after that aren't strictly comparable.

We're now applying these steps to the demo business and will report what moved, with the same questions measured before and after. You'll be able to watch it happen in the live demo and read every answer. For the earlier, shorter version of this playbook, see how to rank in ChatGPT. For the other AI systems, see how to rank in AI Overviews, AI Overview tracking for near me searches, Perplexity vs ChatGPT, ChatGPT citations and 5 AI systems, 35 questions.

How to track your own

You can work through all ten steps by hand: ask your buying questions in ChatGPT logged out, a few times a week, note who gets named and which sites are cited, and keep going for a month.

Or let LLM Visibility Lens do it. Every morning it asks your questions in ChatGPT, Gemini, Google AI Mode and AI Overviews, plus Perplexity on the Pro and Agency plans, and keeps every answer and source. It shows each system's named rate with its range, and each month it turns the data into an action plan built around these steps. The plans start free, and agencies can send reports under their own name. See the agency plan.

There's also a plain comparison with Semrush's AI visibility tool, the changelog, the FAQ page and the security page. If you'd like me to look at which of these steps would help your business most, get in touch.

Questions people ask

Harjeet Singh Bedi Founder, LLM Visibility Lens · SEO consultant, Bedi SEO

Harjeet has worked in SEO since 2010, first in agencies and now running Bedi SEO for local businesses, online stores and agencies. He built LLM Visibility Lens after clients kept asking whether they showed up in ChatGPT and other AI tools and nobody could give them an honest, measured answer. Every figure in this article comes from answers his tool collected.

Connect with Harjeet on LinkedIn