How to get your business recommended by ChatGPT: a step-by-step plan
You can't apply or pay to be recommended by ChatGPT. It names businesses from what its model already knows and from pages it finds when it searches the web. So the work runs in this order:
- Record what ChatGPT says about you today.
- Make sure OpenAI's search crawler, OAI-SearchBot, can reach your site.
- Make your name, category, location and prices match everywhere.
- Publish pages that answer buyer questions plainly.
- Get mentioned on the sites ChatGPT cites for your category.
- Re-check the same prompts every month.
Nobody outside OpenAI knows exactly how it picks businesses, and nobody can promise you a mention. What you control is whether ChatGPT finds accurate, specific facts about you when it looks. Each step depends on the one before it.
What to have in place before you start
- Edit access to your website, including robots.txt. On Shopify that means the robots.txt.liquid theme template. On WordPress it's usually your SEO plugin or the file on your server. On Umbraco it's typically a file or a package setting.
- Access to your CDN, host or firewall settings. Crawler blocks can sit here as well as in robots.txt.
- Logins for your Google Business Profile and directory listings (local businesses) or your review platform profiles (B2B companies).
- Google Analytics 4 on the site, or Shopify's analytics if you run a store.
- A ChatGPT account and a spreadsheet.
- Time: an afternoon in week one for the baseline and technical checks, then a regular weekly block for pages and outreach.
How ChatGPT picks businesses to recommend
People ask how to rank in ChatGPT, but there is no fixed list of ten results to climb. Each answer is written fresh. It comes from one of two places, and the difference decides how fast your work can show up.
Answers from the model's own knowledge
When ChatGPT answers without searching, it draws on what it learned from text it was trained on, up to a cutoff date. If your business was written about often and consistently before that cutoff, the model may know you. If you opened recently or are rarely mentioned, it probably doesn't. You can't edit this directly. It changes when OpenAI trains and releases a new model.
Answers built from a live web search
ChatGPT can also search the web and write its answer from the pages it finds. If the answer lists sources or links, it searched. These answers can change much faster, because they depend on what's on the web now and whether OpenAI's crawler can reach it. OpenAI's help article for Enterprise and Edu workspaces mentions Bing as a search provider. OpenAI doesn't publish a full list of providers for every plan.
What OpenAI doesn't say
OpenAI doesn't publish ranking factors. It doesn't say which local listing sources it uses (Google Business Profile, Bing Places, Yelp or others). Its crawler documentation doesn't say whether OAI-SearchBot runs JavaScript, and the OpenAI pages we checked don't say it uses llms.txt or schema markup. Any advice that says ChatGPT "relies on" one particular source is a guess. This plan covers the likely sources instead of betting on one.
Step 1: Record what ChatGPT says about you today
Do this first. Without a baseline you can't tell later whether anything changed, and you'll miss the wrong facts you need to fix in Step 3.
Write 10 to 20 buyer prompts
Write them the way a customer would type them, not as keywords. Pull phrasing from sales calls, support emails and the queries report in Google Search Console. Cover each type:
- Category: "Who's the best emergency dentist in Sacramento?" or "Best bookkeeping firms for construction companies"
- Problem: "I cracked a tooth on a Sunday, where can I go in Sacramento?"
- Price: "How much does a dental crown cost in Sacramento?"
- Comparison: "[Your business] vs [a competitor]" and "Alternatives to [a competitor]"
- Brand: "Is [your business] any good?" and "What does [your business] charge?"
- Near me: "Dentist open Saturdays near me" (note where you ran it from)
Run each prompt three times and save the answers
The same prompt gives different answers from one run to the next, so a single run tells you little. Use a temporary chat, which doesn't use your memory, or turn memory off. Run a few prompts logged out as well. For local prompts, put the city in the prompt and write down where you were when you ran it. Note which model you used.
Here's a sample row to show the format. The businesses and details in it are made up.
| Prompt | Run | Date | Searched? (sources shown) | Businesses named, in order | Named you? | What it said about you | Cited URLs | Errors to fix |
|---|---|---|---|---|---|---|---|---|
| Best emergency dentist in Sacramento (example) | 1 of 3 | Nov 3 | Yes | Business A, Business B, you | Yes, 3rd | "Open Saturdays" (example) | A local news roundup, a directory listing, your site | Old address |
Fifteen prompts at three runs each gives you 45 answers. Your baseline score is simple: named in X of 45. Save the full text of each answer too. The wording ChatGPT uses about you matters as much as whether you appear.
Step 2: Make sure ChatGPT can reach your site
If OpenAI's search crawler can't fetch your pages, nothing you publish in Step 4 can show up in answers built from your own site.
Allow OAI-SearchBot in robots.txt
OpenAI runs separate crawlers for separate jobs, listed on its crawler documentation page. OAI-SearchBot is used for ChatGPT search, and OpenAI recommends allowing it if you want to appear in search answers. ChatGPT-User fetches a page when a user asks ChatGPT to visit it, and OpenAI says robots.txt rules may not apply to it. GPTBot is listed separately and tied to training OpenAI's models. Read its current description on that page before you decide whether to block it.
A robots.txt that allows search and blocks GPTBot looks like this:
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /Leave out the GPTBot lines if you don't want to block it. Then open yourdomain.com/robots.txt and read the live file. Look for a blanket Disallow: / under User-agent: *, or a rule against OAI-SearchBot added by a plugin, theme or a previous developer. OpenAI's crawler page also says how long robots.txt changes take to reach search. Check the current figure there rather than a number quoted elsewhere.
Blocking a page in robots.txt doesn't remove it from ChatGPT completely. OpenAI's publishers and developers FAQ says a blocked page may still appear as a link and title, and that a noindex tag is how you keep it out.
Check every layer that can block crawlers
robots.txt is a request. A firewall rule is a locked door, and it applies no matter what robots.txt says. Check each layer between a crawler and your site that you have:
- Bot settings at your CDN
- Security plugins, for example on WordPress
- Your host's security or firewall settings
- Rate limits
Look for rules that mention AI crawlers, OpenAI or unknown bots. Defaults differ by provider, so read your own settings instead of assuming.
OpenAI publishes the IP ranges OAI-SearchBot uses at openai.com/searchbot.json and recommends allowing them. If you can read your server logs, search for OAI-SearchBot requests and check the status codes. 200 means pages are served. 403, 429 or 503 means something is turning the crawler away. Faking the user agent with a command-line tool doesn't test IP-based rules, so logs are the better check.
Make sure key content is in the HTML
OpenAI doesn't document whether OAI-SearchBot runs JavaScript. As a precaution, make sure your key facts are in the raw HTML. Open a page, view the source (Ctrl+U on Windows, Cmd+Option+U on a Mac) and search for your phone number, address, prices, service names and review count. If something is visible on the page but missing from the source, JavaScript added it. Common culprits are review widgets, pricing tables loaded by apps, booking widgets, tabs filled in by script and sites built as single-page apps. Put the text that matters into the page itself.
If you run WordPress or Umbraco, FoundWell's free plugins run an AEO audit on every page, add schema automatically and generate an llms.txt file. OpenAI doesn't document using schema or llms.txt, so treat them as cheap extras. Crawler access and content in the HTML come first.
Step 3: Make your facts consistent everywhere
Start with the errors from your Step 1 sheet. If ChatGPT gave an old address or a service you dropped, find where that fact still lives. The cited URLs in your sheet usually show you.
Then make these match across your site and every listing you control:
- Business name: one version. Not "Smith Dental" in one place and "Smith Family Dentistry LLC" in another.
- Category: the same words for what you are, such as "family dentist" or "bookkeeping firm for contractors".
- Address, service area, phone number and hours.
- Prices or price ranges, on a pricing page that matches what listings say.
- Your About page: year founded, who runs it, who you serve and what you don't do.
- Google Business Profile, Bing Places, Apple Business Connect, Yelp, Facebook and your LinkedIn company page.
- Industry directories and review platforms in your field.
Because OpenAI doesn't say which listings it uses, update all the major ones rather than one. For listings you don't control, contact the site owner with the correct details.
If ChatGPT gets something wrong without listing sources, that error comes from the model's built-in knowledge and you can't edit it directly. Publishing the correct facts widely helps answers that use web search now and gives future models better material.
Step 4: Publish pages that answer buyer questions directly
When ChatGPT searches, it can only use what your pages actually say. Vague marketing copy gives it nothing to quote. Compare these two openings:
- "We offer comprehensive dental solutions for the whole family."
- "Smith Dental is a family dentist in East Sacramento. We see patients from age 2, accept most PPO plans and open at 7 a.m. on Saturdays."
The second can be lifted into an answer. The first can't. Build these pages, guided by the gaps your prompts revealed:
- A pricing page with real numbers or ranges, what's included and what costs extra.
- One page per service, and one per location if you genuinely serve several areas. Copies that only swap the city name don't count.
- Comparison pages against categories of alternatives: "Dental implants vs bridges" or "Outsourced bookkeeping vs hiring in-house".
- A "who we're for and who we're not for" section. It helps ChatGPT match you to the right questions.
- Answers to the questions you hear on sales calls, one question per section.
Put the answer in the first sentence of each section. Use numbers, names and places. Add a date to anything that changes, like prices or opening hours.
Step 5: Get mentioned on the pages ChatGPT cites
Collect and sort the cited URLs
List every URL from the "Cited URLs" column across all your runs. Count how often each domain appears. A domain that shows up across several prompts is a priority, because it keeps feeding answers in your category. Then sort each one by type.
One action for each type of source
| Source type | Examples | What to do |
|---|---|---|
| Directory or listing | Yelp, BBB, industry directories | Claim the profile, complete every field, match the facts from Step 3. |
| Best-of list or roundup | "Best dentists in Sacramento" on a local news site or blog | Email the author with specific, checkable facts and ask to be considered at the next update. If you're offered a paid spot, ask whether it will be labeled as sponsored. |
| Review platform | Google reviews, G2, Capterra, Clutch | Ask recent customers for reviews and reply to them. Never buy reviews or only ask happy customers. |
| Press or trade publication | Local paper, trade journal, industry newsletter | Pitch something they can use: your own data, a useful expert comment, a real story. |
| Forum or community | Reddit, industry forums | Answer questions as yourself and say who you are. Don't post from fake accounts. |
| A competitor's site | Their pricing or comparison page | You can't get onto it. Note what they publish that you don't, and cover it in Step 4. |
| Your own site | Any of your pages | This is working. Keep that page accurate and up to date. |
Step 6: Re-check every month and track visits from ChatGPT
Once a month, run the same prompts with the same settings, three times each. Record how many times you were named out of the total, which new sources appeared and whether the errors from Step 1 are gone. Compare against your baseline, not against last week's mood.
For traffic, OpenAI's publishers and developers FAQ says ChatGPT adds utm_source=chatgpt.com to the links it sends, which you can track in Google Analytics for sites that allow OAI-SearchBot. In GA4, open the Traffic acquisition report, set the dimension to session source and look for chatgpt.com. In Shopify, look at sessions by UTM source or referrer. Then check which landing pages those visits hit. That tells you which pages ChatGPT links to.
Know what this can't measure:
- A mention that doesn't lead to a click isn't counted anywhere.
- Someone who reads your name in ChatGPT and later searches Google for you won't be credited to ChatGPT. The visit is recorded under whatever channel they used next.
- A redirect that drops query strings strips the parameter, and the visit lands under another source.
Add "ChatGPT or another AI assistant" as an option on your "How did you hear about us?" question to catch some of what analytics misses.
The local business track vs the B2B track
The six steps are the same for everyone. What you test, where you get mentioned and which pages you build differ a lot between a dentist and a software or services company.
| Local business | B2B company | |
|---|---|---|
| Typical prompts | "Best plumber in Columbus", "open now near me", "how much does X cost in [city]" | "Best [category] for [industry]", "alternatives to [competitor]", "[product] pricing" |
| Listings to get right first | Google Business Profile, Bing Places, Apple Business Connect, Yelp | LinkedIn company page, review platform profiles, industry directories |
| Pages to build | Service and area pages, pricing, hours, insurance or payment options | Pricing, comparisons against categories of alternatives, use cases by industry, integrations, who it's for |
| Review sites | Google reviews, Yelp, sites specific to your trade | G2, Capterra, Clutch, Trustpilot, depending on your field |
| Press and mentions | Local news, neighborhood blogs, city best-of lists | Trade publications, industry newsletters, podcasts, comparison articles |
| Key detail | Location: the city has to be in your listings, pages and test prompts | Buyer fit: industry, company size and use case have to be stated plainly |
| What progress looks like | Named more often in "best in [city]" prompts, calls and bookings from chatgpt.com visits | Named in category and alternatives prompts, demo requests from chatgpt.com visits |
What changes in weeks and what takes months
There's no reliable timeline for being recommended, and we won't invent a success rate. What we can tell you is what each kind of change depends on.
| Change | What it depends on | What to expect |
|---|---|---|
| Unblocking OAI-SearchBot in robots.txt | OpenAI picking up the new rules | OpenAI's crawler page states how long this takes. Check the current figure there. |
| Fixing a firewall or CDN rule | Your change, then the next crawl | The block lifts as soon as you change it. Recrawl timing isn't published. |
| New or updated pages | Being crawled, then being picked for an answer | Can appear in answers that use web search once crawled. No fixed time. |
| Reviews and third-party mentions | Other people publishing, then those pages being found | Weeks to months, mostly outside your control |
| The model's built-in knowledge | OpenAI training a new model | Only changes with new model releases. OpenAI doesn't say what goes into each one. |
A three-month work schedule
This is when you do the work. It says nothing about when results arrive.
| Weeks | Work | Must be done first |
|---|---|---|
| 1 | Prerequisites, Step 1 baseline, Step 2 access checks | Site, CDN and analytics access |
| 2 | Step 3: fix errors from the baseline, then align listings | Baseline sheet |
| 3 to 6 | Step 4: pricing, service and comparison pages | Facts settled in Step 3 |
| 3 to 12 | Step 5: outreach to cited sources, ongoing review requests | Sorted list of cited URLs |
| 5, 9, 13 | Step 6: monthly re-check and traffic review | Baseline sheet and analytics |
Can you pay ChatGPT to recommend you?
No. You can buy an ad, but not a recommendation. OpenAI's announcement, Testing ads in ChatGPT, describes a test in the US that shows ads to Free and Go users, with no ads on Plus, Pro, Business, Enterprise and Edu plans. Ads are labeled and appear below the answer. OpenAI's ads help page says: "Ads do not influence ChatGPT's answers," and "advertisers have no ability to shape, rank, or alter ChatGPT's responses." On May 5, 2026, OpenAI opened a self-serve Ads Manager beta for US advertisers.
| What you might try to buy | Available? |
|---|---|
| A labeled ad below a ChatGPT answer | Yes, for US advertisers, as part of OpenAI's ads program |
| A place inside the answer itself | No, according to OpenAI |
| A paid entry on a third-party list that ChatGPT cites | Depends on the list. OpenAI doesn't say how it treats paid entries on other sites. If you're offered one, ask for a sponsored label. |
If someone sells you guaranteed placement inside ChatGPT's answers, they're selling something OpenAI says doesn't exist.
Doing it yourself, with a freelancer or with a service
None of these steps needs special software. They need steady hours every week for months. The baseline is the easy part. The monthly re-checks and outreach are what slip when other work gets busy.
| Option | What it costs you | What to check |
|---|---|---|
| Doing it yourself | Your time: 45 or more answers to run and log each month, plus pages and outreach | Whether you can keep the monthly re-checks going when work gets busy. |
| A freelancer | Varies by person and scope | Whether monthly re-checks and site fixes are written into the agreement, and who covers when they're unavailable. |
| An SEO agency | Usually a monthly retainer. Terms vary. | Whether tracking AI answers is included and whether changes are made in your CMS. See how FoundWell compares to an SEO agency. |
| FoundWell | $1,000 per month per website, month to month, cancel any time | A fixed scope: up to 12 articles and 5 outreach pitches a month. It doesn't promise rankings or AI mentions. |
FoundWell does this work every day inside your WordPress, Umbraco or Shopify site: a daily site audit with fixes, indexing checks, articles, and weekly tracking of what ChatGPT, Perplexity, Claude and Google's AI answers say about you. Outreach pitches are drafted for you, never sent in your name. In approval mode nothing goes live until you approve it. If you're weighing it against writing content in ChatGPT yourself, see FoundWell compared with doing SEO in ChatGPT.
The guarantee covers Google clicks, not AI mentions: 25% more clicks from Google in 90 days, or FoundWell keeps working for free until you get them. It compares days 63 to 90 with the 28 days before you connect, in Google Search Console, and needs at least 90 days of Search Console data at connection, 90 days connected, and autopilot or approvals answered within a week. It's a poor fit for a one-off fix or a brand-new site with no Search Console history.
Want the six steps done for you every month, inside your own CMS? Sign up for FoundWell and start in approval mode.
Frequently asked questions
Can I pay ChatGPT to recommend my business?
No. OpenAI is testing labeled ads in ChatGPT in the US, shown below the answer to Free and Go users, and opened a self-serve Ads Manager beta for US advertisers on May 5, 2026. But OpenAI's ads help page says "Ads do not influence ChatGPT's answers" and that advertisers "have no ability to shape, rank, or alter ChatGPT's responses." You can buy an ad next to an answer, not a place in it.
How do I rank in ChatGPT?
ChatGPT has no fixed ranking. It writes each answer from its built-in knowledge or from pages it finds in a web search. To appear more often, record your baseline across 10 to 20 buyer prompts, allow OAI-SearchBot in robots.txt and your firewall, make your business facts consistent, publish pages that answer buyer questions plainly, get mentioned on the sites ChatGPT cites, and re-check monthly.
Should I block GPTBot?
It's a separate decision from search. OpenAI's crawler documentation lists GPTBot separately from OAI-SearchBot, the crawler used for ChatGPT search, and ties GPTBot to training its models. Blocking GPTBot doesn't block OAI-SearchBot. Read GPTBot's current description on OpenAI's page before you decide, and keep in mind that blocking it may mean future models learn less about you from your own site.
Do llms.txt and schema markup help with ChatGPT?
OpenAI doesn't document using either one on the pages we checked (November 2026). They're cheap to add and may help other systems, but crawler access, key content in the raw HTML and clear, factual pages matter more and should come first.
ChatGPT says something wrong about my business. What should I fix?
Check whether the answer lists sources. If it does, open the cited pages, find the wrong fact and fix it on your site or ask the site owner to update it. If no sources are shown, the error comes from the model's built-in knowledge, which you can't edit. Publish the correct facts consistently on your site and listings so searched answers improve now and future models learn the right version.
Does the same plan work for Perplexity?
Mostly, yes. None of the six steps is specific to ChatGPT: a baseline of buyer prompts, consistent facts, clear pages and mentions on the sites the answers cite all carry over. The crawler check is the part to redo. Find the crawler name in each assistant's own documentation and make sure your robots.txt and firewall allow it.
About the author
Peter Sinke, Founder
Peter Sinke builds websites at Datapad in the Netherlands and made FoundWell from the daily AI SEO routine he ran for Datapad's own site.
$1,000 per month per website. Month to month, cancel any time.