Skip to content
Chalk Labs / AI SEO
How-to guide

How to Get Your Business Recommended by ChatGPT

By Rahil Jain, founder of Chalk LabsUpdated How we research

The short answer

ChatGPT recommends a business when it runs a web search, finds your pages or listings in the results, and sees enough agreeing detail to name you. So let OAI-SearchBot (OpenAI's search crawler) in, publish plain fact pages, keep listings consistent, earn mentions on sites your buyers trust, and test with a fixed list of prompts. No one can promise a recommendation.

  • ChatGPT only searches the web for some prompts, so first learn which of your buyer questions trigger a search.
  • Allow OAI-SearchBot. Opted-out sites are not shown in ChatGPT search answers.
  • For facts about you, ChatGPT leans on your own site. For 'best' and comparison questions, other sites decide more.
  • Local picks are loose: repeat runs of the same prompt overlapped only 23% in one large study.
  • Test with ten fixed prompts, three runs each, and measure enquiries, not just visits.

How does ChatGPT decide which business to recommend?

ChatGPT recommends a business only after it decides to search the web, runs one or more searches, and reads what comes back. OpenAI's guide says the model can choose whether to search, based on the prompt 1. In its agentic mode, the model can also decide to keep searching after it reads the first results 1.

That choice happens less often than most owners assume. Ahrefs tracked 66.6 million free, logged-out ChatGPT responses (one model, August and September 2026). Only 14.4% contained fan-out queries, a close proxy for a web search 2. Those averaged about 1.6 fan-out queries, which are the extra searches an AI runs behind the scenes, and logged-in and paid users may differ 2. We found no published list of which prompts trigger a search, so test your own.

When ChatGPT does search, pages from ordinary search results are the ones it cites most. In an Ahrefs study of 1.4 million prompts, 88.46% of pages retrieved through general search were cited, against 1.93% of Reddit pages 3. The authors conclude that ranking in the search pool is necessary for a citation, and they advise titles and URLs that match ChatGPT's own sub-questions 3.

Do it step by step

0 of 8 done0%

  1. List ten questions a buyer would type: three about facts such as price, three 'best' or comparison questions, two local ones, and two about your brand name. Use the wording from real sales calls and emails. The free Query Fan-Out Explorer shows the extra searches an AI runs for each one.

    Tip: Keep the list fixed for three months so results stay comparable.

  2. Run each prompt three times in a logged-out ChatGPT window with search on. In the prompt sheet from the templates, log whether you were named, cited or absent, and which sites were cited instead. Those sites are your real competitors for each prompt.

    Tip: Screenshot the first run of every prompt. Answers change, and a screenshot is your only record.

  3. Open yourdomain.com/robots.txt and look for any rule that names OAI-SearchBot, or a blanket Disallow under User-agent: *. If needed, use the explicit-allow block from the templates. Then check your firewall or CDN, and wait about 24 hours 4.

    Tip: Done when the AI Crawler Checker shows OAI-SearchBot as allowed for your key pages.

  4. Start with the prompts where rivals were cited and you were absent. Use the fact-page outline from the templates: answer first, price range, who it suits, date checked. One question per page, in plain text a bot can read without scripts.

    Tip: Real numbers beat adjectives. 'From $40 per sitting, checked on 20 October 2026' is quotable.

  5. Compare your name, address, phone, hours and service list across your site, Google Business Profile, Bing Places, LinkedIn and the three directories your buyers use. Fix every mismatch on the same day. Yext links clear location data to being named, though not as proven cause 9.

    Tip: Keep one text file as the master copy and paste from it.

  6. Paste the LocalBusiness JSON-LD from the templates into your contact page, then replace every value. Keep only details that are visible on the same page, because Google's policy says markup must be a true representation of the page content 16.

    Tip: No source we opened shows ChatGPT reads this markup. Treat it as housekeeping, not a lever.

  7. Pick three sites your buyers already read: a trade directory, a local association, a review platform. Ask for a listing or an expert quote with accurate details. Do not buy links or reviews. See how to earn brand mentions with digital PR.

    Tip: Start with the sites that ChatGPT cited in your baseline runs.

  8. Run the same ten prompts, three runs each, and compare with the baseline sheet. Note what you changed and when. Keep the changes that moved a result, and drop the ones that did not.

    Tip: If nothing moved, check access again before you write more pages.

Your ticks are saved in this browser only.

Copy-ready templates

robots.txt: allow ChatGPT's search bot, optionally opt out of training

text
# Let ChatGPT search and user-initiated visits read the site.
# A bot with its own group ignores the User-agent: * group,
# so repeat your private-path rules here.
User-agent: OAI-SearchBot
User-agent: ChatGPT-User
Disallow: /admin/
Allow: /

# Optional: stop OpenAI using pages for model training.
# Delete these two lines if you are happy for training use.
User-agent: GPTBot
Disallow: /

# Everyone else
User-agent: *
Disallow: /admin/

Replace /admin/ with your own private paths. OpenAI documents GPTBot and OAI-SearchBot as separate settings, so the GPTBot block does not remove you from ChatGPT search . ChatGPT-User visits are user-initiated, so robots.txt may not apply to them.

Prompt test sheet (copy into a spreadsheet)

markdown
| # | Type | Prompt (edit the brackets) | Run 1 | Run 2 | Run 3 | Named? | Cited? | Sites cited instead | Action |
|---|------|----------------------------|-------|-------|-------|--------|--------|---------------------|--------|
| 1 | Fact | How much does [service] cost in [city]? | | | | | | | |
| 2 | Fact | What does [your business] do and who is it for? | | | | | | | |
| 3 | Fact | Does [your business] offer [service]? | | | | | | | |
| 4 | Best | Best [service] provider in [city] for [buyer type] | | | | | | | |
| 5 | Best | Best [product] for [use case] under [budget] | | | | | | | |
| 6 | Compare | [Your business] vs [rival]: which suits [use case]? | | | | | | | |
| 7 | Compare | [Rival] alternatives for [buyer type] | | | | | | | |
| 8 | Local | [Service] near me, open on [day] | | | | | | | |
| 9 | Local | Who is the most trusted [service] in [neighbourhood]? | | | | | | | |
| 10 | Trust | Is [your business] legit? What do reviews say? | | | | | | | |

Fill Run 1 to 3 with N (named), C (cited) or A (absent), on three different days. Log the date and whether search was on. Repeat after each change.

LocalBusiness JSON-LD for a contact page

json
{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "Example Bakery",
  "url": "https://www.example.com/",
  "telephone": "+1-555-0100",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "12 High Street",
    "addressLocality": "Springfield",
    "addressRegion": "IL",
    "postalCode": "62701",
    "addressCountry": "US"
  },
  "openingHoursSpecification": [
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": [
        "Monday",
        "Tuesday",
        "Wednesday",
        "Thursday",
        "Friday"
      ],
      "opens": "07:00",
      "closes": "17:00"
    }
  ],
  "areaServed": "Springfield, IL",
  "sameAs": [
    "https://www.linkedin.com/company/example-bakery",
    "https://www.facebook.com/examplebakery"
  ]
}

Every value here is a placeholder. Keep only facts that are visible on the same page . Yext associates schema markup with being named (2.1 times) but does not claim it causes selection . Wrap the block in a script tag of type application/ld+json.

Fact-page outline

markdown
# [Service] in [City]: price, process and who it suits

[Answer paragraph, 40 to 75 words: what it is, the typical price range, who it suits, how to book.]

## How much does [service] cost in [City]?
Typical range: $[low] to $[high]. Checked on [date].
What changes the price: [factor 1], [factor 2], [factor 3].

## Who is [service] for, and who should skip it?
[Two or three plain sentences. Name a case where you are not the right fit.]

## What happens, step by step?
1. [Step]
2. [Step]
3. [Step]

## Where do you work, and when are you open?
[Areas served. Address. Hours. Phone.]

## What do customers say?
[Two or three real quotes with first name, month and permission.]

## Questions people ask us
[Three questions with 30 to 60 word answers.]

Last checked on [date] by [name, role].

One page per buyer question. Plain text, no tabs or accordions, because hidden text may be skipped by AI systems.

What do you need, and how long does it take?

You need edit access to your website, a Google Business Profile if you serve a local area, and about an hour a week for testing. The access fix takes minutes. Writing fact pages and earning mentions takes weeks. These are our estimates, not measured figures.

We hold no measured time from first edit to first ChatGPT recommendation, so we will not invent one. Our published local-search case, Meru Trophy in Bengaluru, took 16 months to reach 27,300 Google clicks. That is Google, not ChatGPT, but it shows how slow organic visibility can be. Be wary of anyone who promises a date.

Can ChatGPT reach your site at all?

Check access first, because a blocked crawler cannot read any page you write. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as links 4. OpenAI recommends allowing OAI-SearchBot in robots.txt, the text file that tells bots which pages they may fetch 4.

Three OpenAI names matter. OAI-SearchBot finds pages for search. GPTBot collects content that may train models, and you can block it without leaving search 4. ChatGPT-User opens a page when a person asks, and OpenAI says robots.txt rules may not apply to those visits 4.

After you edit robots.txt, OpenAI says its systems can take about 24 hours to adjust 4. A firewall or CDN setting can still block the bot while robots.txt says yes. Run the free AI Crawler Checker and follow how to check if AI bots can crawl your site.

Which of your pages does ChatGPT cite, and when?

For facts about your own business, ChatGPT leans on your site. In an Ahrefs test of 20,000 US prompts, official sites received 63.4% of citations for fact questions such as pricing, features and company details. For comparison questions that fell to 46.4% 2.

So two kinds of page do different jobs. Fact pages answer 'what do you charge', 'where are you' and 'who is it for' in plain text. Comparison and 'best' questions are decided partly by other people's sites, which the next section covers.

Write each fact page around one question. Put the answer in the first two sentences, then add the price range, the date you checked it and who it suits. How to write answer-first introductions shows the pattern, and the free Answer Block Checker grades a draft.

Where citations come from, by type of buyer question
Buyer questionWhat the data showsWhat you publish
Fact: price, features, who it suitsOfficial sites got 63.4% of citations in 20,000 US prompts 2.One fact page per question, with a checked date.
Comparison: 'X vs Y', 'best X'Official sites got 46.4% of citations, so other sites decide more 2.An honest comparison page, plus presence on sites that review you.
Local: 'near me', 'in [city]'Across ChatGPT, AI Mode and AI Overviews combined, Google Business Profile was 28.5% of citations 8. BrightLocal says ChatGPT itself rarely uses it, so also fix Bing Places, directories and your own site.A complete profile, plus a location page per area.
Shopping: product searchProduct feeds in ChatGPT are open to approved partners only 11.Strong product pages until you are approved.

How much do mentions on other sites matter?

Mentions on other sites matter, but the evidence is correlation, not proof. Ahrefs studied 75,000 brands and found branded web mentions correlated with AI visibility at roughly 0.66 to 0.71 across platforms (about 0.66 for ChatGPT by our reading of its charts), against 0.352 for branded search volume 5. Domain Rating, an Ahrefs score for backlink strength, scored 0.266 5. The authors state that correlation is not causation 5.

Lists matter too. In an Ahrefs study of 750 top-of-funnel prompts for software, products and agencies, 'best X' blog lists were 43.8% of the source page types cited 6. The tempting move is to publish a 'best' list that ranks yourself first. In an Ahrefs test across ChatGPT, Gemini, Perplexity and Copilot, 82% of new mentions of a little-known conference came from answers that cited its own pages 7.

It is no safe trick, though. In 43% of answers citing those pages the AI named other conferences, and in one query ChatGPT never named it. That was one brand and one test 7. Sources also shift fast. In Ahrefs data, Reddit citations in US ChatGPT answers fell 88% and Wikipedia 80% from March to September 2026, while original sources gained 2. Aim to be a source, not to ride one platform.

What changes if you serve a local area?

Local prompts depend on place, and the results are loose. BrightLocal tested about 200,000 local prompts across ChatGPT, Google AI Mode and AI Overviews over 60 days. ChatGPT named 4.1 businesses per response on average, and repeat runs of the same prompt overlapped only 23% 8.

Distance and listings carry weight. Across all three AI engines, 52.8% to 59.3% of picks fell within 5 km of the search point and Google Business Profile made up 28.5% of citations. BrightLocal says ChatGPT uses Business Profile little, so do not rely on it for ChatGPT 8. It calls these associations, not proven causes 8.

Yext compared four AI engines on 491,972 shared local cases from February to August 2026. All four named the same top business in 4% of cases, and the highest-rated business ranked first only 40% of the time 9. Yext links clear location data to being named, but says one field alone does not cause selection 9.

Reviews still matter to the people who read the answer. BrightLocal's 2026 survey of 1,002 US adults found 45% use ChatGPT or other generative AI for local recommendations, up from 6% a year earlier, and 47% will not use a business with fewer than 20 reviews 10. For the full local playbook, read how to optimize a local business for AI search.

What did an AI assistant search for when we asked this question?

On 20 October 2026 a Claude agent with web search ran five searches for 'how do I get my business recommended by ChatGPT', covering the method, selection, local, mentions and crawler access.

That is one run on one engine, so read it as a sample, not a rule.

Which mistakes waste the most time?

Most wasted effort comes from four habits. The table pairs each with what the evidence says and a better move.

Common mistakes when chasing ChatGPT recommendations
What people doWhy it backfiresDo this instead
Block every AI bot after a news scareSites opted out of OAI-SearchBot are not shown in ChatGPT search answers 4.Block GPTBot only if you want to stop training use. Allow OAI-SearchBot.
Test one prompt once and decideRepeat runs overlapped only 23% for ChatGPT in local tests 8.Run each prompt three times and log all runs.
Publish a 'best' list that crowns yourselfIn an Ahrefs test across four AI engines, 43% of answers citing a conference's pages named other conferences 7.Disclose, include alternatives, show your criteria.
Copy last year's Reddit and Wikipedia playbookReddit and Wikipedia citations in US ChatGPT answers fell 88% and 80% in six months 2.Become a source with original facts.

How do you check that it worked?

Check with a fixed prompt panel, not one lucky search. Run each of your ten prompts three times, on different days, in a logged-out window with search on, and log whether you are named, cited or absent. Because repeat runs overlapped only 23% in local tests, a single appearance proves little 8.

Then look for visits and leads. Google Analytics 4 has a default channel called AI Assistant for visits from sources such as ChatGPT, Gemini, Copilot and Grok, and it excludes Google's AI Overviews and AI Mode 13. Expect small numbers: in Ahrefs' September 2026 benchmark, ChatGPT sent about 4.3 million, 0.34% of traffic, across 115,254 sites 14.

Add a 'how did you hear about us?' field to your enquiry form, as in how to set up self-reported attribution. For a fuller method, read how to test whether your AI SEO changes worked.

When is this not worth your time?

Skip ChatGPT work if your buyers do not use it. Menlo Ventures' July 2026 survey of 5,067 US adults found 26% used AI to research topics, and ChatGPT reached 60% of AI users 15. Ask ten recent customers whether they asked an AI assistant before they contacted you.

Pause if you cannot edit robots.txt or your CDN settings, or if your pages carry advice that needs expert review. Fix access first, then wording. For regulated fields, read AI SEO in regulated industries before you publish claims.

What an AI assistant searches when it answers this

We gave the question below to AI search tools and recorded the web searches they ran behind the scenes. This is called query fan-out. Each phrase is a page you could publish to be found and cited.

The prompt

“how do I get my business recommended by ChatGPT”

The searches it ran
  • how to get recommended by ChatGPT
  • how ChatGPT search chooses sources to cite
  • ChatGPT local business recommendations optimize
  • brand mentions ChatGPT study Reddit Wikipedia citations
  • OAI-SearchBot robots.txt allow ChatGPT crawler

Observed October 20, 2026 using Claude web search. Claude agents were told to run 6 to 10 searches, so that count is our setting, not Claude's own choice. Assistants differ and change often, so treat this as a sample, not a rule. See how we collect it.

Frequently asked questions

Can I pay to be recommended by ChatGPT?

OpenAI's crawler page lists OAI-AdsBot, which checks the safety of pages submitted as ads on ChatGPT 4. That is a separate paid route. This guide covers the unpaid route only, where you earn a place through access, clear pages and outside mentions.

How long until ChatGPT recommends my business?

No one can say. We hold no measured time to a first ChatGPT recommendation, and our published local-search case, Meru Trophy, took 16 months to reach its Google clicks (Google, not ChatGPT). Expect to test for several weeks with a fixed prompt panel and judge the trend, not one answer.

Does ChatGPT use Google or Bing results?

OpenAI's guide does not name its search provider 1. Ahrefs found only about 12% of links cited by ChatGPT, Gemini and Copilot sat in Google's top 10 for the same prompt, and ChatGPT's in-text citations were 8.0% 17. So ranking first on Google is neither required nor enough.

Should I add an llms.txt file for ChatGPT?

It is optional and unproven. SE Ranking found no correlation between having llms.txt and AI citation frequency across nearly 300,000 domains 12. Spend your first hour on crawler access and fact pages. Read what llms.txt is before you decide.

Do reviews affect whether ChatGPT names my business?

Reviews influence the people who read the answer, and Yext says ratings and review depth help rank a business once it is in the set 9. Yext does not claim reviews alone cause selection 9. Keep reviews recent, specific and answered, and never buy them.

About the author

Rahil Jain

Rahil Jain is an AI SEO (answer engine and generative engine optimisation) consultant and co-founder of Chalk Labs.

LinkedIn · Method and credentials

Sources

  1. Web search (tools guide) · OpenAIOpened October 20, 2026. Supports: The model can choose whether to search based on the prompt; agentic search can keep searching; responses include inline citations; the guide does not explain how sources are chosen or name a search provider.
  2. How ChatGPT Does Query Fan-Out (15 Million Queries Studied) · AhrefsOpened October 20, 2026. Supports: 14.4% of 66.6 million free, logged-out responses triggered a search, averaging about 1.6 fan-out queries; official sites got 63.4% of citations for fact questions and 46.4% for comparison questions in 20,000 US prompts; Reddit citations fell 88% and Wikipedia 80% from March to September 2026.
  3. Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts) · AhrefsOpened October 20, 2026. Supports: Citation rates by retrieval channel (search 88.46%, Reddit 1.93%); the authors' conclusion that ranking in the search pool is necessary; advice on titles and URLs matching fan-out sub-questions.
  4. Overview of OpenAI Crawlers · OpenAIOpened October 20, 2026. Supports: Roles of OAI-SearchBot, GPTBot, ChatGPT-User and OAI-AdsBot; opted-out sites are not shown in ChatGPT search answers; robots.txt may not apply to ChatGPT-User; about 24 hours for robots.txt changes.
  5. Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) · AhrefsOpened October 20, 2026. Supports: Correlations with ChatGPT brand mentions: branded web mentions 0.664, branded search volume 0.352, Domain Rating 0.266; the authors' warning that correlation is not causation.
  6. Do Self-Promotional 'Best' Lists Boost ChatGPT Visibility? Study of 26,283 Source URLs · AhrefsOpened October 20, 2026. Supports: In 750 top-of-funnel prompts, 'best X' blog lists were 43.8% of the source page types cited.
  7. Self-Promotional Content Works, Until It Backfires (AI SEO Experiment) · AhrefsOpened October 20, 2026. Supports: 34 self-promotional pages and 9,886 answers; 82% of new mentions of the little-known brand came from answers citing the author's pages; ChatGPT never named it; 43% of answers citing conference pages recommended a rival; the author's ethical guidance.
  8. What 200k local AI searches tell us about what businesses get recommended · BrightLocalOpened October 20, 2026. Supports: ChatGPT averaged 4.1 businesses per response; repeat-run overlap of 23%; 52.8% to 59.3% of picks within 5 km; Google Business Profile 28.5% of citations; business websites 42% of 1,967,762 citations; findings described as associations.
  9. Four AIs, four different answers · Yext ResearchOpened October 20, 2026. Supports: Across 491,972 shared local cases, four engines named the same top business in 4% of cases and the highest-rated business ranked first 40% of the time; clear location data and schema markup are associated with being named; the study does not claim one field causes selection.
  10. Local Consumer Review Survey 2026 · BrightLocalOpened October 20, 2026. Supports: In a survey of 1,002 US adults, 45% use ChatGPT or other generative AI for local recommendations, up from 6%; 47% will not use a business with fewer than 20 reviews.
  11. Get started with product feeds · OpenAIOpened October 20, 2026. Supports: Onboarding product feeds in ChatGPT is currently available to approved partners.
  12. LLMs.txt study · SE RankingOpened October 20, 2026. Supports: Article dated 15 October 2026: of nearly 300,000 domains, 10.13% had an llms.txt file, and the study found no correlation between the file and AI citations.
  13. Default channel group definitions · Google Analytics HelpOpened October 20, 2026. Supports: The AI Assistant default channel covers visits from sources such as ChatGPT, Gemini, Deepseek, Copilot and Grok, and excludes Google's AI Overviews and AI Mode.
  14. AI vs Search traffic benchmark · AhrefsOpened October 20, 2026. Supports: September 2026: ChatGPT sent 4.3 million visits, 0.34% of total traffic, across 115,254 sites tracked with Ahrefs Web Analytics.
  15. 2026: The State of Consumer AI · Menlo VenturesOpened October 20, 2026. Supports: Survey of 5,067 US adults in July 2026: 26% used AI to research topics; ChatGPT reached 60% of AI users.
  16. General structured data guidelines · Google Search CentralOpened October 20, 2026. Supports: Markup must be a true representation of the page content, and content not visible to readers must not be marked up.
  17. Only 12% of AI Cited URLs Rank in Google's Top 10 for the Original Prompt · AhrefsOpened October 20, 2026. Supports: Of 15,000 long-tail prompts, about 12% of links cited by ChatGPT, Gemini and Copilot appeared in Google's top 10; ChatGPT in-text citations 8.0%.
How this page was made. Built on Rahil Jain's AI SEO method and drafted with AI research assistance. Every statistic links to the page we read it on (opened October 20, 2026 or nearby). Published October 20, 2026; last updated October 20, 2026. Chalk Labs sells AI SEO services, so we have a commercial interest in this topic. Spotted an error? Tell us and we will fix it.