Short answer
To get recommended by ChatGPT, make sure OAI-SearchBot can crawl your site, publish pages that answer your buyers' exact questions in the first paragraph with checkable facts, and get named accurately on the third-party lists, reviews and directories ChatGPT already cites for your category. Keep your business facts consistent everywhere. Then check the same blind questions many times with search on, and only trust changes that beat the noise. There is no way to buy an organic recommendation.
On this page
- How ChatGPT builds a recommendation
- Step 1: let the right crawler in
- Step 2: find the questions that matter
- Step 3: see what ChatGPT says today
- Step 4: publish pages worth citing
- Step 5: get on the pages ChatGPT reads
- Step 6: keep your facts consistent
- Step 7: measure, then repeat
- How we help
- Frequently asked questions
- Sources
How ChatGPT builds a recommendation
When ChatGPT searches the web, it reads a handful of pages and writes an answer from them. OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which opt out of OAI-SearchBot will not be shown in ChatGPT search answers. When it does not search, it answers from training data.
So there are two paths to a recommendation: be in the pages it reads when it searches, and be well described on the public web so future models learn accurate facts about you. The first path is the one you can move this quarter.
Step 1: let the right crawler in
Open yourdomain.com/robots.txt and look for rules that name OpenAI's agents. OpenAI lists three you should know:
| Agent | What OpenAI says it does | Block it? |
|---|---|---|
| OAI-SearchBot | Surfaces websites in ChatGPT search results | Only if you want to be left out of ChatGPT search answers |
| GPTBot | Collects content to improve OpenAI's foundation models | Your choice. Blocking signals your content should not be used for training |
| ChatGPT-User | Visits pages when a user asks ChatGPT to | robots.txt may not apply, as these are user actions |
Step 2: find the questions that matter
List 20 to 50 questions a real buyer might ask ChatGPT before choosing a business like yours. Do not include your brand name. Cover these types:
- Discovery: "best billing software for a small pharmacy"
- Comparison: "Tool A vs Tool B for GST invoicing"
- Alternatives: "cheaper alternative to Tool A"
- Pricing: "how much does billing software cost per year in India"
- Problem: "how do I stop stock mismatch in my shop"
- Local: "billing software support in Indore"
Step 3: see what ChatGPT says today
- Ask each question with a note to search the web first. Do not mention your brand.
- Repeat each question several times, ideally in fresh chats, because answers vary.
- For each run, write down the businesses named, their order, and the sources linked.
- Count how often you appear. Note any wrong facts about you.
- Group the sources: your site, competitor sites, list articles, review sites, directories, forums, news.
Using a clean setup matters. A logged-in chat with memory may be shaped by your own past conversations, and a buyer will not see that.
Step 4: publish pages worth citing
For each high-value question where you are missing and no page of yours answers it, write one. Use the citation-first format:
- A 50 to 100 word direct answer under the title.
- The reader's problem first, your company later.
- Specific facts: prices with dates, named options, steps, tables.
- Links to official or primary sources.
- A worked example.
- A "Last verified" date.
- An honest limits note.
- Four to six real FAQs.
Step 5: get on the pages ChatGPT reads
Look at the source list from step 3. For each third-party page that appears often:
- If it is a directory or review site, claim and complete your profile, and ask real customers for honest reviews.
- If it is a list article, send the author a short, factual note with evidence. Never pay to be listed without saying so.
- If it is a forum thread, answer helpfully only if you have something real to add, and say who you are.
- If it is a news article, consider whether you have genuine news or data worth sharing with that publication.
Step 6: keep your facts consistent
ChatGPT pieces together a picture from many pages. If your site says you serve "small businesses", a directory says "enterprises", and an old article says you only do payroll, the answer may be vague or wrong. Write one entity sentence and one fact sheet, and update your About page, directory profiles and social bios to match.
Step 7: measure, then repeat
Re-run the same questions on a schedule. Report a rate with a range, not a single yes or no. Keep some questions you did not work on as a control group. If both groups rise together, something outside your work, such as a model update, probably caused it.
How we help
We run steps 3 and 7 for you on ChatGPT and other AI apps, with blind questions and forced search, and draft the pages and outreach notes for steps 4 and 5 for you to review.
Frequently asked questions
Can I pay OpenAI to be recommended?
Organic answers are not sold. OpenAI documents a separate crawler for checking ads on ChatGPT, and ads are a different product from organic recommendations.
Does blocking GPTBot hurt my ChatGPT visibility?
OpenAI treats GPTBot as the training crawler and OAI-SearchBot as the search crawler. Blocking GPTBot does not, by OpenAI's description, remove you from search answers. Blocking OAI-SearchBot does.
Why does ChatGPT describe my business wrongly?
It is repeating what it read or learned. Find the pages with the wrong facts, fix the ones you control, and ask owners of the others to update them.
Should I ask ChatGPT about my own brand to check?
That tells you what it says about you when asked, which is useful for accuracy. It does not tell you whether buyers hear your name. For that, use blind questions.
How many times should I run each question?
Enough to see a pattern. A handful of runs gives a rough view. Dozens of runs across weeks give a rate you can compare with a confidence range.
Does ChatGPT use my llms.txt file?
OpenAI's crawler documentation does not say it does. An llms.txt file is a low-cost extra, not a replacement for clear, crawlable pages.
Sources
Last verified: 17 September 2026. Facts on this page come from the public documents below. Brands such as Tallybook are fictional and used only as examples.