How to get your business recommended by ChatGPT

There is no way to rank in ChatGPT in the sense of buying or fixing a position. What a business can do is make itself eligible, findable and easy to cite. When ChatGPT searches the web to answer a question, it sends searches to search providers and builds its answer from the pages that come back. The steps below follow from that, and each one is tied to what OpenAI has published or to named research. OpenAI's own wording is that placement is not guaranteed.

How ChatGPT finds businesses

OpenAI's help page on ChatGPT search describes the process:

• ChatGPT may search the web automatically when a question would benefit from current information.
• It typically rewrites the question into one or more targeted searches and sends them to search providers. It may send follow-up searches after reading the first results.
• For local questions it may use the user's approximate location, for example turning a "near me" request into a search for a named city.
• It ranks results using several factors intended to surface relevant, reliable information. OpenAI does not list those factors.
• Answers may include citations, and OpenAI warns that results and citations can be incomplete, outdated or incorrect.

Not every answer involves a search. The steps here concern the part a business can influence: what ChatGPT finds when it does search the web.

Step 1: let OpenAI's search crawler in

This is the one hard requirement. OpenAI's crawler documentation states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Its help page adds that the site's host or content delivery network must allow traffic from OpenAI's published search crawler addresses.

Check three things:

• The robots.txt file does not disallow OAI-SearchBot.
• The firewall, security plugin or content delivery network is not blocking it.
• If a change is made, allow about 24 hours. OpenAI says that is how long its systems take to pick up a robots.txt update.

OpenAI runs separate crawlers for separate jobs, and each can be set independently. GPTBot collects content that may be used to train models. A site can block GPTBot and still allow OAI-SearchBot.

Step 2: be findable for the specific search

ChatGPT does not send a whole conversation to a search provider. It sends targeted searches, sometimes with a city added. A page has to be returned for those searches before ChatGPT can use it.

This is an inference from OpenAI's description of how search works, not a statement OpenAI makes, but it leads to a practical rule. Each service and each location a business wants to be recommended for needs a page that is plainly about that service in that place. A business with one general services page gives a search engine nothing specific to return.

OpenAI's help page points to Microsoft's privacy policy for how search providers handle queries, which suggests Bing is among those providers. Checking that the site is indexed in Bing as well as Google is a sensible precaution.

Step 3: state the facts in plain text

An AI system can only repeat what it can read. On each key page, state in ordinary sentences:

• What the business does
• Where it operates and which areas it serves
• Who it is for
• What makes it different, in specific terms
• How to make contact

Keep these details the same everywhere the business appears online. Conflicting details in different places give a system no reliable answer to repeat.

Step 4: add a summary and questions and answers

Stryx's approach is to get core SEO right first, then add two things to each important page:

• A short summary at the top that answers the page's main question directly
• A set of real questions with clear answers

Both give a system a clear, accurate statement about the business to work from.

Step 5: give it something worth citing

The main academic study on this subject, GEO: Generative Engine Optimization, tested nine ways of changing page content and measured how much of the page then appeared in AI-written answers. According to the paper, adding quotations, statistics and cited sources worked best, and keyword stuffing did worse than making no change.

The study did not test ChatGPT. It used a simulated engine and Perplexity. The direction of the finding is still a reasonable working rule: specific, checkable substance was used more than repeated keywords. For a service business that means real detail about how the work is done, who does it and what a customer can expect.

Step 6: check what ChatGPT says

• Ask the questions a customer would ask, in a new chat, and note whether the business appears and what is said about it.
• Repeat the questions from time to time. OpenAI notes that the wording of a question, the user's location and saved memories all shape the results, so one test proves little.
• Watch website analytics for visits referred by ChatGPT, and count the enquiries that follow.
• If ChatGPT states something wrong about the business, look for the source it cites and correct the information there.

What will not help

• Paying for placement. OpenAI's help page does not describe any way to pay for a better position in search answers.
• An llms.txt file. The published evidence does not show an effect. See does llms.txt matter.
• Keyword stuffing. In the GEO study it reduced visibility.

Stryx offers this work as AI search optimisation, and an AI visibility audit records what AI systems currently say about a business. The crawler checks in Step 1 are part of the technical SEO audit checklist. To discuss a site, get in touch.