How to get your brand to appear in ChatGPT

For your brand to appear in ChatGPT, the assistant must be able to read your site when it searches the web and must find pages that answer the user's question. That means allowing OAI-SearchBot in robots.txt, publishing pages that answer specific questions and keeping the company's name, address and services the same everywhere they appear.

No one controls an AI assistant's answer. You can control what it can read about you and the number of sources that confirm that information. This guide explains each of those pieces and how to check the result.

How do AI assistants choose sources?

An assistant such as ChatGPT, Claude or Perplexity answers from two sources of information:

• Training data. Text collected from the web before the model was trained. What is there stays fixed until the next model. If your site changed after that collection, the model does not know.
• Web search at the time of the question. The assistant runs one or more searches, reads some pages and cites them in the answer. This is where a page published today can appear tomorrow.

Each company uses different robots for each of these functions. At OpenAI, GPTBot collects content that can be used to train models and OAI-SearchBot is used to show sites in ChatGPT search results. OpenAI states that sites that block OAI-SearchBot do not appear in ChatGPT search answers. Anthropic separates ClaudeBot (training), Claude-SearchBot (indexing for search) and Claude-User (requests made by users). Perplexity uses PerplexityBot to show and link sites in results and states that this robot does not collect content to train models.

For a small business, web search is the part you can work on. Training data changes slowly and does not depend on you.

What can you control on your site?

Allow search robots in robots.txt

The robots.txt file sits at the root of the domain and tells each robot what it can read. Many sites block AI robots without knowing it, through rules inherited from a theme, a security plugin or a hosting setting. Open yoursite.com/robots.txt and look for these names:

• OAI-SearchBot and GPTBot (OpenAI)
• Claude-SearchBot, Claude-User and ClaudeBot (Anthropic)
• PerplexityBot (Perplexity)

If you want to appear in answers, the search robots (OAI-SearchBot, Claude-SearchBot, PerplexityBot) must not have Disallow: /. Training robots (GPTBot, ClaudeBot) are a separate decision: blocking them says you do not want your content in training data, and according to OpenAI this does not remove the site from ChatGPT search. OpenAI and Perplexity state that a change to robots.txt can take about 24 hours to take effect.

Pages that answer specific questions

An assistant cites the page that answers the question asked. If someone asks "how much does an accountant for freelancers in Braga cost", the page that states the price, the city and the type of client is more likely to be cited than a home page with a generic sentence.

• One page per service, with what you do, for whom, where and from what price.
• Answers to the questions customers ask you by email or phone, written on the site.
• The answer in the first paragraph and the details after it.
• Headings that repeat the question, so the reader and the assistant know what the section answers.
• Text in HTML. Information that exists only inside images or PDFs is harder to read.

Structured data

Structured data describes the page in a standardised format. Google recommends the JSON-LD format and Schema.org types. For a company, the Organization type (or LocalBusiness, if you have a customer-facing address) lets you state the name, address, logo and, in the sameAs field, the profiles on other sites. Google says this markup helps it tell your company apart from others with similar names. Test the markup in Google's Rich Results Test.

Google also says there is no special markup to appear in AI Overviews or AI Mode. Structured data describes the company; it does not guarantee citations.

Same name, address and services in every directory

When the assistant searches your brand, it finds your site, Google Business Profile, sector directories, social networks and review platforms. If the name, address, phone number or list of services changes from one place to another, the information contradicts itself. Make a list of the profiles you have, correct the outdated details and always use the same spelling of the company name.

Mentions on third-party sites

When the question is "who do you recommend for X", the assistant looks for pages that compare or list companies: local press articles, association lists, sector directories, interviews, podcasts with a transcript. A brand that appears only on its own site has fewer sources confirming what it does. Some ways to earn mentions:

• Apply for membership in associations and directories of your sector.
• Reply to requests from journalists in your field.
• Publish customer cases with permission and with the customer's name.
• Ask for reviews on platforms that appear in searches in your sector.

The llms.txt file

llms.txt is a Markdown file, placed at the root of the site, with the site name, a summary and a list of links to the main pages. It was proposed by Jeremy Howard, of Answer.AI, in September 2024. llmstxt.org describes it as a proposal open to contributions, not as a formal standard.

None of the sources consulted for this guide confirms that OpenAI, Anthropic or Perplexity use llms.txt to choose which sites to cite. Google says you do not need to create text files for AI to appear in AI Overviews or AI Mode. Creating the file takes little time and does no harm, but it does not replace the pages and robots.txt.

How do you measure whether your brand appears?

Ask the assistants

Write 5 to 10 questions that a potential customer would ask before hiring you, without including your brand name. For example: "which wedding photographer do you recommend in Porto" or "what is the best invoicing software for freelancers". Ask the same questions in ChatGPT, Claude and Perplexity, with web search on, and record:

• whether your brand is mentioned;
• whether your site appears in the cited sources;
• which competitors and which pages appear in your place.

Answers vary from one conversation to another. Repeat the questions once a month and compare the lists, instead of drawing conclusions from a single answer.

Read the server logs

Your hosting's access logs (in cPanel, under "Raw Access" or "Visitors") show the user agent of each request. Look for OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User and PerplexityBot. If they never appear, check robots.txt and the hosting firewall rules. If they appear, see which pages they visit: those are the pages being read to answer questions.

Use Search Console for Google

Google states that traffic from AI Overviews and AI Mode goes into the Search Console Performance report, under the "Web" search type. It does not appear separately, but it shows whether the pages continue to be displayed.

What mistakes are common?

• Blocking all AI robots at once. A rule meant to prevent training also ends up blocking the search robots.
• Firewall or CDN blocking robots. robots.txt allows access, but the hosting returns a 403 error or a verification challenge.
• Site built only in JavaScript. If the text only appears after the JavaScript runs, some robots read an empty page.
• Service pages without price, area or audience. The assistant has nothing to cite when the question includes those details.
• Company name written in different ways. Acronyms, old versions and commercial names different from the site split the mentions.
• Expecting results from llms.txt. It is a proposal with no confirmed effect on citations.
• Measuring with a single question. One isolated answer does not show the pattern.

Where to start?

1. Open your robots.txt and confirm that OAI-SearchBot, Claude-SearchBot and PerplexityBot are not blocked.
2. Write the 5 questions your customers ask before buying and confirm that each one has an answer on a page of the site.
3. Add Organization or LocalBusiness structured data to the home page, with the sameAs field filled in.
4. Correct the name, address and services in the directories where the company appears.
5. Ask the questions to the three assistants and save the answers to compare the following month.

If you want to know how SEO relates to this, read the guide AEO vs SEO. If the site needs to be rebuilt, see the website with an AI assistant service.