Skip to content

How to check if ChatGPT mentions your brand (AI visibility)

Published: Last updated:

AI visibility (also called LLM visibility) describes how often AI assistants such as ChatGPT and Claude mention, cite and recommend your brand when people ask them a question. This guide shows how to check it by hand, what to record, which sources matter and where manual checks reach their limits.

Why AI answers matter for your brand

More and more people put questions to an AI assistant that they used to type into a search engine. Ask for the best tool, agency or product in a category and you get a short written answer with a handful of names instead of a page of links.

ChatGPT and Claude can both search the web to answer. ChatGPT shows the sources it used for answers with web search, and Anthropic states that every response from Claude's web search includes citations. That gives AI visibility two parts: whether the assistant names your brand, and whether your own website appears among the sources.

An AI answer has no fixed position like number 3 on a results page. What you can measure is how often your brand shows up across many answers. The steps below show how.

Step 1: Write the questions your customers ask

Start with five to ten questions that a potential customer might type into an assistant. Use their words rather than your marketing language, and write them in the language of your customers and for their country. Mix different kinds of questions:

  • Category questions: What is the best software for scheduling jobs for tradespeople?
  • Comparison questions: Which is better for a small trades business, tool A or tool B?
  • Problem questions: How do I keep track of quotes across several building sites?
  • Local questions: Which field service apps are popular in Austria?
  • Brand questions: What is Kantwerk and who is it for?

The examples use Kantwerk, a fictional brand. Write every question down word for word, so you can ask it again unchanged.

Step 2: Track your brand in ChatGPT and Claude

Ask each question in ChatGPT and in Claude with web search switched on, because that is when the assistants cite sources. Claude searches the web when it helps, and you can also ask for it directly, for example with “Search the web” in your prompt. Three things distort the result if you ignore them:

  • Answers vary. OpenAI states that model outputs may differ from request to request. Ask every question several times, for example five times, and count how often your brand appears instead of trusting a single answer.
  • Your own account knows you. ChatGPT can reference saved memories and past chats, and Claude can use memory and search past chats. That way an answer can look different for you than for other people. Use a temporary chat without personalization in ChatGPT and an incognito chat in Claude, or switch memory off in the settings.
  • Language and country matter. Ask in the language your customers use. Answers can differ by language and country, so note in which language and from which country you test.

What to record: mentions, recommendations, citations and competitors

For every answer, note the following points:

  • Mention: does the answer name your brand at all?
  • Recommendation: is your brand recommended, and at which position in a list?
  • Citation: does your own domain or a specific page of yours appear among the sources?
  • Competitors: which other brands are named, and how often?
  • Share: the answers that name you divided by all answers, per question and overall. If 3 of 9 answers name your brand, your share is one third.
  • Description: does the assistant describe your brand the way you would?

A simple spreadsheet is enough: one row per answer with the date, question, assistant, run number and one column for each point. Repeat the check on a fixed schedule, for example every month. A single snapshot says little, the trend is what counts.

Which sources shape your AI search visibility

When an assistant searches the web, it lists the pages it used. Collect these sources for every answer and look at which websites come up again and again, whether they are yours or not. Those pages shape the answers in your category.

For each recurring source, ask how your brand could get onto it: a correct entry in a directory, a mention in an article, a review, a contribution to a specialist publication. A source that is cited for many of your important questions, and does not mention you, is your next PR or content task. If your own pages are cited, note which ones and keep them accurate and up to date.

Let AI crawlers read your website

An assistant can only cite what it can read. The providers publish crawlers with different purposes, and you control them through your robots.txt:

  • OAI-SearchBot (OpenAI) surfaces websites in ChatGPT search. OpenAI states that sites that are opted out of it will not be shown in ChatGPT search answers.
  • GPTBot (OpenAI) crawls content that may be used to train OpenAI's models. Disallowing it signals that your content should not be used for training.
  • ChatGPT-User (OpenAI) visits a page when a user asks something in ChatGPT. Because these actions are initiated by a user, OpenAI says robots.txt rules may not apply.
  • Claude-SearchBot (Anthropic) analyzes content to improve search result quality, and Claude-User retrieves content when a user asks Claude a question. Anthropic notes that blocking either may reduce your visibility. ClaudeBot collects content that may be used for training.
  • PerplexityBot (Perplexity) surfaces and links websites in Perplexity search results and is not used to train foundation models.

Blocking training bots and allowing search bots are therefore separate decisions. Allowing the search crawlers is a prerequisite for being cited, not a guarantee. OpenAI and Perplexity both say that robots.txt changes can take up to about 24 hours to take effect.

The free robots.txt checker for AI crawlers shows which of these bots your robots.txt allows, whether you have an llms.txt and a sitemap, and drafts an llms.txt for you.

From manual checks to ChatGPT brand monitoring

A manual check is a good start and costs nothing. It does not scale: ten questions with five runs each in two assistants make 100 answers per round, for one market and one point in time. Comparing months, following competitors and keeping every question identical takes more time than most teams have.

Listentic automates exactly this. Every week it asks ChatGPT and Claude, both with web search, the questions your customers ask, with several runs per question (3 by default, up to 10), and shows rates such as “3 of 9 answers” instead of a single result. It reports whether you are mentioned, recommended and cited, how often you are named compared with up to 5 tracked competitors, and which sources the assistants rely on. Questions run per market, so one project can ask in German for Austria and in English for the US.

Listentic asks without chat history, memories or personal settings. It uses the official APIs of OpenAI and Anthropic with web search. That comes close to what people see in the apps but is not identical, because the apps sometimes use other models and settings.

The AI visibility tool page describes all features. To compare providers first, see the overview of AI visibility tools or the comparisons with Peec AI and Otterly.AI. Plans are listed in the pricing section, and the Starter plan comes with a free 14-day trial. A payment method is required at sign-up, and nothing is charged until the trial ends.

Frequently asked questions

How often should I check my AI visibility?

One check tells you little, because answers change. Repeat the same questions on a fixed schedule, for example monthly by hand or weekly with a tool, and compare the share of answers that name your brand.

Why does ChatGPT give a different answer each time?

Model outputs can differ from request to request, as OpenAI states. That is why you should look at how often your brand appears across several runs, not at a single answer.

Which AI assistants should I check?

Start with the assistants your customers use. ChatGPT and Claude both search the web and show their sources, which makes their answers easy to check. Listentic covers both.

Does blocking GPTBot remove my site from ChatGPT answers?

OpenAI documents its bots separately: GPTBot relates to training, while OAI-SearchBot decides whether a site is shown in ChatGPT search answers. Check both in your robots.txt.

Is there a free way to check whether AI crawlers can read my site?

Yes. The robots.txt checker for AI crawlers is free and shows which AI crawlers your robots.txt allows.