How to Manually Check Your Brand's Appearance in AI

A growing share of shoppers now ask ChatGPT, Gemini, Claude or Perplexity for recommendations before they ever open a search engine. That shift means a brand can rank perfectly on Google and still be invisible in the answer an AI assistant gives someone comparing options. The good news is that checking this yourself does not require expensive software, just a bit of patience and a consistent method.

Start With a Logged-Out Session

Before typing a single prompt, open an incognito or private browser window. Being logged into an account can let past conversations and personalisation quietly shape the answers you get back, which will skew your results. Testing from a clean, logged-out session gives you a much closer approximation of what a new customer would actually see.

Write Prompts the Way Customers Actually Ask

Resist the urge to simply ask an AI model what it knows about your company. Real customers rarely search that way. Instead, mix in indirect and category level questions, such as asking for the best options in your space, a comparison between you and a competitor, or a recommendation for a specific problem your product solves. Run the same set of prompts across ChatGPT, Gemini, Claude and Perplexity, since each model pulls from different sources and can produce noticeably different answers.

Check the Sources, Not Just the Answer

Tools like Perplexity and Gemini often show the pages they used to generate an answer. Whenever your brand appears, click through to see which article, review or listing is being cited, and note whether the information is accurate and current. Whenever a competitor appears instead of you, look at what made their page a better fit and treat that as a gap in your own content.

Log Everything in a Simple Spreadsheet

Before reaching for a paid monitoring platform, a plain spreadsheet is usually the better starting point. For every prompt, record which model you used, whether your brand was mentioned, where it appeared in the answer, the overall tone, and any factual errors you spotted. Repeating this on a weekly or monthly basis matters more than doing it once, because AI answers shift as models are updated and as new content gets indexed.

Fix What the Models Get Wrong

If you spot outdated pricing, an old product line-up or a factual mistake about your brand, the fix is not to argue with the chatbot, it is to update the pages, listings and structured data those models are drawing from. Keeping core facts about your business consistent across your own site, review platforms and third party listings is still the most reliable way to influence how AI describes you, and a manual audit like this is the simplest way to know where you currently stand.