How to Check If ChatGPT Recommends Your Business

1. Why this matters: Your buyers now ask AI, not Google

ChatGPT has hundreds of millions of weekly active users. A massive chunk of your buyers are no longer opening traditional search engines to research software, services, or products. Instead, your buyers ask AI assistants for recommendations.

Think about the profound shift in the buyer journey. When someone searches Google for "best accounting software," they expect to do the work of clicking through five different blogs to synthesize an answer. When they ask ChatGPT, they expect the AI to just give them the final verdict. If ChatGPT doesn't bring up your website in that verdict, you're essentially invisible to a highly motivated audience that is ready to buy.

2. The manual way to check (Free, 15 minutes)

You don't need expensive software to get a pulse check on your AI visibility. You can do a manual baseline test right now using the major AI platforms (ChatGPT, Perplexity, and Google Gemini).

When you run these prompts, you are looking for a specific type of output: an explicit citation. Here is an example of what a successful Answer Engine citation looks like:

3. Why the manual way is unreliable

While running those five prompts is a great starting point, manual testing has massive blind spots. Here is why you cannot rely on manual ChatGPT queries to build your marketing strategy:

• Personalization Bias: ChatGPT tailors its responses based on your past chat history, your location, and your custom instructions. If you frequently visit your own website, the AI might suggest it to you, but hide it from a prospect halfway across the country.
• The Probabilistic Nature of LLMs: Large Language Models are essentially advanced autocomplete. They roll the dice on every word. If you ask the exact same question three times in three different browser windows, you will likely get three completely different lists of recommendations.
• Lack of Historical Tracking: You cannot track progress manually. Knowing you were cited today doesn't tell you if your visibility dropped by 40% next week following an OpenAI model update.

4. The signals that decide whether you get cited

If you run the manual tests and find that you are completely invisible, it usually means you are failing on one of four critical technical signals. Generative engines do not recommend sites randomly; they extract from sites that are structurally optimized for them.

• Crawler Access

  AI cannot recommend what it cannot read. If your robots.txt file blocks GPTBot, ClaudeBot, or PerplexityBot, you are willingly erasing your brand from the AI's training and retrieval dataset.

• Brand Entity Presence

  Does the AI trust you? Generative engines rely heavily on the Knowledge Graph. You must establish your brand through strict Organization JSON-LD schema, Wikipedia mentions, and high-authority backlinks.

• Content Extractability

  LLMs do not render heavy Javascript or complex CSS grid layouts well. They want clean, semantic HTML. You must provide 40-to-60 word definition blocks directly following your H2 headings.

• The llms.txt File

  The new gold standard for AI visibility. Providing a markdown-formatted llms.txt file in your root directory gives AI agents a clean, noise-free summary of your pricing and features.

5. How to check all of it automatically

If you want to accurately measure your AI visibility, you need a clean-room approach. That means querying the AI models directly via their APIs, completely stripping away any user-level personalization bias, and doing it at scale.

You need to track whether AI engines mention your brand programmatically.

With Igris Radar, the platform sends these queries live to the engines (ChatGPT, Claude, Perplexity), reads the generated responses, and programmatically searches for your brand name. It also runs sentiment analysis to ensure the AI isn't actively telling users to avoid your software.

6. What to do if you're not being recommended

If your tracking reveals an abysmal visibility score, do not panic. It simply means you have not optimized for the new answer engines yet. Here is your prioritized fix list:

1. Fix your robots.txt immediately: This is a 5-minute fix. Ensure GPTBot and ClaudeBot are allowed.
2. Run an AEO Audit: Scan your high-value landing pages to see if your content is actually extractable by a machine.
3. Restructure your comparison pages: AI engines love tables. If you have a "Vs Competitor" page, make sure the actual feature comparison is in a clean HTML <table>, not a complex CSS grid.
4. Build your llms.txt file: Create a markdown summary of your product and host it at the root of your domain.