The Prompt Test: How to Check Your Presence in ChatGPT & Perplexity Yourself
Increasingly, customers are asking ChatGPT, Perplexity, or Gemini before they make a purchase. The obvious question for every shop operator is: Does the AI even know me – and what is it saying about me? Many test this with a quick question in their own chat window. Unfortunately, the result is often misleading. In this post, we'll show you why and provide you with four prompts to reliably check your AI presence yourself.
Why quick self-tests often yield false results
We repeatedly encounter four errors in reasoning in practice:
| Common Error | What happens | Better way |
|---|---|---|
| Testing in your own logged-in account | The AI's history and memory already know your brand – the response is personalized | Test in a private browser window, logged out, or with a new account |
| Only asking for your own brand name | Your customers often don't know your name yet – they ask about products and problems | Ask category and problem-based questions without mentioning the brand |
| A single attempt | AI responses fluctuate; a hit or miss can be a coincidence | Ask each prompt three to five times on different days |
| Evaluating the response without sources | The AI might answer from old training data instead of reading your current website | Pay attention to cited sources: Is your website linked? |
The four prompts for your self-test
Replace the placeholders in brackets with your own details. Tools with live web search, such as Perplexity or ChatGPT with search enabled, are the most revealing.
1. Accessibility: Can the AI read your website?
Please summarize the homepage of [your-domain.de]. Which products are offered and roughly how much do they cost? Cite the source.
How to interpret the result: If the AI answers evasively, with outdated information, or without your domain as a source, it might not be able to access your site. Common causes include a firewall rejecting AI agents or content that is only loaded via JavaScript. In the AI Trust Audit, the "AI Access: WAF & Firewall" check verifies this.
2. Trust: What does the AI know about your reputation?
I am considering ordering from [your-domain.de]. How reliable is this shop? Are there reviews, seals of approval, or experiences from other customers?
How to interpret the result: If the AI mentions no reviews despite you having many, they are often not machine-readable – for example, because they are only displayed via a widget. Structured review data and clear information regarding the imprint, shipping, and returns help most here.
3. Visibility: Are you recommended without being named?
I am looking for [product category, e.g., high-quality porcelain tableware] and would like to order online in [Country]. Which shops can you recommend and why?
How to interpret the result: This is the most important test because this is how your future customers ask. If only competitors appear, it's worth looking at their content: Do they answer specific questions? Do they have meaningful category texts and FAQs? This is exactly what the Content category evaluates in the audit.
4. Product Understanding: Does the AI understand your offers?
Please compare [specific product] from [your-domain.de] with similar offers. State price, availability, delivery time, and key features.
How to interpret the result: If price or availability are missing or incorrect, it usually means structured product data (Schema.org Product and Offer) or unique product identifiers like the GTIN are missing. For shopping agents that will compare and order autonomously in the future, this data is crucial.
What the self-test can achieve – and what it cannot
The four prompts give you an honest feel for how AI models perceive your shop today. However, they only show the symptom, not the cause. Whether a missing recommendation is due to the firewall, unreadable reviews, thin content, or missing product data cannot be determined in the chat window.
This is where the AI Trust Audit comes in: It checks your website technically and contextually in the areas of Crawling, Legal, Content, Reputation, Technology, and Structure, showing you the three levers that will have the greatest impact on your visibility in AI models.
Our tip for the process
- Run the four prompts in a private window and note the answers.
- Simultaneously start a free AI Trust Audit for your domain.
- Compare the two: The recommendations for action in the audit usually explain exactly the gaps you noticed in the self-test.
- Repeat the test a few weeks after the first improvements – this way you can see what has changed.
Check now: Start your free AI Trust Audit and find out what AI models can know about your shop.