How to Check If AI Recommends Your Brand (5-Minute Self-Audit)

The fastest way to know whether AI is working for or against your brand is to ask it the same questions your customers ask. Five questions, two models, twenty minutes — and you will have a clear picture of where you stand today.

Why Does the Audit Use Five Question Types?

Each question type maps to a distinct stage of the buyer funnel, which is why you need all five rather than a single search. A buyer at awareness stage asks "what is open-ear headphone technology?" A buyer at decision stage asks "should I buy [Brand X] or [Brand Y]?" AI models answer these differently, and your brand can be strong in one stage and completely absent in another.

The five types are:

1. Brand awareness — "What is [your brand]?" or "Tell me about [your brand]." This reveals whether the model has any coherent representation of your brand at all.
2. Category recommendation — "What are the best [product category] available right now?" This is the highest-value query: it shows whether you appear when someone has money to spend but no brand preference yet.
3. Scenario recommendation — "I need [specific use case], what should I look at?" This tests whether AI connects your product to the problems it actually solves.
4. Comparison — "How does [your brand] compare to [main competitor]?" This surfaces both whether you appear and whether the comparison is accurate.
5. Purchase decision — "Is [your brand] worth buying?" or "What do real users say about [your brand]?" This is what late-stage buyers ask before they hand over a credit card.

Running only the category query misses the scenario and comparison stages where many purchases are actually decided.

Which Models Should You Test?

Test in at least two models — ChatGPT (GPT-4o) and Gemini Advanced — as a minimum. Add Perplexity AI and Claude if you have time. Models are trained on different data mixes and retrieve information differently, so your brand can be well-represented in one and absent in another.

Start a fresh conversation for each question. Never ask all five in a single chat thread — earlier answers prime later ones, which inflates perceived visibility. If you sell in more than one language, run the full five-question sequence in each language you sell in. AI models trained on different language corpora can hold fundamentally different impressions of the same brand.

How Do You Score Each Answer?

Score every AI response on four dimensions:

• Present? — Yes or No. Did your brand appear at all?
• Position? — If yes, what rank? 1 is named first, 5 means buried late in a list.
• Accurate? — List any factual errors: wrong price, wrong category, wrong origin, discontinued products described as current.
• Who's talking? — When the model cites sources (Perplexity and browsing-mode ChatGPT always do; others sometimes do), list every domain. This is the most actionable column in the whole audit.

The domain list tells you which third-party sites currently shape what AI says about your category. If TechRadar appears three times and your own site appears zero times, you know where to focus next.

What Do the Three Visibility Buckets Mean?

Your results will fall into one of three buckets, each requiring a different response.

Invisible means your brand did not appear across most or all of the five question types. This is the most common state for brands that have strong Google rankings but have not yet invested in off-site authority building. The fix is building presence on the source types that AI actually cites — professional review media, vertical blogs, Reddit, and YouTube.

Mentioned-but-wrong means you appear but the information is incorrect or outdated. This is often more damaging than being invisible, because AI answers are delivered with confidence. A buyer who reads that your product has a feature it no longer has, or costs half what it actually costs, converts poorly and churns fast. Priority here is fixing your own site's structured data and fact-sheet consistency across every third-party domain that AI cites.

Recommended-but-behind means you appear in category queries but rank third or fourth behind competitors. This is a narrative and authority problem. The model has enough signal to mention you, but not enough to lead with you. The fix is systematically building more citation-worthy content and securing additional placements on high-authority domains.

What Does a Real Audit Look Like?

In June 2026 an open-ear headphone brand ran this audit across ChatGPT, Gemini, and Claude. Results: ChatGPT did not mention the brand unprompted in either category or scenario queries. Gemini mentioned it once in a comparison query but got the price wrong by 40%. Claude mentioned it in a scenario query and was factually accurate.

When sources were logged across the Perplexity and browsing-mode ChatGPT responses for the full category, five citations appeared: TechRadar twice, YouTube twice, Trustpilot once. The brand's own website appeared zero times across all five citation slots. The brand ranked in the top three on Google for its primary category keyword.

This is the defining pattern for brands making the Google-to-AI transition: high Google ranking built on technical SEO and backlink acquisition, but the third-party content ecosystem that AI actually reads has not been cultivated. Google ranks pages; AI cites sources that explain, review, and compare.

What Are the Honest Limits of This Audit?

Five questions is a sample, not a census. AI models are probabilistic — run the same prompt twice and you may get different brand rankings. The audit tells you your approximate visibility today; it does not guarantee the same result every time.

The audit also cannot tell you about purchase attribution. Adobe's Q1 2026 data showed AI-referred traffic converting 42% better than average web traffic, with revenue per visit 37% higher. That means AI visibility has real commercial consequences, but this audit measures top-of-funnel presence, not downstream conversion. Use it to diagnose a gap and set a baseline, then re-run monthly to track whether your off-site authority building is working.

One more limit: the audit covers English by default. If your brand sells in German, Japanese, Spanish, or any other language, run separate sessions. AI visibility in one language does not predict visibility in another.