Auditing what AI says about your brand category means systematically querying ChatGPT, Perplexity, Google AI Overviews, and Gemini with the prompts your buyers actually use, then recording which brands get named, how they are described, and where yours ranks. This gives you a real picture of your AI share of voice and tells you exactly where your brand is missing from the conversations that increasingly replace Google searches.

Key Takeaways

  • AI answer engines pull from review sites, editorial roundups, Reddit threads, and structured product data. Your visibility there determines whether you appear in answers.
  • A useful audit needs 30-50 buyer-intent prompts tested across at least three AI platforms, logged in a spreadsheet with exact responses.
  • The output is not a vanity score. It is a gap list that tells you which content to create, which third-party sites to get listed on, and which claims to correct.

Why This Matters Now

When someone asks ChatGPT "best moisturiser for dry skin" or Perplexity "sustainable running shoes under $150," the answer names 3-5 brands. If yours is not one of them, you are invisible in a channel that is growing fast and has no ad units you can buy.

This is different from SEO. Google shows ten blue links and you can rank for keywords. AI answer engines synthesise a single response. You are either in it or you are not. And unlike paid search, you cannot bid your way in.

Step 1: Build Your Prompt List (30-50 Queries)

Start with the questions a real buyer would type into an AI assistant. These fall into four buckets:

Discovery prompts

"Best [category] for [use case]", e.g. "best protein powder for women over 40"

Comparison prompts

"[Brand A] vs [Brand B]" and "alternatives to [competitor]"

Problem prompts

"How do I fix [problem your product solves]", e.g. "how to stop acne from hormonal changes"

Purchase prompts

"Is [your brand] worth it" and "[your brand] reviews"

Pull these from your search console queries, customer support logs, and Amazon search terms. Add the exact phrasing your customers use, not marketing language. Aim for 30 minimum, 50 if your category has many sub-segments.

Step 2: Query Four Platforms and Log Responses

Test each prompt on these four surfaces:

  1. ChatGPT (GPT-4o, free tier, no plugins), the largest user base
  2. Perplexity, cites sources, so you can see where it pulls from
  3. Google AI Overviews, visible to anyone searching on Google
  4. Gemini (gemini.google.com), Google's standalone AI, draws from different data than AI Overviews

For each query, log in a spreadsheet:

Prompt Platform Brands mentioned (in order) Your brand present? How described Sources cited

Use a fresh session or incognito window for every query. AI models personalise responses based on conversation history, so a dirty session gives you skewed results.

This takes 3-4 hours for 40 prompts across four platforms. Do it manually the first time. Automated tools exist (Profound, Otterly, Peec AI) but they cost $100-300/month and you need to understand what you are measuring before you automate it.

Step 3: Score Your Visibility

Count across all queries:

  • Mention rate: what percentage of relevant prompts include your brand
  • Position: when mentioned, are you first, second, or last in the list
  • Sentiment: are descriptions accurate and positive, or outdated and wrong
  • Source trail (Perplexity only): which URLs feed the answers

Compare your mention rate against your top 3 competitors. If competitor A appears in 70% of category prompts and you appear in 15%, that gap is your priority.

Step 4: Identify Why You Are Missing

AI models build brand associations from specific source types. When you are absent from answers, it is almost always one of these:

  • No presence on roundup sites. AI heavily weights editorial lists ("10 best X of 2026") from sites like Wirecutter, Byrdie, RunRepeat, or category-specific review blogs. If you are not on these lists, you will not be in the answer.
  • Thin Reddit footprint. ChatGPT and Perplexity pull heavily from Reddit. Search Reddit for your category terms and see if your brand appears in threads.
  • No structured product data. Google AI Overviews and Gemini favour brands with clean schema markup, Google Merchant feeds, and consistent product specs across retailers.
  • Outdated or negative information. If the top search results about your brand reference old formulations, past controversies, or discontinued products, that is what the AI repeats.

Step 5: Build a Fix List

Turn your audit into a prioritised action list:

  1. Pitch roundup editors. Identify the 5-10 roundup articles that Perplexity cites for your top category prompts. Contact those editors about inclusion. This is old-school digital PR, and it now has a direct AI visibility payoff.
  2. Seed Reddit and community discussions. Genuine participation in subreddits where your buyers hang out. Not astroturfing. Answer questions, share honest experience with your product category.
  3. Fix your product data. Ensure Google Merchant Center feed is complete with GTINs, accurate descriptions, and current pricing. Add FAQ schema to your product pages addressing the exact prompts from your audit.
  4. Publish comparison content on your own site. "[Your brand] vs [Competitor]" pages that are factual and genuinely useful. AI models pick these up when they have real substance.
  5. Correct wrong information. If AI is repeating something inaccurate, trace it to the source URL (Perplexity makes this easy) and get it corrected at the source.

Step 6: Re-Audit Monthly

AI model outputs change as training data updates and retrieval indices refresh. Run the same prompt list once a month. Track your mention rate over time to see if your efforts are working.

Keep the same prompts for consistency, but add 5-10 new ones each month based on new search trends or product launches.

Common Mistakes

  • Testing with a logged-in, personalised session. ChatGPT adjusts responses based on your history. If you have been talking about your own brand, it will mention you more. Always use incognito or a fresh account.
  • Only checking ChatGPT. Each platform has different training data, retrieval methods, and biases. A brand that dominates in Perplexity answers may be absent from Gemini.
  • Treating this as a one-time project. AI outputs shift frequently. A monthly cadence is the minimum to catch regressions and measure progress.
  • Focusing on branded queries only. Your brand name queries matter, but the real value is in unbranded category queries where buyers have not decided yet.
  • Trying to manipulate AI directly. There is no meta tag or hack that forces inclusion. The path runs through the sources AI models read: editorial sites, forums, structured data, and your own well-built content.

Want this strategy executed for your brand? Adsome builds and runs AI ad campaigns for DTC brands.