You have no idea what AI is saying about your brand right now.
Maybe ChatGPT recommends you for the right reasons. Maybe it recommends your competitor instead. Maybe it describes you incorrectly. Or maybe you simply don't exist in the conversation at all.
The only way to know is to run an audit. Here's exactly how to do it.
Step 1: Define your audit scope
Before you start typing prompts into ChatGPT, you need a plan. Random testing gives you random insights.
Pick your platforms. The five that matter most right now: ChatGPT, Perplexity, Claude, Gemini, and Grok. Each has a different user base, different data sources, and different recommendation patterns. Testing only one gives you a partial picture.
Build your prompt list. Create 15-20 prompts that reflect what your potential customers actually ask. These fall into three categories:
- Discovery prompts. "What are the best [your category] tools?" or "Recommend a [your service] for [use case]."
- Comparison prompts. "What's the difference between [you] and [competitor]?" or "Which [category] tool is best for [specific need]?"
- Brand-specific prompts. "What is [your brand]?" or "Tell me about [your brand]."
The discovery prompts are the most valuable. They reveal whether AI thinks of you when someone is looking for a solution you provide.
Step 2: Run the audit
Open each platform. Enter each prompt. Record the results in a spreadsheet.
For every prompt and platform combination, capture four things:
- Were you mentioned? Yes or no.
- Position. Were you the first recommendation, third, or buried at the end?
- Description accuracy. Did the AI describe you correctly? Did it get your category, pricing, or key features right?
- Competitor presence. Who else appeared in the response?
This gives you a matrix. 20 prompts across 5 platforms means 100 data points. That sounds like a lot of manual work, and it is. But the first audit is worth doing by hand because you learn things a dashboard can't show you, like how the AI frames its reasoning, what language it uses, and where it hedges.
Step 3: Score your results
Now turn those data points into something actionable.
Mention rate. What percentage of prompts included your brand? Across all platforms? Per platform? A 40% mention rate on Perplexity but 5% on Claude tells you something specific.
Accuracy rate. Of the times you were mentioned, how often was the description correct? Being recommended as "a project management tool" when you're actually an invoicing platform is worse than not being mentioned at all.
Competitive share. For each prompt, count how many competitors appeared alongside you. If three competitors show up consistently and you don't, those are the battles you're losing.
Step 4: Identify the three types of gaps
Your audit will reveal problems that fall into three categories.
Complete invisibility. AI doesn't mention you at all for relevant prompts. This usually means weak online authority, inconsistent brand information across the web, or insufficient content on your site that addresses those queries.
Incorrect descriptions. AI mentions you but gets the details wrong. This happens when your messaging is inconsistent across platforms, when outdated information dominates your web presence, or when AI confuses you with a similarly named brand.
Competitor dominance. AI consistently recommends competitors over you. This means they have stronger authority signals: more third-party mentions, better reviews, more comprehensive content, or clearer positioning.
Each gap type requires a different fix.
Step 5: Create your action plan
For invisibility gaps, the fix is building presence. Create content that directly answers the prompts where you're missing. Get mentioned on credible third-party sites. Make sure your brand information is consistent and clear across directories, review sites, and social profiles.
For accuracy gaps, clean up your web presence. Update your website copy so AI has correct information to pull from. Fix outdated listings. Make your positioning crystal clear so there's no room for confusion.
For competitor dominance gaps, study what they're doing differently. Often they have more comprehensive content, stronger review profiles, or better third-party coverage. Your AI search optimization strategy should address these specific weaknesses.
Prioritize by business impact. A gap on a high-intent prompt like "best [category] for [your ideal customer]" matters more than a gap on a generic informational query.
Step 6: Automate ongoing monitoring
Here's the uncomfortable truth about LLM audits: a single snapshot is misleading.
AI responses change. Models get updated. New training data gets incorporated. Your competitors are working on their visibility too. The audit you ran today could be outdated in two weeks.
Running a manual audit every month is possible but painful. 100 data points, every month, recorded in a spreadsheet. Most teams do it once, learn a lot, then never do it again because it takes half a day.
Mentionable automates this. You set up your prompts (it even suggests relevant ones based on your URL), choose your platforms, and get continuous monitoring. When your mention rate drops on a specific platform or a competitor starts appearing where you used to, you know immediately instead of finding out three months later.
The value isn't just in the data. It's in the trend. Knowing your ChatGPT mention rate went from 35% to 50% over two months proves your content strategy is working. Seeing it drop from 50% to 30% after a model update tells you something changed and you need to investigate.
What a good audit reveals
A thorough LLM audit typically uncovers surprises. Brands that rank well on Google often assume they're visible in AI too. They're frequently wrong.
Common findings:
- Strong Google presence doesn't guarantee AI visibility. Different signals, different outcomes.
- Visibility varies wildly across platforms. You might be well-represented on Perplexity (which searches live) but invisible on Claude (which relies more on training data).
- Competitor rankings in AI don't match competitor rankings in Google. The playing field is different.
- Small, focused brands sometimes outperform larger competitors in AI because they have clearer positioning and more specific content.
The audit gives you facts instead of assumptions. From there, you can build a strategy based on what's actually happening rather than what you think is happening.
For a deeper look at tracking specific platforms, see our guides on tracking AI mentions and the best AI visibility tools.
