Key takeaways
- Most brands have no idea what AI search engines say about them — or whether they're mentioned at all. Running a manual audit takes under 30 minutes and reveals your true AI visibility baseline.
- The audit has three layers: direct brand queries, category queries (where buyers actually start), and competitor comparison queries. You need all three.
- Sentiment and accuracy matter as much as presence. Being mentioned with wrong information is sometimes worse than not being mentioned.
- Manual audits give you a snapshot; tracking tools give you the trend. Both matter.
- Fixing gaps requires content changes, not just technical tweaks. AI models cite sources — you need to be a source worth citing.
Here's something most marketing teams haven't done yet: open ChatGPT, type "what's the best [your category] tool for [your use case]," and read the response carefully.
Go ahead. I'll wait.
If your brand didn't show up, you're not alone. In a Reddit thread where one marketer audited 20 brands across ChatGPT, Gemini, Claude, and Perplexity, the majority either weren't mentioned or were described inaccurately. Some had outdated pricing. A few had their core product described wrong. One brand that had been acquired was still described as independent.
This is the new brand audit. Not just "what does Google say about us" but "what do AI models say about us when a buyer asks a real question." The answers are often surprising, sometimes alarming, and almost always actionable.
Why this matters more than you might think
AI search is no longer a future trend. ChatGPT has over 400 million weekly active users. Perplexity is handling hundreds of millions of queries per month. Google's AI Overviews appear on a significant portion of searches. When someone asks "what CRM should I use for a small sales team" or "best project management software for agencies," they're increasingly getting a direct answer from an AI model — not a list of ten blue links they have to sift through.
The difference from traditional SEO is significant. In Google search, you could rank on page one and still get clicks even if you weren't the top result. In AI search, the model synthesizes an answer and typically recommends two or three options. If you're not in that short list, you don't exist for that buyer.
What makes this harder is that AI models don't tell you why they chose the sources they did. They don't show you a ranking. You have to go find out.
The three-layer brand audit
A proper AI brand audit isn't just typing your company name into ChatGPT. That's the least useful query you can run. Buyers don't search for your brand name unless they already know you — the real risk is the queries where they're discovering options for the first time.
Layer 1: Direct brand queries
Start here to understand what AI models actually know about you. These queries establish your baseline.
Run each of these across ChatGPT, Claude, Perplexity, and Gemini:
- "What is [Brand Name]?"
- "What does [Brand Name] do?"
- "Tell me about [Brand Name] — what are their main features and pricing?"
- "Is [Brand Name] a good option for [your primary use case]?"
- "What are the pros and cons of [Brand Name]?"
For each response, note:
- Is your brand mentioned at all?
- Is the description accurate?
- Is the pricing correct (or at least in the right ballpark)?
- Are the features described correctly?
- What's the sentiment — positive, neutral, negative, or hedged?
- What sources does the model cite (especially in Perplexity, which shows citations)?
You'll often find inconsistencies between models. Claude might describe you accurately while ChatGPT has outdated information. Perplexity might cite a review from two years ago. Gemini might conflate you with a competitor. All of these are useful data points.
Layer 2: Category and problem queries
This is where most of the real buyer traffic lives. Someone who doesn't know your brand yet is asking:
- "What are the best tools for [your category]?"
- "How do I solve [problem your product solves]?"
- "What should I use for [specific use case]?"
- "Compare the top options for [your market]"
- "[Your category] for [specific buyer type, e.g. small teams, enterprise, agencies]"
Run 8-10 of these queries per platform. For each response, track:
- Does your brand appear?
- Where in the list (first, second, buried at the end)?
- What framing is used (recommended, mentioned as an alternative, noted as a niche option)?
- Which competitors are consistently appearing that you're not?
This layer is the most commercially important. These are the queries where buyers form their shortlists. If you're not here, you're not on the shortlist.
Layer 3: Competitor comparison queries
The third layer reveals where you're losing deals before they even start:
- "[Your brand] vs [Competitor]"
- "Is [Competitor] better than [Your Brand]?"
- "Alternatives to [Competitor]" (are you listed as one?)
- "Why do people choose [Competitor] over [Your Brand]?"
These queries are particularly revealing because they show you how AI models frame your competitive position. Sometimes you'll find that a model consistently recommends a competitor for reasons that are outdated or simply wrong — which is fixable if you know about it.
What to record and how to score it
Create a simple spreadsheet. Columns: Query, Platform, Mentioned (Y/N), Position, Accuracy (1-5), Sentiment (positive/neutral/negative), Notes.
Run the same queries across all four platforms. You'll end up with a matrix that shows you exactly where you have visibility, where you're missing, and where you're being described incorrectly.
A rough scoring framework:
| Score | What it means |
|---|---|
| Not mentioned | You don't exist for this query on this platform |
| Mentioned, inaccurate | Potentially worse than not mentioned |
| Mentioned, neutral | You're on the list but not recommended |
| Mentioned, positive | You're being actively recommended |
| Featured prominently | You're the first or primary recommendation |
The goal isn't to be mentioned everywhere — it's to be mentioned accurately and positively for the queries that matter to your buyers.
The four platforms behave differently
One thing that catches people off guard: ChatGPT, Claude, Perplexity, and Gemini don't all behave the same way, and their responses aren't consistent even for the same query.
ChatGPT (especially with browsing enabled) will pull from recent web content. Without browsing, it's working from training data that has a cutoff. This means your brand might be described accurately in one mode and outdated in another.
Perplexity is the most transparent — it shows you its sources. This is extremely useful for an audit because you can see exactly which pages it's pulling from. If it's citing a three-year-old review or a Reddit thread with complaints, you know what to address.
Claude tends to be more cautious and hedged. It often says things like "I don't have current pricing information" or "you should verify this directly." That's honest, but it also means it may be less likely to give a strong recommendation for brands it doesn't have solid data on.
Gemini is deeply integrated with Google's index, so your traditional SEO footprint matters more here. Strong Google rankings tend to correlate with stronger Gemini visibility.
Common things you'll find (and what they mean)
After running audits across dozens of brands, a few patterns come up repeatedly:
You're mentioned but with wrong pricing. This usually means the AI is pulling from old content — a review, a blog post, or a comparison page that hasn't been updated. Fix: publish fresh, clearly dated pricing pages and get them cited by authoritative sources.
You're not mentioned in category queries but appear in direct brand queries. The model knows you exist but doesn't associate you strongly with your category. Fix: you need more content that explicitly connects your brand to the category and use cases buyers are searching for.
Competitors are mentioned in "alternatives to [Competitor]" lists but you're not. This is a content gap — there's no page on your site (or elsewhere) that positions you as an alternative. Fix: create explicit comparison and alternative content.
You appear in some platforms but not others. This often comes down to which sources each model weights. Perplexity might cite sources that ChatGPT doesn't index. Fix: diversify where your brand is mentioned — not just your own site, but reviews, industry publications, Reddit, YouTube.
The sentiment is negative or hedged. The model is pulling from critical reviews or forum complaints. Fix: this is harder, but generating positive authoritative content that outweighs the negative sources is the long-term play.
Tools to scale beyond the manual audit
The manual audit is essential for understanding the landscape, but it's a snapshot. AI model responses change as models update, as new content gets indexed, and as competitors publish new material. You need ongoing tracking.
Several tools have been built specifically for this:
Promptwatch goes beyond monitoring to show you exactly which prompts competitors rank for that you don't, then helps you generate content to close those gaps. It tracks 10 AI models including ChatGPT, Claude, Perplexity, Gemini, and others, and includes AI crawler logs so you can see which pages AI bots are actually reading on your site.
For teams that want dedicated AI visibility tracking, a few other options worth knowing:
Otterly.AI
Profound
Here's a quick comparison of what these tools cover:
| Tool | Platforms tracked | Content generation | Crawler logs | Citation analysis | Best for |
|---|---|---|---|---|---|
| Promptwatch | 10 (ChatGPT, Claude, Perplexity, Gemini, Grok, DeepSeek, Copilot, Meta AI, Mistral, Google AI) | Yes (built-in AI writer) | Yes | Yes | Teams that want to track AND fix |
| Otterly.AI | ChatGPT, Perplexity, AI Overviews | No | No | Basic | Basic monitoring |
| Peec.ai | ChatGPT, Perplexity, Claude | No | No | No | Simple tracking |
| Profound | 9+ AI engines | No | No | Yes | Enterprise monitoring |
| Rankshift | ChatGPT, Perplexity, AI search | No | No | Basic | Brand tracking |
The core difference: most of these tools will tell you where you're invisible. Promptwatch also tells you what to do about it — which prompts to target, what content to create, and whether that content is working.
What actually moves the needle
Once you know where your gaps are, the question is how to fix them. A few things that actually work:
Publish content that directly answers the queries where you're missing. AI models cite sources. If there's no page on your site (or anywhere on the web) that clearly answers "what is [Your Brand] good for in [specific use case]," you won't appear for that query. Create that page.
Get cited by sources AI models trust. Perplexity's citation panel is a cheat sheet here — look at what it cites for your category queries and figure out how to get mentioned on those sites. Industry publications, well-trafficked review sites, and authoritative blogs all matter.
Fix outdated information at the source. If a major review site has your old pricing or an outdated feature description, update it. Reach out to the publication. AI models will keep pulling from that source until it changes.
Don't ignore Reddit and YouTube. Both platforms are heavily cited by AI models, especially for "what do real users think" type queries. A well-written Reddit comment or a YouTube review can influence what AI models say about your brand more than a press release.
Be consistent across platforms. Inconsistent descriptions across your website, G2, Capterra, and third-party reviews create confusion for AI models. They average out the signals — if half your reviews say you're great for enterprise and half say you're for small teams, the model will hedge or pick one arbitrarily.
How often to run this audit
The manual audit is worth doing quarterly at minimum. AI models update, competitors publish new content, and your own content changes. What was true in January may not be true in April.
For ongoing monitoring, a weekly or bi-weekly automated check across your core queries is more realistic. That's where tracking tools earn their keep — you don't want to manually run 50 queries across four platforms every week.
Set up alerts for significant changes: if you drop out of a category query you were previously appearing in, that's worth investigating immediately. If a competitor suddenly starts appearing in queries where they weren't before, that's also worth understanding.
The brands that will win in AI search are the ones treating this as an ongoing discipline, not a one-time project. The audit gets you started. The tracking keeps you honest.