A 30-minute manual AI visibility audit
No tool required: 20 prompts, two AI assistants, one spreadsheet. The free baseline that tells you whether you have a discovery problem or an optimization problem.
A manual AI visibility audit takes 20 buyer prompts, asks them in ChatGPT and Perplexity, and logs which answers mention your brand. Thirty minutes of work, zero budget — and the result tells you which of two very different problems you have: undiscovered (never mentioned → crawler/definition issue) or underquoted (mentioned rarely → content/mentions issue).
Step 1: collect 20 prompts (15 min, once)
- 5 category questions ("what is [category]", "how does [category] work").
- 10 comparison questions ("[your category] vs [alternative]", "best tool for [job]"). Mine the last month of sales calls for exact buyer wording.
- 5 problem questions ("how do I solve [problem]"). These reveal consideration sets you didn't know existed.
Step 2: ask both assistants (10 min, weekly)
Same 20 prompts, same week, in ChatGPT and Perplexity (free tiers suffice). Use fresh chats, no follow-ups — follow-ups contaminate the sample. Log per answer: mentioned yes/no, position in answer (first/half/afterthought), description accurate yes/no, competitors named.
Step 3: read the pattern (5 min)
- Zero mentions everywhere → discovery problem. Check robots.txt crawler access, publish a category-definition page, then re-run in two weeks.
- Mentions in Perplexity, silence in ChatGPT → the most common split. ChatGPT leans on training data and broad coverage: push third-party mentions and entity clarity.
- Mentioned but misdescribed → accuracy problem. Find the source being quoted (usually an outdated roundup) and fix it at the source, then on your site.
Free telemetry you're probably ignoring
Check GA for utm_source=chatgpt.com referrals — AI referral traffic is measurable today without any tooling, and it validates whether visibility turns into visits. Pair it with assisted conversions in GA4 before declaring any channel ROI.
Related reading
Frequently asked questions
How much time does the manual audit really take?
About 30 minutes once the prompt list exists: 20 prompts × ~1 minute across two assistants, plus 10 minutes logging. Building the prompt list the first time takes another 30–60 minutes from sales calls and search data.
When should I switch from manual to a tool?
When the set grows past ~50 prompts, when you need daily (not weekly) data, competitor tracking, or alerts on description changes. Manual auditing is a baseline method, not a scaling method.
Check how visible your brand is in AI answers
See where ChatGPT, Perplexity and Gemini cite you — and where competitors win. Start your free 7-day trial.
Try ChatBakers free