AI visibility audit template for baselines that hold up over time
Sample the prompts your buyers use, record who gets cited, and track share across engines instead of checking once.
- Google Sheets
- Notion
- Looker
- ChatGPT
- Perplexity
- Google Gemini
Baseline our visibility for "ai marketing agent" prompts across ChatGPT and Perplexity.
- Locked the prompt set32 prompts sourced from sales calls and search-before-search
- Ran each prompt 8xAcross ChatGPT, Perplexity, Gemini, and Google AI Overviews
- Scored share of answerMention 41%, citation 18%, competitor cited 34%
- Diffed against last quarterCitation share up 6pts; two prompts newly cited
Citation share moved from 12% to 18% quarter over quarter, driven by two prompts where the FAQ rewrite landed. The competitor cited most often is the one with the freshest stats — a signal to update ours.
An ai visibility audit template is the artifact that turns a vague question — are we showing up in ChatGPT? — into a measured one. It samples the prompts your buyers actually use, runs them across the engines that matter, and records whether you appear, how you are described, and which sources are quoted instead of you. Because generation is probabilistic, the baseline repeats prompts and aggregates rather than checking once.
What is an ai visibility audit template?
A repeatable sampling procedure: a fixed prompt set, a fixed engine list, a run count, and a scoring rubric that records mention, citation, and description quality per prompt. The template fixes the method so two baselines a quarter apart are comparable.
Build the prompt set
- Source prompts from sales, support, and search-before-search behavior
- Separate commercial prompts from definitional ones
- Flag prompts where AI overviews are appearing in classic SERPs
Score and aggregate
- Run each prompt five to ten times per engine
- Record mention, citation, and description quality
- Report share of answer, not single-run outcomes
Which prompts should the ai answer engines be sampled with?
The prompts your buyers use, not the ones you rank for. Source them from sales calls, support tickets, and the queries people type into ChatGPT before they type into Google. A prompt set built from your own keyword list measures SEO, not visibility.
How often should llm citations be baselined?
Monthly for the top prompts, quarterly for the long tail. Generation variance means a single sample tells you nothing — the baseline repeats each prompt five to ten times and reports the share, not the single outcome.
What does an ai overview tracking baseline record?
For each prompt: whether you appear, whether you are cited, which competitors appear, which sources are quoted, and the wording used to describe your category. The wording matters — it is how a buyer who has never visited your site forms their impression.
How the ai visibility audit template fits your stack
The agent pulls your tracked competitors and topic list from workspace Knowledge, runs the prompt set across ChatGPT, Gemini, Perplexity, and Google AI Overviews, and writes the baseline to Sheets or Notion so the next run diffs against it. Results export to your reporting tool for the monthly review.
- Google Sheets
- Notion
- Looker
- ChatGPT
- Perplexity
- Google Gemini
Who uses this ai visibility audit template
- SEO teams
- Move from tracking rankings to tracking citations.
- Brand marketers
- See how the category is described when you are not in the room.
- Founders
- Know whether AI answer engines are sending buyers your way.
How to run this ai visibility audit template in Metaflow
Lock the prompt set
Twenty to fifty prompts your buyers use, sourced from real conversations.
Set the run count
Five to ten runs per prompt per engine — enough to smooth variance.
Run the baseline
The agent executes, records, and aggregates share of answer.
Diff against last quarter
The report shows what moved and which prompts to dig into.
What you provide
- Buyer prompt set
- Engine list
- Run count
- Tracked competitors
What you get back
- Baseline report
- Share-of-answer table
- Description wording log
- Quarter-over-quarter diff
Why use this ai visibility audit template?
Repeats prompts and reports share, not a single yes/no
Sources prompts from buyer behavior, not your keyword list
Records the wording used to describe your category, not just presence
Diffs against the prior baseline so movement is visible
AI visibility audit template FAQs
How do you measure AI visibility?
Sample the prompts your buyers use, run them across engines repeatedly, and record mention and citation share per prompt. One run is noise — the baseline repeats and aggregates so the trend is the signal.
Which AI answer engines should you track?
The ones your buyers use: ChatGPT, Google AI Overviews, Perplexity, and Gemini. Claude matters for developer-facing brands. Track all of them, because share moves independently across engines.
How often should you run an AI visibility audit?
Monthly for the prompts that map to revenue, quarterly for the long tail. Anything less frequent and a drift goes unnoticed for too long; anything more frequent and you spend on noise.
What is a good AI visibility score?
Share of answer — the fraction of prompts where you are mentioned or cited — not a single number. A baseline of 30% share that holds steady is healthier than 80% share that swings every run.
Similar AEO checklists and AI visibility templates
Key takeaways
- Visibility is share of answer over time, not a single check
- Source prompts from buyer behavior, not your keyword list
- Repeat prompts and aggregate — single runs are noise