The AI Visibility Audit: 10 Questions Every Marketing Team Should Be Asking

Most marketing teams have no idea how they appear inside AI assistants — which means they can't tell whether they're winning share of voice or quietly losing it to competitors. The fastest way to fix that is a short, honest audit. Ten questions. Each one either has a data-backed answer or reveals a blind spot you should close this quarter.

Use this list as-is. Score yourself out of ten. Anything below eight is a strategy running on assumption.

1. Does ChatGPT recommend us for our core products?

The most direct question there is. For the two or three prompts that describe what you sell in your customers' words, does ChatGPT name you? If it does, you have baseline visibility. If it doesn't — and your competitors are named — you have a citation gap on the exact terms that matter most.

2. Which competitors appear more often than we do?

Every category has an AI shortlist of three to five brands. Knowing who is on it — for your category, right now — tells you who the AI has anointed as the reference set. Beating one of them off the list is a concrete, quarterly-scale goal.

3. Which prompts never mention us?

Coverage gaps aren't random. They cluster around specific topics, specific comparisons, and specific buying stages. A list of prompts where you never appear is a content brief in disguise.

4. Which sites does AI cite instead?

When the AI links out, whose pages does it pick? Usually the same handful of domains — review sites, industry publications, category leaders. Those are the citation surfaces you need presence on. They're also the sites you should be reading to understand how your category is being described.

5. Are we recognised as an authority in our space?

Authority in AI shows up in two ways: being named without qualification ("X is a leader in Y"), and being linked to as a source. If you're being mentioned only as a member of a list, you have presence but not authority. Authority is what makes you the default, not the alternative.

6. Which topics do competitors own?

Some topics belong to a competitor in the model's eyes. Ask the AI a question in that area and the same brand comes up every time. Knowing which topics they own tells you where to compete directly, where to reposition around, and where to let go.

7. Has our AI visibility improved over time?

A single snapshot is anecdote. A trend line is signal. Whether your citation frequency is climbing, flat or falling over months is the single most important number in an AI visibility programme.

8. Are product pages or educational content driving our citations?

When AI cites you, what does it cite? If it's mostly product pages, your marketing content isn't earning trust. If it's mostly educational content, your product surface isn't answerable enough. Balanced citation across both is a healthy sign.

9. Do different AI models produce different answers about us?

They will — and the pattern matters. Being strong in ChatGPT and absent from Perplexity means one training corpus knows you and another retrieval system doesn't. Each engine has different levers: fresh content, structured data, authoritative third-party coverage, llms.txt.

10. What's our biggest AI content gap right now?

If you had to ship one content asset this month to close a real visibility gap, which prompt would it target and what shape would the answer take? A team that can name that piece specifically is running a strategy. A team that can't is running a hope.

Using the answers

Not every question has to score well immediately. The point of the audit is to expose which ones you can answer — and to make the ones you can't the next quarter's work. Ten questions. Ten measurable answers. That's the strategy.

FAQ

Do I need a tool to answer these? For prompt-level, cross-engine measurement, effectively yes — it's too much to do manually at any useful cadence. For a first pass, sampling by hand across the four major engines is fine.

How often should we re-audit? Monthly is a reasonable cadence for the full ten. Weekly for the citation trend line (question 7) and the priority prompts (questions 1 and 3).