How to run an AI visibility audit in one afternoon
By haylow.ai editorial team
In short
An AI visibility audit is a one-time review of whether AI assistants can read your site, whether they name your brand for your key questions, and which sources they cite instead.
What is an AI visibility audit?
It is a structured review of how AI assistants treat your brand today. You check that they can reach your site, test whether they name you for the questions customers ask, and note which sources and competitors appear instead. The output is a ranked list of fixes.
Set aside an afternoon and one person. You need a spreadsheet, access to the main assistants and your site's robots.txt and sitemap.
Step 1: Which questions should you test?
Write twenty questions a real customer would type: category questions, comparison questions and problem questions. Avoid questions that already contain your brand name, because they test recall, not discovery. Keep the wording fixed so you can repeat the audit.
Category
What is the best tool for X, who offers X in my city.
Comparison
X vs Y, alternatives to Y, is X worth it.
Problem
How do I fix X, how do I measure X.
Step 2: How do you test the assistants?
Run every question in ChatGPT, Gemini, Claude and Perplexity, and in Grok if your audience uses it. Record whether you appear, your position in any list, the wording used about you and the sources cited. Run each question twice, on different days, because answers vary.
Copy the full answer into your sheet. A yes or no hides the detail you will need later, such as the description and the cited sources.
Step 3: What technical checks should you run?
Confirm robots.txt does not block answer-engine crawlers, the sitemap lists your key pages, pages are indexable with a canonical address, and you have clear titles, descriptions and structured data. A free technical checker covers most of this in a few seconds.
Fix blockers first. A blocked crawler makes every other effort pointless for that engine.
Step 4: Which sources do assistants cite instead of you?
List the domains cited across all answers and count how often each appears. The most repeated ones are where your category is being described. Those sites, whether reviews, directories, news or competitors, become your priority list for earning a mention.
Also note which competitors are named most often and how they are described. That is the standard you are being compared against.
Step 5: How do you turn the audit into a plan?
Rank findings by impact and effort and keep five. Typical top items: remove a crawler block, add a direct answer and FAQ to a key page, create one comparison page, earn one placement on a frequently cited source, and set up repeated tracking to measure change.
Re-run the same questions in four to six weeks. Because you kept the wording and stored the answers, you can see what moved.
Key takeaways
- An audit has three parts: technical access, brand mentions and cited sources.
- Use twenty real customer questions, not brand-name queries, or you will measure the wrong thing.
- Record every answer in full so you can compare after you make changes.
- End the audit with a ranked list of five fixes, not a long report.
Frequently asked questions
How long does an AI visibility audit take?
A first pass takes an afternoon for twenty questions across four assistants. A repeat run is faster because the questions and sheet already exist.
How many questions should I test?
Twenty is a workable starting set. Fewer than ten gives a weak picture; more than fifty is better handled by a tracker.
Can I automate an AI visibility audit?
Yes. A tracker runs the same questions on several assistants and stores the answers, which makes repeat audits quick and comparable.
What is a good result?
There is no universal benchmark. Aim to be named in most runs for your core questions and to improve against the competitors named alongside you.
See what AI says about your brand
Add one domain and get your first visibility report across five assistants. Free to start, no card required.