AI Visibility Audit: Step-by-Step Method

Step 1: Decide what the audit must answer

Write down the brand, its product names and aliases, three to five real competitors, the market and language, and the AI surfaces you will look at. A consumer chat app, a model API response, Google AI Overviews, and AI Mode are different observations, so name the surface beside every result.

Choose one job for this audit: finding whether the brand is mentioned at all, checking how it is described, or finding which sources the answers rely on. An audit that tries to answer all three at once usually answers none of them well.

Step 2: Build a question set from real buyer decisions

Use category discovery questions, use-case recommendations, named comparisons, pricing and implementation questions, and a few direct questions about your own brand facts. Take the wording from sales calls, support tickets, and search queries rather than inventing it, and keep the set small enough that a person can read every answer.

Freeze the list before you collect anything. If you add or reword questions mid-audit, you can no longer compare the first answers with the last ones. The prompt portfolio guide covers how to balance the set.

Discovery: “Which tools help teams track how AI describes their brand?”
Comparison: “How does Brand A differ from Brand B for a mid-size team?”
Fact check: “What does Brand A cost, and what is included?”

Step 3: Collect answers you can reread

Ask each question more than once, because generated answers vary. For every attempt, save the exact prompt, date, surface or model, market, the full answer, any displayed source URLs, and whether the attempt was usable. Count failed or empty attempts separately; a failed run is a missing observation, not evidence that the brand is absent.

Keep the denominator visible. A mention rate means little unless the report says how many valid answers it is based on.

Step 4: Score presence, role, accuracy, and sources separately

Read each answer for four different things and record them in separate columns. A brand can be named but described wrongly, recommended without any source, or cited for a page that does not support the claim. Folding these into one number hides the gap you need to fix.

Step 5: Check the pages and facts the answers depend on

Compare what the answers say with your own pages. Look for outdated pricing, inconsistent product names, missing comparison or documentation pages, and facts that appear only in a hard-to-read format. Confirm that the crawlers you want to reach public pages are not blocked by robots rules or by pages that require a login.

A crawler check on your key URLs is a quick, concrete part of this step. It tells you whether a page can be fetched; it does not tell you whether an AI system will use it.

Step 6: Compare competitors and group the sources

Run the same questions for each competitor and compare presence and framing over the same valid-answer set. Then group the displayed source URLs: your own pages, independent reviews, community threads, and reference pages play different roles. A competitor named more often may simply reflect a question set tilted toward its use case, so inspect results question by question before drawing a market-level conclusion.

Step 7: Write up three findings and one test

Report the scope, the question count, the valid-answer count, the three clearest gaps with example answers, and the sources involved. Pick one bounded change with a named owner, such as correcting a product fact, adding a missing comparison page, or earning one relevant independent reference.

Record the baseline and schedule a recheck with the same question set. A changed answer afterwards is an observation; it does not on its own prove which change caused it.

Compare the approaches

Audit layer	What you check	Common mistake
Presence	Whether the brand is named in valid answers, and in what role	Counting failed runs as “not mentioned”
Accuracy	Whether pricing, features, and category are described correctly	Treating any mention as a good result
Sources	Which URLs the answers show, and whether they support the claim	Assuming a displayed source caused the recommendation
Site readiness	Whether key pages are crawlable, current, and consistent	Fixing markup before fixing missing facts
Competitors	Who else appears for the same questions	Comparing different question sets

Common questions

What is an AI visibility audit?

It is a dated review of how AI-generated answers mention, describe, and cite a brand for a defined set of buyer questions, followed by a short list of evidenced gaps and a plan to recheck them.

How often should I run an AI visibility audit?

Run a full audit when you set a baseline and after a meaningful content or positioning change. Between audits, re-ask the same fixed question set on a regular schedule so you can separate normal answer variation from a real shift.

Can a free AI visibility checker replace an audit?

A free checker can give a quick first look, but check which surface it queried, how many answers it based the result on, and whether you can read the full answers and sources. A score you cannot inspect is hard to act on.

Does an audit show why an AI system recommended a competitor?

It shows what the answers said and which sources they displayed. It cannot show hidden model weights, so treat displayed sources as evidence of what the interface showed, not as proof of cause.

Sources and methodology

Digraph publishes this guidance and provides AI visibility software. Worked examples are illustrative unless explicitly identified as measured results. A research sample does not establish total market coverage or guarantee future recommendations.