How to Run an AI Visibility Audit for Your Employer Brand

What an AI Visibility Audit Is

An AI visibility audit is a structured check of what AI systems say when someone asks about your organization. You ask the questions a candidate, customer, or journalist would ask, record the answers from each major platform, and score them for accuracy, completeness, and source. The result is a baseline you can fix against and measure again later.

It differs from a traditional SEO audit. An SEO audit asks where your pages rank. An AI visibility audit asks whether the answer an AI gives is about the right company, says true things, and draws on sources you would want it to trust. For background on the concept, see What Is AI Visibility? and AI Visibility vs Traditional SEO.

You need a spreadsheet, an hour or two for a first pass, and clean browser sessions. No special tools are required for a manual audit, though tracking tools save time once you repeat it.

Step 1: Define the Scope

Decide what you are auditing before you ask anything.

• Entity names: List every name your organization goes by, including the legal name, the brand name, abbreviations, former names, and common misspellings.
• Platforms: Include ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, and Grok. Answers differ across platforms, so one platform is not enough.
• Audience and market: Note which regions and languages matter. An answer in one market can differ from another.
• Topics: For an employer brand, cover culture, benefits, pay, management, interview process, career growth, diversity and inclusion, and leadership.

Writing this down first keeps later audits comparable. Changing the scope between rounds makes the results impossible to compare.

Step 2: Build the Question Set

Write 10 to 20 questions in the words real people use. Start with these patterns and adapt them to your organization:

• What is it like to work at [Company]?
• Is [Company] a good place to work?
• How does [Company] treat its employees?
• What benefits does [Company] offer?
• How is management at [Company]?
• What is the interview process at [Company]?
• Does [Company] offer career growth?
• What do employees say about [Company]'s CEO?
• Has [Company] had layoffs?
• Who owns [Company]?

Add a few unbranded questions too, such as "best companies to work for in [your industry]," to see whether you appear when the person doesn't know your name yet.

Step 3: Run the Questions Cleanly

Consistency matters more than volume. For each question and each platform:

1. Open a fresh session so earlier conversation doesn't shape the answer.
2. Ask the question exactly as written.
3. Save the full answer, the date, the platform, and any cited sources.
4. Ask a second phrasing of the same question and compare the two answers.
5. Repeat the run a few days later. AI answers can change from one run to the next, so a single answer is a sample, not a verdict.

Keep a log with one row per answer. A simple structure works:

Date	Platform	Question	Answer saved	Entity correct?	Facts accurate?	Sources cited

Step 4: Score Each Answer

Score every answer on a short rubric so the results can be compared across platforms and over time. A three-point scale for each criterion is enough:

• Identity: Is the answer about your organization at all? (0 = wrong company, 1 = mixed, 2 = correct)
• Accuracy: Are the stated facts true and current? (0 = false claims, 1 = partly outdated, 2 = accurate)
• Completeness: Does it cover what a candidate needs? (0 = vague or empty, 1 = partial, 2 = useful)
• Sources: Does it cite sources you would want it to trust? (0 = only noise sources, 1 = mixed, 2 = verified sources)
• Tone: Is the overall picture fair? (0 = distorted, 1 = skewed, 2 = balanced)

Add the scores across questions and platforms. The totals show where the problems cluster, such as one platform getting your identity wrong, or one topic like pay being consistently thin.

Step 5: Map the Sources

For every answer that cites sources, record what it cited. Group them: your own site, news coverage, employee review sites, social threads, directories, and third-party verified profiles.

This step often explains the whole audit. If the answers about your culture keep drawing on a handful of old review posts, those posts are shaping your reputation in AI, whatever your careers page says. If a verified source such as a certification profile appears in answers, you can see it working. Noting which sources the models favor tells you where new, verified content will have the most effect. To understand why platforms favor certain sources, read How AI Search Engines Choose Sources.

Step 6: Check for Identity Problems

Review every answer where the identity score was 0 or 1. The cause is usually one of three:

• A sound-alike: A similarly named organization absorbs your information.
• A former parent or acquirer: An older, larger entity keeps the narrative after a spinout, rebrand, or acquisition.
• A thin record: There is too little verified information about you, so the model fills gaps with whatever it finds.

Identity problems come first in the fix order, because every other improvement is wasted if the model is describing someone else. If you find invented facts, treat them as a separate category and read AI Hallucinations About Your Brand.

Step 7: Run a Technical Check

Check the foundations that help systems identify you correctly:

• Organization schema: Your site should publish structured data naming the organization, with consistent legal name, description, and logo.
• sameAs links: Structured data should link your official profiles, such as LinkedIn and verified directory listings, so a model can connect them as one entity.
• Consistent naming: The same canonical name should appear across your site, careers pages, and profiles.
• Crawl access: Confirm that important pages aren't blocked from the crawlers that feed AI systems, and that key facts appear as text on the page, not only inside images or scripts.
• Freshness: Check that leadership, headcount, and ownership details on your own pages are current.

Step 8: Prioritize and Assign Fixes

Rank the findings in this order:

1. Wrong entity: Answers about the wrong organization.
2. False facts: Specific claims that are untrue.
3. Outdated facts: True once, no longer true.
4. Missing facts: Topics where the answer is empty or vague.
5. Tone and sources: Fair but skewed answers drawing on weak sources.

For each item, assign an owner, the fix, and a check date. Typical fixes include publishing structured, verified content, correcting the facts on your own properties, adding third-party verified profiles, and building cross-references across independent sources. A real example of this sequence is the DFIN case study, where an employer brand that AI platforms could not see became one that five of six major platforms mentioned, after structured entity data, 21 published articles, and 1,238 tracked AI mention checks.

Step 9: Re-Run on a Schedule

AI systems are updated continuously, so a clean audit today doesn't stay clean. Re-run the full audit monthly for the first quarter, then at least quarterly. Keep the same question set so results are comparable, and track the trend in three numbers: the share of answers with the correct entity, the average accuracy score, and the number of platforms that mention you at all.

Common Audit Mistakes

Using one platform. Each system draws on different sources, so a clean result on one says little about the others.

Treating one answer as the result. Answers vary between runs. Repeat the questions and look for patterns, not single outliers.

Asking leading questions. "Why is [Company] such an outstanding employer?" produces a flattering answer by design. Use neutral phrasing that matches how real people ask.

Skipping the log. Without dates, platforms, and saved answers, you can't show progress or prove that a fix worked.

Fixing content before fixing identity. More pages about an entity the model confuses with another only adds to the confusion. Settle identity first.

What a Good Result Looks Like

A healthy audit shows the right organization named on every platform, accurate and current facts, answers that cover what candidates ask, and citations that point to verified sources rather than only to anonymous threads. Perfect scores are unrealistic, because answers vary. The goal is steady improvement and no identity errors.

How Visipage Helps

Visipage runs this audit continuously across the major AI platforms, builds the verified profile and structured data that give models a reliable source, and re-checks on a monthly cadence. See the methodology for how it works, or run a free check to see how AI describes your organization today.