How to run an AI visibility audit without buying another platform

What does an AI visibility audit measure?

AI search has created a new reporting problem for businesses.

A team might rank well in Google and still wonder whether buyers see the brand when they ask ChatGPT, Gemini, Perplexity, Copilot, or Google's AI features for recommendations and comparisons.

The wrong response is to chase a new acronym every week.

The better response is to build a baseline.

An AI visibility audit asks a practical set of questions.

• Where does the brand appear?
• Where does it not appear?
• Which competitors appear more often?
• Which sources get cited?
• Is the information accurate?
• Which pages or third party mentions seem to shape the answer?
• Does the website have a crawl or indexing problem?
• Which gaps connect to real buyer intent?

That gives the team something measurable before it starts changing content.

Google's current guidance matters here.

For AI Overviews and AI Mode, Google says the same core SEO practices still apply. A page needs normal Search eligibility. Helpful, original content, internal links, crawl access, page experience, useful images, and accurate structured data still matter.

Google also says there is no special AI schema or required AI text file for those Search features.

That removes a lot of noise from the audit.

Build the audit

Start with buyer prompts, not brand ego.

Create a set of questions people might ask before choosing a provider, tool, service, or product.

Use a mix.

Category prompts: "best website designer for SaaS"

Problem prompts: "who helps fix a website that gets traffic but few leads"

Comparison prompts: "Framer designer or development agency for a marketing site"

Evaluation prompts: "website redesign partner with SEO and analytics"

Branded prompts: "what does Ahmed Jamal design"

The exact prompt set depends on the business.

Keep it stable enough to repeat later.

Then run the prompts across the AI systems that matter to the audience.

Record the answer.

Do not only record whether the brand appears.

Track: Mention. Position in the answer. Citation. Source cited. Accuracy. Competitors mentioned. Type of recommendation. Useful or misleading description.

AI answers vary.

That means one run does not deserve a large conclusion. Ahrefs recommends a structured audit that tracks mentions, citations, share of voice, source patterns, branded accuracy, and competitor gaps over time.

Next, check Google separately.

As of August 31, 2026, Google's generative AI performance reports in Search Console are available worldwide. They show impressions, pages, countries, devices, and dates for visibility in generative AI features.

That data belongs in the audit because it comes from Google's own Search Console rather than a third party estimate.

Then review the website foundation.

• Is the important content indexed?
• Does Search Console show crawl or indexing issues?
• Are important pages linked clearly?
• Is the main content present as readable text?
• Does the site explain the business clearly?
• Are service pages specific?
• Are case studies detailed enough to support trust?
• Does structured data match the visible page?
• Do third party sources describe the brand accurately?

Repeat a stable audit sample

Do not turn the findings into fifty new articles.

Google's current generative AI guidance warns against creating separate pages for every possible query variation.

That is useful advice beyond Google.

Fix the highest value gaps first.

If AI answers describe the company incorrectly, improve the primary source pages and external profiles that support the facts.

If competitors get cited for a topic you genuinely know well, publish a stronger page with first hand detail.

If the site has thin generic content, replace summaries with real process, examples, comparisons, decisions, and evidence.

If important pages are not indexed or easy to find, fix the technical structure.

If third party sources matter, work on real mentions, reviews, community participation, partnerships, or useful public work rather than manufactured citations.

Then repeat the same audit later.

Monthly works for an active search program. Quarterly is enough for many smaller businesses.

The goal is not to "win AI" in one sprint.

The goal is to understand where the brand appears, why it appears, where it is missing, and which improvements also make the website more useful for real people.

That is a much better starting point than chasing another trick.