AI visibility audit checklist for B2B brands

What should an AI visibility audit prove?

It should show whether the brand can be discovered, whether its pages contain useful retrievable evidence, whether platform answers describe it accurately, and whether visibility contributes to qualified behavior.

The audit is not a promise that a platform will cite the brand. It is a structured record of eligibility, evidence, observed answers, gaps, and next actions. Because answer outputs change, keep the prompt set and testing method stable enough to compare over time.

Start with buyer questions that connect to a real decision. A hundred vanity prompts about the brand name can look impressive while creating no commercial insight.

1. Check technical discovery

Confirm successful status codes, rendered main content, internal links, canonical consistency, sitemap coverage, and the crawler rules for platforms you intend to serve.

Crawler controls are platform-specific. OpenAI documents OAI-SearchBot separately from GPTBot. Google documents Google-Extended as a separate control and says it does not affect inclusion or ranking in Google Search.

• Googlebot access for Google Search features
• OAI-SearchBot policy for ChatGPT search discovery
• PerplexityBot policy for Perplexity search
• CDN and bot protection logs for unexpected blocking
• No important answer hidden only inside images or interactions
• Stable canonical URLs and internal discovery paths

2. Check answer usefulness and extractability

Pages should answer the question directly, explain when the answer changes, name evidence, and remain understandable when a passage is read out of context.

• One clear page intent
• A direct answer near the relevant heading
• Specific examples, methods, and limitations
• Primary sources for platform or legal claims
• Useful tables and checklists where comparison matters
• Accurate publication and update dates

3. Check entity consistency

Names, people, services, locations, contact details, pricing context, and claims should agree across the website and independent profiles.

Create a fact ledger with the canonical statement, source URL, owner, and last review date. When an answer system gets a fact wrong, first ask whether the public evidence is inconsistent or weak. Correct your own sources before blaming retrieval.

4. Check whether claims are corroborated

Original evidence and relevant independent mentions are more useful than a larger volume of interchangeable summaries.

Review which domains answer the target questions today. Classify them as primary sources, specialist publications, communities, directories, competitors, or aggregators. Look for evidence gaps Branddirr can honestly fill, not passages to imitate.

5. Build a stable buyer-question set

Use real sales questions, Search Console queries, community language, and service decisions. Record the exact prompt, platform, date, location context, answer, cited sources, and accuracy.

6. Connect visibility to behavior

Track cited URLs, answer-engine referrals, engaged sessions, service-page movement, diagnosis starts, qualified opportunities, and assisted conversions.

Referral data can be incomplete, and some answer interactions will not create a click. That is why prompt observation and web analytics belong together. Neither alone proves commercial value.

7. Prioritize fixes by evidence and business value

Fix blocking eligibility and factual errors first. Then improve pages tied to valuable buyer decisions where the brand has real expertise and a credible service path.

• Blocked or broken discovery
• Incorrect public business facts
• Missing answer on a high-value buying question
• Weak firsthand evidence on a relevant page
• Inconsistent service or person entities
• Cited page with no useful conversion path
• Unmeasured referrals or duplicate conversion events