Your first AI visibility audit: a step-by-step walkthrough

What is an AI visibility audit?#

An AI visibility audit evaluates how ready your website is to be cited by AI search engines. It checks three layers: whether AI crawlers can access your content, whether your content is structured for citation, and whether AI platforms already mention your brand.

Think of it as a health check for your presence in the AI search world. A traditional SEO audit checks your Google ranking factors. An AI visibility audit checks the factors that ChatGPT, Claude, Perplexity, Gemini, and other AI engines use to decide citations.

The audit produces an actionable score with specific fixes. You will know exactly which crawlers are blocked, which schema types are missing, which pages lack citation-ready structure, and how your brand performs across each AI platform. BrandCited automates this entire process, but understanding each step helps you interpret results and prioritize fixes.

Step 1: Check your crawler access#

Open your robots.txt file at yourdomain.com/robots.txt. Look for rules that mention these AI crawler user-agents: GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, anthropic-ai, Google-Extended, PerplexityBot, Bytespider, CCBot, Applebot-Extended, and FacebookBot.

If you see "Disallow: /" for any of these bots, that AI platform cannot crawl your content and will not cite you. If your robots.txt has a blanket "User-agent: * / Disallow: /" rule without specific exceptions for AI bots, every AI crawler is blocked.

Check beyond robots.txt too. Cloudflare Bot Fight Mode, CDN-level bot blocking, server-side WAF rules, and IP-based rate limiting can all block AI crawlers even when robots.txt allows them. Review your server logs for the past 30 days and look for hits from these user-agents. If you see no AI crawler activity despite allowing them, something else is blocking access.

Score this step pass/fail for each platform. A single blocked crawler means zero citations from that AI engine.

Step 2: Audit your structured data#

Visit your homepage and key landing pages. View page source and search for "application/ld+json" to find existing schema markup. Check for these types:

Organization schema on your homepage tells AI engines who you are, what you do, and where to find you online. Without it, AI models piece together your identity from scattered signals across the web. With it, they have a definitive reference.

Article schema on blog posts and guides signals authorship, publication date, and topic. AI engines use this to assess recency and author authority. Anonymous articles without dates get skipped.

FAQPage schema on pages with Q&A content gives AI engines pre-structured question-answer pairs ready for extraction. Pages with FAQ schema receive roughly 40% more citations than similar pages without it.

HowTo schema on tutorial content maps directly to "how to" queries that users ask AI assistants. If you teach processes, this schema captures those queries.

Product schema on pricing and product pages helps AI engines answer comparison and recommendation queries. When someone asks "best CRM under $50 per month," Product schema gives the AI your pricing data.

For each schema type, check: Is it present? Is it valid (no errors in Google's Rich Results Test)? Is the data accurate and current? Missing or broken schema is a gap you can close in a single development sprint.

Step 3: Evaluate your llms.txt file#

Navigate to yourdomain.com/llms.txt. If you get a 404, you do not have an llms.txt file, and you are missing a direct communication channel with AI engines. As of early 2026, only 5-15% of websites have implemented llms.txt. That gap is your opportunity.

If you have an llms.txt file, evaluate its quality. Does it start with an H1 containing your brand name? Does it include a blockquote description that explains what your site does in factual (not marketing) language? Does it list your 20-50 most important pages organized by section? Are the URLs current and not pointing to 404 pages?

Common mistakes to flag: including every page from your sitemap (dilutes the signal), using promotional language in descriptions ("industry-leading world-class platform" teaches crawlers to distrust your file), and not updating the file when content changes. The best llms.txt files are curated, factual, and maintained quarterly.

Step 4: Score your content readiness#

Pull up your ten most important pages. For each page, check five content factors.

Does the page lead with a direct answer? The first 60 words of each section should contain a clear, self-contained answer to the question posed by the heading. AI engines extract these opening blocks for citations. Content that builds up to an answer in paragraph three loses to content that states the answer first.

Does the page include specific data? Numbers, percentages, dates, and named sources signal authority. "We reduced customer churn by 34% over six months using automated onboarding sequences" gets cited. "We saw significant improvement in retention" does not.

Are headings structured as questions? Headings like "How much does schema markup cost?" match how users query AI assistants. Headings like "Implementation Details" do not align with query patterns.

Does the page contain citation-ready blocks? Look for 40-60 word paragraphs that can stand alone as complete answers. If every paragraph depends on surrounding context to make sense, AI engines cannot extract a clean citation.

Is the content comprehensive? Pages that cover a topic with multiple sections, specific examples, and practical steps signal depth. A 300-word overview loses to a 2,000-word guide with eight sections on the same topic.

Score each page from 1-5 on each factor. Pages scoring below 15 out of 25 need rewriting.

Step 5: Check your current AI citations#

The final step measures your actual performance across AI platforms. Open ChatGPT, Claude, Perplexity, Gemini, and Grok. Ask each one 5-10 questions that your target customers would ask about your industry or product category.

Record whether your brand appears in the response, whether it is recommended positively, and which competitors get mentioned instead. This manual process works for a baseline check but does not scale for ongoing monitoring.

BrandCited automates this step by querying all seven major AI platforms with industry-relevant prompts and tracking your citation frequency, sentiment, and share of voice over time. The scan takes minutes instead of the hours required for manual testing.

Compare your citation performance across platforms. You might discover that Perplexity cites you consistently but ChatGPT does not, which points to a brand authority gap in ChatGPT's training data. Or you might find that all platforms cite a competitor for topics where your content is stronger, which suggests a technical or structural issue preventing AI engines from finding your content.

Interpreting your results and next steps#

Your audit produces scores across four areas: crawler access, structured data, content readiness, and citation performance. Prioritize fixes based on impact.

Blocked crawlers are the highest priority. No other optimization matters if AI engines cannot reach your content. Fix robots.txt and infrastructure-level blocking first.

Missing structured data is second. Adding Organization and Article schema takes hours and immediately improves how AI engines parse your content. FAQ and HowTo schema can follow in the next sprint.

Content gaps are third. Rewriting your top pages with citation-ready structure takes more effort but drives the biggest improvement in citation frequency. Focus on the pages that map to your highest-value queries first.

Run this audit quarterly. AI platforms update their models frequently, competitors publish new content, and your own site evolves. What scores well today may need adjustment in three months. BrandCited's automated auditing handles the quarterly cadence with alerts when scores change.