AI Visibility Audit: What the Data Shows and Which Fixes Move the Needle
Most AI visibility audits reveal the same five patterns. A Zenodo study of 2,729 businesses found 69% are invisible to AI engines. Here is the diagnostic map and the fix sequence that moves numbers fastest.
I wrote the full 7-step AI visibility audit framework earlier this year. This piece covers what you actually find when you run one — and which fixes move your numbers fastest.
The short version: most brands are invisible to AI engines, the gap between being mentioned and being cited is where pipeline leaks, and the fix sequence matters more than the fix list.
69% of Businesses Are Invisible to AI Engines
A peer-reviewed dataset published on Zenodo analyzed 2,729 businesses across 14 verticals and five AI engines — ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews — using 266,844 prompt-level observations. The headline finding: 69% of businesses are completely invisible. Only 2% appeared across all five models. The average business showed up in 0.6 out of 5 engines.
That is not a ranking problem. That is an existence problem.
Walker Sands' H1 2026 B2B AI Search Visibility Benchmark, reported by Search Engine Land, analyzed 45 million queries across 828 enterprise B2B companies. The median brand earns citations in just 3% of the AI Overviews that appear on its ranking keywords. Top quartile reaches 4.5%. Bottom quartile sits at 1.7%.
Your SEO dashboard might show thousands of ranking keywords. Walker Sands found AI Overviews appear on roughly 50% of those queries. If you are cited in 3% of them, your AI visibility is functionally zero for the queries that increasingly drive purchase research.
The Five Findings That Repeat in Every Audit
Five findings show up consistently, usually in the same order of severity.
Finding 1: AI crawlers are blocked. The highest-leverage, lowest-cost fix is also the most overlooked. Enterprise sites routinely block GPTBot, ClaudeBot, PerplexityBot, and OAI-SearchBot through robots.txt rules, CDN rate limits, or JavaScript-only rendering. If crawlers cannot access your pages, nothing else in this audit matters. Check robots.txt, server logs, and CDN/WAF rules before optimizing a single word of content.
Finding 2: AI engines cannot resolve your brand entity. Ask ChatGPT, Perplexity, and Gemini "What is [your company]?" and compare the answers. Most brands discover inconsistent descriptions, confused product attributions, or outright conflation with competitors. Entity confusion comes from fragmented signals — your homepage says one thing, your LinkedIn says another, and Wikipedia says something different.
Finding 3: Your content is not extractable. AI engines extract structured claims, not narrative prose. Semrush's 2026 AI Visibility Index — analyzing 126 million U.S. AI search prompts across 22 industries — found that on Gemini, the overlap between mentioned brands and cited domains can be as low as 30%. Only 36 brands maintained top-100 visibility across all four AI platforms every month. You get mentioned without being cited, or cited without being named. The fix is structural: answer-first formatting, standalone declarative claims, comparison tables, and definition blocks that engines can extract and attribute cleanly.
Finding 4: Earned media is thinner than you assume. Ahrefs' study of 75,000 brands found that off-site brand signals predict AI visibility 2-3x more strongly than backlinks. YouTube mentions showed the strongest single correlation at 0.737, followed by branded web mentions at 0.664. Backlinks — the metric most PR and SEO teams optimize — correlated at just 0.218. AuthorityTech's research on AI search citation factors found that earned media accounts for 82-89% of all AI citations. Your owned content is table stakes. Your earned media footprint is what AI engines actually weigh.
Finding 5: You have no measurement baseline. Most brands discover during their first audit that they have no historical data for AI citation rate, share of voice across engines, or accuracy tracking. Without a baseline, every future measurement is a data point with no context.
Which Fixes Move the Needle First
The fix sequence matters more than the fix list. Here is the order I use, ranked by leverage and implementation speed:
| Priority | Fix | Time to implement | Why this order |
|---|---|---|---|
| 1 | Crawler access — unblock GPTBot, ClaudeBot, PerplexityBot in robots.txt and CDN/WAF rules | One afternoon | Binary gate: blocked means invisible regardless of content quality |
| 2 | Entity resolution — consistent descriptions across homepage, LinkedIn, Wikipedia, Google Business Profile, Organization schema | 1-2 weeks | AI engines need to resolve your brand before they can cite it |
| 3 | Content extractability — answer-first openings, standalone claim blocks, comparison tables, FAQ sections on top 10 pages | 2-4 weeks | Fixes the ghost citation problem — gets your brand named, not just linked |
| 4 | Earned media authority — sustained editorial placements in publications AI engines cite | 90+ days | TNW reported editorial placements cited 2.3x more than wire coverage, and content published within 12 months cited 2.9x more than older coverage |
| 5 | Measurement baseline — monthly tracking across ChatGPT, Perplexity, and Google AI Overviews using a paid monitor plus Bing Webmaster Tools and GA4 | 1 week setup | Bing Webmaster Tools now reports AI performance data — critical because ChatGPT and Copilot both run on the Bing index |
Each step unlocks the next. There is no point optimizing content extractability if crawlers cannot reach your pages. There is no point building earned media if your entity signal is so fragmented that the coverage reinforces the wrong brand identity. Work the stack from the bottom up.
The Source Architecture Gap
The pattern these findings expose is not a content problem or a technical SEO problem. It is a source architecture problem — the structured relationship between owned content, earned media, entity signals, and technical accessibility that determines whether AI engines can resolve, trust, and cite your brand.
The brands winning in AI discovery are not the ones with the most content. They are the ones whose source architecture is machine-legible end to end.
FAQ
What is a normal AI visibility audit score for a B2B brand?
Low. Walker Sands found the median enterprise B2B brand is cited in just 3% of AI Overviews on its ranking keywords. A Zenodo-published study of 2,729 businesses found 69% are completely invisible to AI engines. If your first audit shows single-digit citation rates, you are in the majority — and there is room to improve with the fix sequence above.
How long does it take to improve AI visibility after an audit?
Crawler access and entity resolution fixes can show results within weeks. Content restructuring takes two to four weeks per page batch. Earned media — the highest-impact lever — requires at minimum 90 days of sustained placement activity before AI citation patterns shift measurably.
Is AI visibility the same as SEO visibility?
No. The Walker Sands benchmark found that AI Overviews appear on 50% of B2B search queries, yet brands ranking for those keywords are cited in only 3% of the AI answers. Ahrefs data confirms that traditional ranking factors like backlinks and domain authority show weak correlations with AI visibility, while brand mentions and YouTube presence are 2-3x stronger predictors.