An AI visibility audit measures whether AI search engines can find your business, understand it, trust it, and quote it when a buyer asks for a company like yours. It returns a single score from 0 to 100 and a prioritized list of fixes. Ours takes 60 seconds and scores six factors. This is what each one checks and why it counts, so you know exactly what the number means before you act on it.
1. Crawler access
The first check is the most literal: can AI crawlers actually reach your pages. It inspects robots.txt for the AI user-agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended and others) and confirms the server returns your content to them rather than blocking or redirecting. This is a gate, not a nicety. If the crawler cannot fetch your page, nothing else you do matters, which is why it is scored first. Perioskoup, a live client of ours, reached 95 out of 100 on AI-crawler access, and that access is what everything else is built on.
2. Content visibility (rendering)
The second check reads your raw HTML the way a non-JavaScript crawler does, and asks whether the real content is actually there. Modern single-page apps built in React, Vue, or Webflow often paint their content with JavaScript after load. A human sees a full page; the crawler sees an almost-empty shell. This factor catches that gap directly. The fix is server-side rendering or static prerendering, so the first HTML response already contains your headings, copy, and structured data.
3. Structured data
The third check looks for machine-readable facts: Organization, LocalBusiness, Service, FAQPage, and Review schema, expressed as clean JSON-LD. Structured data is how you tell an AI parser what you are, where you are, and what you sell without making it infer from prose. The audit checks that the schema is present, valid, and flat rather than buried in a nested graph wrapper that some parsers cannot traverse. Getting this right is one of the fastest ways to move the score.
4. Entity clarity
The fourth check is subtler: is it unambiguous what your business is. An AI engine only recommends what it can describe confidently. This factor looks at whether your name, category, location, and offering are stated consistently and plainly across the page and the structured data, so the model can form a clean mental model of you as an entity. This is the layer where Perioskoup already wins: ask an AI engine what Perioskoup is and it answers correctly and in detail. Entity recognition is built first, because the model has to know what you are before it will name you in a category answer.
5. Answer-first content
The fifth check reads your page the way an assistant does when it needs a sentence to quote. It looks for direct, self-contained answers to real buying questions, ideally with the question as a heading and the answer stated immediately underneath. Pages that bury the answer in marketing language give the model nothing to lift. This is also where the research is loudest: the GEO study (Aggarwal et al.), the first peer-reviewed paper on Generative Engine Optimization, found that answer-first structure with citations, quotations, and statistics can raise a source's visibility in AI answers by up to 40 percent.
6. Technical hygiene
The sixth check covers the fundamentals that quietly drag a score down: HTTPS, a responsive mobile viewport, page speed and Core Web Vitals, clean titles and meta descriptions, and canonical tags. Performance is itself an AI visibility signal, because a slow or broken page is a weaker source to cite, and because the same speed the audit rewards is the speed your human visitors feel. This is the least glamorous factor and one of the most reliable to fix.
Why the order matters
The six factors are not weighted equally, and they are not independent. They stack. Crawler access and rendering are gates: fail them and your structured data, entity clarity, and answer-first prose are never even read, so the score collapses regardless of how good the rest is. Structured data and entity clarity are the middle layer, where the model decides what you are. Answer-first content and technical hygiene are the polish that decides whether you get quoted and how fast you load. A prioritized fix list follows the same logic, which is why the audit tells you not just what is broken but which broken thing to fix first for the biggest gain.
What a score actually tells you
Add the six together and you get a number from 0 to 100. A high score means an AI engine can reach you, read you, trust your facts, know what you are, quote your answers, and load you fast. A low score tells you which of those is broken and, more usefully, which cheap fix moves the number most. We publish our own: our site currently scores 72 out of 100 with a public fix log, and Perioskoup reached 61 against a frozen baseline with that 95 on crawler access, its entity layer already solid and category citations the next thing to earn. Run the free 60-second audit, read your six factors, and start with the gate that is costing you the most.