The eight categories, and what each one is looking for
| Category | What is checked | The failure it catches |
|---|---|---|
| Crawlability and indexation | robots.txt, sitemaps and their accuracy, canonical logic, indexed-to-submitted ratio, redirect chains, orphan pages | Pages that exist but were never going to be found |
| Rendering | Whether primary content is present in the served HTML or depends on client-side JavaScript Google Search Central, JavaScript SEO basics | A marketing site that looks complete to a browser and empty to a crawler |
| Structured data and entity | Whether the schema graph parses, whether it matches visible content Google Search Central, intro to structured data markup, and whether the organisation is described consistently | Invalid or contradictory markup, which is silently ignored rather than reported |
| Core Web Vitals | Field data where available, layout shift sources, render-blocking resources Google Search Central, understanding Core Web Vitals | A fast lab score hiding a slow real-world experience |
| Mobile and accessibility | Viewport behaviour from 375 to 1920, tap targets, contrast ratios, heading order | A site that technically responds but is not usable on a phone |
| Security and hygiene | HTTPS consistency, mixed content, header basics, exposed staging or admin surfaces | Trust signals that quietly fail in the background |
| AI-crawler readiness | Whether assistant crawlers are permitted, whether content is reachable without JavaScript, and whether claims carry sources OpenAI, bots and crawler documentation | A site invisible to the surface a growing share of buyers start from |
| Content and conversion | Depth against the query, ideal-customer clarity, the conversion path, and the quality of the writing | A technically perfect site with nothing worth citing on it |
The AI-crawler checks, in more detail
This category is the one most audits still skip, and it is mostly a robots.txt and rendering question rather than anything exotic. Assistant operators document their user agents publicly: OpenAI documents OAI-SearchBot for surfacing sites in ChatGPT search, GPTBot for model training and ChatGPT-User for user-triggered fetches OpenAI, bots and crawler documentation. Anthropic documents ClaudeBot, Claude-User and Claude-SearchBot with distinct purposes Anthropic Support, does Anthropic crawl data from the web. Perplexity documents PerplexityBot for surfacing and linking sites in its results, and Perplexity-User for user-initiated visits Perplexity, PerplexityBot documentation.
The check is whether a site blocks the ones it does not mean to block. A blanket disallow added years ago, or a security product blocking unfamiliar agents, routinely removes a site from assistant answers while leaving Google Search untouched. Google’s own control token works the other way round: Google-Extended governs whether crawled content may be used for training and grounding in Gemini apps, and explicitly does not affect inclusion or ranking in Google Search Google Search Central, Google's common crawlers.
What comes back, and in what order
The output is a written read, not a tool export. It names what is wrong, puts it in the order it should be fixed, and separates the things that need an engineer from the things that need an editor. Where a check passes, it is not listed; a 52-item list of green ticks is a sales document rather than an audit.
It includes the recommendation not to hire anyone where that is the right answer. The most common version of that is a site whose problem is a product-market or demand problem rather than a search problem, where content investment would be the expensive way to discover something a week of customer calls would show.
The honest limits of a 48-hour audit
It is a read of public surfaces plus whatever the applicant shares. Without Search Console access it cannot see query-level performance, and without analytics it cannot see conversion behaviour. It will say which of those it needed and did not have.
It is also a point-in-time read. Core Web Vitals field data lags, index coverage changes, and an assistant answer that cites you today may not tomorrow. The audit is the start of a plan, not a certificate.
Frequently asked questions
Is the 52-check audit really free?
Yes, with no obligation and no call required. Four are taken a month because each one is run personally.
Do I need to give access to Search Console?
No, but the audit is better with it. Without it, query-level performance cannot be read and the report says so rather than guessing.
How is this different from a crawl tool report?
A crawl tool lists everything it can detect. This prioritises what matters for your specific situation and includes the non-technical layer a crawler cannot assess, such as whether the content is worth citing.
What if the audit finds nothing serious?
Then it says so, and the recommendation will be that you do not need an agency right now. That happens, and it is a more useful answer than a manufactured problem.
Does the audit check whether AI assistants can read my site?
Yes, that is one of the eight categories: whether assistant crawlers are permitted, whether content is reachable without JavaScript, and whether claims carry sources that can be cited.