Skip to content

Site Audit

Crawl your own site for the technical and AI-visibility issues that keep engines from citing you.

On this page

What it does

Site Audit crawls your workspace's own website and scores it out of 100 across ten categories, then ranks the fixes by how much each one should move AI visibility. It's a standalone crawl: it doesn't use Search Console, GA4 or any other connection, and it shares no data with the Pages tab.

It only ever crawls the website on the workspace — you can't point it at an arbitrary domain.

Running one

Open Site Audit and click Run audit. A run takes about two minutes; small sites finish in seconds. One audit runs per workspace at a time, and asking for a second while one is in flight is refused rather than queued.

There's no schedule and no recurring audit — it's on demand. Nova can also start one for you.

What gets crawled

The crawler starts at your homepage, reads robots.txt for any sitemaps it declares, falls back to /sitemap.xml, and then follows same-origin links up to six levels deep. It respects robots.txt disallow rules and crawl delay, and identifies itself as MentionovaSiteAudit — worth allowlisting if your WAF blocks unfamiliar agents.

A sitemap isn't required, but it improves coverage a lot on sites with weak internal linking.

The report

A health score out of 100 with a one-paragraph summary, then counts of errors, warnings and notices and the number of pages crawled. Below that: the categories that scored under 100, a Prioritized fixes list of four to eight items ordered by AI-visibility impact, and every finding grouped by category.

Each finding expands to the sentence explaining it, the fix to make, and the affected URLs — copyable in bulk. Findings that apply to the whole site are labelled site-wide rather than counted per page.

The ten categories

Crawlability & indexing · On-page technical · Structured data · Performance & Core Web Vitals · Mobile · Security · Internal linking · International · Duplicate content · AI engine optimization.

The last one is the reason to run this rather than a generic SEO crawler. It checks for an llms.txt, whether your pages carry question-style headings an engine can lift, whether you publish Organization or WebSite entity schema with sameAs, and whether FAQ schema exists anywhere on the site.

Core Web Vitals come from Google's field data, which only exists for sites with enough Chrome traffic. When it isn't available those checks are skipped, and that's normal rather than a fault.

How the score works

The score is deterministic — the same crawl always produces the same number. Every finding subtracts from 100 by severity (an error weighs most, a notice least), scaled by how much of the site it affects: an issue on one page of five hundred costs a fraction of what the same issue costs when it's everywhere.

Category scores are amplified on purpose so a category with real problems reads clearly low. They aren't comparable to the overall health number.

Limits worth knowing

  • Pages per crawl — 25 on a trial, 100 on Starter and Launch, 500 on Scale and Growth, 2,500 on either Enterprise. Hitting the cap isn't an error: the audit finishes and says so, and issues on deeper pages stay unknown.
  • Audits per month — 1, 2, 5 or 10 per workspace by plan, per calendar month. Failed audits don't count against it.
  • It isn't charged in tasks. The page and monthly caps are the whole limit.
  • JavaScript-rendered sites are the big caveat: the crawler reads raw HTML and doesn't run a browser, so a client-rendered site can report missing titles, H1s and schema that a visitor would see. Treat a suspiciously bad score on an SPA as a rendering artefact first.
  • Nothing behind a login — no cookies or credentials are sent, so a staging site behind basic auth returns 4xx and tanks the score.
  • Apex vs www — if the workspace URL is the apex but your site serves from www, the crawler treats every internal link as off-site and stops after a page or two. An audit that returns one page is almost always this.

Sharing a report

Share report mints a public read-only link to the audit, revocable at any time — the same shape as a shared insights report, and it carries your branding if your plan includes white-label.

Who can run it

Every plan includes Site Audit, including the trial — what changes is depth and frequency. The exception is a read-only organization (an expired trial, a cancelled plan, or demo mode), where audits are blocked along with every other write.

Site Audit has no checkbox in the per-member tab controls, so restricting a member to specific tabs hides it from them.