Deep AI SEO audit
Every page, with the URLs
The free audit reads ten of your pages and tells you what is wrong. This one crawls the whole site and tells you which pages, in what order to fix them, and how you compare with three competitors you name.
What changes when we read all of it
The checks are the same checks. What changes is how much of your site they run over, and whether the report can name the pages. Most people should run the free audit first: if it finds something blocking, fix that before paying anyone for a longer list.
| Measurement | Free audit | Deep audit · $9.99 |
|---|---|---|
| Pages read | 10, sampled from your sitemap | Up to 1,000, following your internal links as well as the sitemap |
| Findings | Every category, every blocking problem, 2 or 3 worked examples | Every finding, with every URL it affects |
| Crawler access | Which AI crawlers your robots.txt turns away | The full ledger: every named crawler, the group it belongs to, your rule for it, and what your server actually did when we knocked |
| Competitors | None | 3 domains you name, run through the same checks, readings side by side |
| Fix order | By severity | By severity and by how many pages it affects, so the top of the list is the change that moves the most pages |
| After you fix something | Run it again whenever you like | 5 re-runs in 30 days, each with a list of what changed since last time |
Included at no extra cost on AI Visibility Starter and above. See the plans.
Seven things a sample cannot tell you
These need the whole site, and most of them need the map of what links to what. They are the reason the deep audit is a different measurement rather than a bigger sample of the same one.
- Does your site start refusing us partway through?
- A site can serve ten pages happily and start blocking at two hundred. Rate limits and bot protection often key on volume, not on who is asking, so this failure is invisible to every tool that only samples. We record where the refusals begin, and a crawl that gets shut down becomes the finding rather than an error message.
- What lives on your other subdomains?
- Crawlers treat docs.yoursite.com and help.yoursite.com as separate sites with their own robots.txt. Your main robots.txt says nothing about them, and your documentation is usually the part an assistant would want to quote. We list every separate host we find in your links and sitemap, and read the rules each one publishes.
- Which pages does nothing link to?
- Crawlers follow links. A page that exists only in your sitemap is reached late or never, however good it is. We build the link graph of your whole site while we crawl, which is the only way to tell a well-connected page from an orphan.
- How many addresses does the same page have?
- Duplicate titles and headings across pages, plus www and non-www, http and https, and trailing-slash variants that all return a page. Each split spreads the same content over several addresses, none of which is clearly the one to quote.
- Are your pages telling crawlers to look elsewhere?
- A canonical tag pointing at a different URL tells a crawler this page is not the source. On one page that is housekeeping. Across a template it quietly removes a whole section of your site, and you only see it by checking every page.
- How much of your crawl budget goes to pages that do not exist?
- Pages that return a normal 200 response while saying "not found" are crawl waste, and they are common on large sites. Measured on production server logs, ChatGPT's crawler spends 34.8% of its fetches on missing URLs against Googlebot's 8.2%, so waste costs you more here than it does in search.
- How big is the part of your site you have blocked?
- "You block /resources/" only means something once you know 240 pages live there. We match your robots.txt rules against the real inventory of URLs we found, so a block is reported with the number of pages behind it.
What we promise, and what we do not
- We check we can read your site before you pay.
- Homepage, robots.txt and sitemap, in about five seconds, free. If your site turns crawlers away or needs JavaScript to show anything, no payment page appears at all. You get the free audit and the reason instead.
- If we read less of your site than we hoped, the report says so first.
- A crawl can be cut short by throttling or by pages that refuse us. When that happens the coverage sits above the score, in the report and in the first line of the email, rather than in a footnote under it.
- Every line is a measurement, not a guess.
- No language model is used anywhere in the audit. Each finding carries how firm it is: measured on your site, backed by published measurement, or a convention nobody has confirmed anyone reads. Only the first two can move your score.
- Being readable is not the same as being recommended.
- This audit measures your own site, and most of what an assistant quotes is somebody else writing about you. Fixing your site is necessary and nowhere near sufficient, and no audit of your own pages can tell you whether assistants bring you up. That is the question the subscription measures.
Questions before you buy
What do I get that the free audit does not give me?
Coverage and page-level attribution. The free audit reads ten pages and tells you what is wrong; the deep audit crawls up to 1,000 and tells you which pages, in what order to fix them, and how you compare with three competitors you name. Seven of its checks, including orphan pages and duplicate addresses, are impossible on a ten-page sample because they need the whole link graph.
What happens if you cannot read my site?
You never reach a payment page. We check your homepage, robots.txt and sitemap before showing a price, and a site we cannot crawl is turned away with the reason. Being unreadable to crawlers is the most valuable thing we could tell you anyway, and the free audit explains it in full.
How long does it take?
A few minutes for most sites, and up to twenty for the largest. You do not have to sit and watch it: the page updates by itself, and we email you the link the moment it is finished.
Do I need an account?
No. You pay, you get a link, the link is yours. There is no password to set and nothing to cancel, because there is no subscription.
How many times can I run it?
Six times in 30 days: the first crawl and 5 re-runs. Each re-run shows what changed since the last one, which is what makes it useful after you fix something. Re-runs are capped rather than unlimited because a crawl costs us worker time even though it costs no AI usage at all.
Is it included in a subscription?
Yes. Every AI Visibility plan from Starter upwards includes one deep audit a calendar month, with the same re-runs. The audit answers whether assistants can read you, and a subscription answers whether they bring you up, so the first is a prerequisite for the second rather than a separate purchase.
Will you tell me what to write?
No, and that is deliberate. Every line in the report is something we measured on your site, with the URL it came from. No language model is used anywhere in the audit, which is exactly why a re-run costs you nothing and why every finding can be checked against your own pages.
Do you crawl politely?
Yes. We obey your robots.txt for our own crawler, hold to three requests at a time, slow down when your server asks us to, and stop entirely if it keeps refusing. Our crawler identifies itself as babel42bot, so you can see exactly what we did in your logs.