Technical audit
195 checks, one score, and a ranked list of what to fix
A crawler that renders when it must, a score that shows its arithmetic, and fixes ranked by what they return.
- checks, in six categories
- 195checks, in six categories
- point score, with its arithmetic shown
- 100point score, with its arithmetic shown
- every rate, so crawl size does not decide the number
- Per pageevery rate, so crawl size does not decide the number
- catalogue version, stamped on every crawl
- v6catalogue version, stamped on every crawl
195 checks in five categories
Every one of them fires on a page or on the site as a whole, and every finding names the pages it came from.
Indexability & crawl
90Status codes, redirect chains, robots and canonical rules, sitemap agreement - the checks that decide whether a page can rank at all.
Metadata
26Titles and descriptions: missing, duplicated, too long, too short, or written for a page that has since changed.
Content
20Thin pages, duplicated bodies, heading order, readability, and the pages nothing links to.
Links & media
22Broken internal and external links, nofollow where it costs you, images with no alt text or dimensions.
Performance
16Time to first byte, render-blocking assets, Core Web Vitals, and the weight a visitor actually pays for.
It renders the pages that need rendering
A crawler that only reads HTML measures the wrong thing on a site built in JavaScript - and one that renders every page is slow and expensive for no gain on a site that is not.
So the decision is made per page, by script weight, and the report says how many were rendered and what rendering found. On a server-rendered site the honest answer is none, and it says that rather than pretending to have worked.
rendered where the HTML was thin
nothing to gain, nothing spent
And audits that sample for accessibility
axe-core against WCAG 2.0, 2.1 and 2.2 at levels A and AA, run in the same rendered browser - because a contrast ratio or a missing label is only true of the page a person actually sees.
A sample, and it says so. Every audit records how many pages it audited out of how many were eligible. This is automated conformance testing, which finds a real and well-defined subset of problems - it is not an accessibility audit and does not certify compliance with anything.
- pages sampled and audited
- 25 of 25pages sampled and audited
- levels A and AA
- WCAG 2.2levels A and AA
The sample is chosen for template diversity rather than by traffic - one page per URL shape finds more distinct problems than twenty copies of the busiest one.
Ranked by return, not by severity alone
Sorting by severity puts ten critical findings on one obscure page above a warning that affects every page you have.
Findings are ranked by impact multiplied by how easy the fix is, scaled by how many pages carry it - so the top five are the top five for a reason you can check.
What the report gives you
Grouped by cause
One broken target linked from forty pages is one finding with forty sources, not forty findings.
Marked as you go
Every finding carries a status - to do, done, ignored - so the next crawl shows what is new rather than what you already decided about.
Compared to last time
Two crawls side by side: what appeared, what went away, and which pages changed underneath.
Your own rules
Add a check the catalogue does not have. Custom findings stay out of the health score on purpose - your rule is not our scoring model.
Segments
Score a slice of the site on its own - one path prefix, one template - without paying for a second crawl.
Shareable
A report link a client can open without an account, and a PDF for the ones who want it in an email.
An internal linking tool, over the crawl you already ran
The crawl stores the whole link graph, so the internal linking analysis needs no second pass and no vendor: internal PageRank is computed over it with the standard 0.85 damping and normalised so the average page sits at 1.0. That gives every page a number you can compare to its neighbours rather than to nothing.
Then it suggests the links worth adding. A target qualifies when it ranks in Google’s striking-distance band - position 11 to 20, where the query already matches - and has two or fewer internal links pointing at it. The source is a page that already holds authority and already takes impressions for overlapping queries, which is stronger evidence of relevance than matching words in a title.
Orphans come out of the same graph: a page nothing links to is flagged, and a page nothing links to that is still getting traffic is flagged louder.
Try six of these checks on one URL, free
No account and nothing stored. Each one is the same check the audit runs, with the same thresholds - pointed at a single page instead of all of them.
- Robots.txt testerIs this URL allowed, and by which rule
- Meta description checkerLength, and the width Google actually cuts at
- Title tag checkerDoes the title fit, in pixels
- Hreflang checkerValid codes, self-reference, x-default
- XML sitemap checkerValid XML, URL count, protocol limits
- Canonical tag checkerSelf-referencing, or pointing elsewhere
Looking for a Screaming Frog alternative?
Sometimes the honest answer is no. A desktop crawler and a hosted audit are different tools, and which one you want depends on whether the crawl is a one-off investigation or a record somebody else has to read.
Stay on the desktop crawler when
- You want to point a crawler at anything, right now, without an account or a credit balance.
- You need custom extraction - pull an arbitrary element off every page with your own selector.
- You are doing one-off forensic work: a migration diff, a redirect chain hunt, a log file question.
- You want the raw export and you would rather do the analysis yourself in a spreadsheet.
Come here when the crawl has to be a record
- It runs on a schedule and keeps the history, so two crawls can be compared instead of remembered.
- The 195 checks produce a score and a ranked list of fixes, not an export you still have to triage.
- Search Console and Analytics are joined per page, so a finding can be read against what that page actually earns.
- A client can open the report without installing anything, and a colleague can pick up where you left off.
- Findings carry a status - to do, done, ignored - so the next crawl shows what is new.
Plenty of teams run both, and that is a reasonable answer too - the desktop crawler for the investigation, this for the record of what was found and whether it got fixed.
It respects the site it is crawling
The crawler obeys robots.txt and Crawl-delay, identifies itself as SeodarBot, and holds a delay between requests.
It is the same crawler on your site and on a competitor’s, which is why the two can be scored against each other at all.
Questions
- Penalties per category, weighted by severity and capped so no single problem drowns out the rest, expressed as a rate per page. The report shows the arithmetic - you can see which category cost you what.
- So a 5,000-page site and a 200-page site can be read against each other. A count would tell you mostly how large the crawl was.
- When the page needs it. Pages are judged by how script-heavy they are rather than by a list of framework markers, because those go stale - one such list was three years old and matched nothing on a modern site.
- Every crawl records which catalogue version scored it. Two scores from different catalogues are not compared silently - the report says so, and the comparison view refuses to subtract them.
Start free
One site and 100 credits - up to 200 pages a scan, about 5 scans, and the whole report. No card. Paid plans add more sites, bigger crawls and the AI rewrites.