Broken links and typos — what a visitor notices first
We re-fetch every link on the site and read the text of every page. You get a list to work through, not a report to interpret. No law requires this — but a dead link in the menu makes a visitor doubt everything else on the page.
This is not a compliance module
Nobody is fined for a broken link. The quality module never affects your site's compliance status and its findings never enter a compliance report — they sit on their own, in the action plan, as what they are: craft. The module is included in every plan and starts switched off, so you turn it on when you want it.
What we check
Two halves you can switch on independently. The link check is deterministic — it fetches the link and reads the status code. The spell check is the only part that uses AI.
Broken links
Only 404 and 410 count as broken: the server states outright that the page does not exist. A site that is temporarily down never becomes an accusation. We show the link text, the address, and how many pages the link sits on — a link in the menu on 150 pages is fixed in one place.
Links to review
Everything that answered something else: 403, server errors, timeouts, or a page that answers OK while saying “not found” itself. We report what we saw and let you judge. Hosts that block all automated checks — common on large sites — are collapsed into one line instead of a hundred.
Spelling in the page text
An AI reads the visible text and suggests suspected misspellings, with the sentence they appear in. Header and footer are checked once and apply to every page — otherwise one typo in the menu would show up on every page in the list.
Dismiss with a reason
A link you know is fine gets dismissed with a note. The decision is saved with name and date, survives every rescan, and can always be undone. If the link changes — it suddenly answers 404 — your decision is flagged for another look, but never undone automatically.
How it works
- 1We use the pages we already fetched
The quality module reads the same pages as the rest of the scan, so your server does not get a second crawl for this.
- 2Every link is re-fetched
At most two concurrent requests per host, with a pause between them — against your own server and against everyone you link to. Internal links we already fetched are checked without a single extra request.
- 3The text is proofread
The page's visible text is read by a language model that sees the sentence each word sits in. Unchanged pages are not re-read on the next scan.
- 4You get three items, not three hundred
Broken links, links to review, and spelling — three rows in the action plan. The full list opens when you want to work through it.
What the module does not do
- We only call 404 and 410 broken. Everything else is left to you, because a site that is down right now is not a broken link.
- We always say how many links out of the total we actually managed to check, and which hosts refused us. A count without that number would mislead.
- Spelling is judged by an AI reading the text in context. It can be wrong — especially about place names, personal names and specialist terms. Every word is a suggestion to check, never a verdict.
What you get
- A list with link text, address and page count — sorted so what sits on the most pages comes first.
- A per-link dismissal, with a reason, that holds over time and can be undone.
- An “I have fixed it” button that re-fetches the links and re-reads the pages, with no AI and no guesswork.
See what is broken on your site
Run a free scan. You see the result straight away, and you switch the quality module on when you want it.
Run a free scan