Check Engine / Checks / Reference

Do successful pages show a not-found message?

It finds short successful pages containing one of the stored missing-page messages.

What Check Engine measures

The test reads successful pages and skips any with more than 250 words. It searches the title plus the first 1500 characters of text for the literal not-found markers. Matching pages are separated according to whether another crawled page links to them. The presence of an observed source link raises severity from medium to high. Matching words alone leave the cause of the response unknown. For example, a page above 250 words is left out even if it discusses a not-found message.

Crawl records supply successful responses, short page text, titles and incoming links. This test does not use the separate site-level probe of a nonexistent address.

How to read the result

The finding reports a conflict between a successful response and an error-like body. A short article discussing an error message can need review before being treated as a missing page, because the check is a phrase test with a length guard.

Why it matters

Missing-page messages and successful statuses tell readers' software different things. An internal link to that response also gives you a source to repair. Review the message and the linking page together.

How to fix it

Open the affected URL from its source link. Restore the intended content if it should exist, or return a proper missing-page status if it does not. Fix internal links that still lead there. If the page is a genuine explanation of an error, verify its content and response before changing it solely because a phrase matched.

How to fix soft 404s

What we found in field audits

Problems in Crawlability & indexing were found on 12 of the 14 sites we audited. These are area totals. Read the field-audit method.

Method note

backend/seo_engine/checks/indexability.py, soft_404_pages, revision 0bb8e8d3.

Common questions

How long can a page be and still enter the error-body check?

Pages above 250 words are skipped. Short successful responses can be searched for stored not-found wording.

Does an internal link to an error-like page raise severity?

Yes. An observed source link moves the result from medium to high. The text still needs review before calling the page missing.

How much text is searched for not-found markers on successful pages?

It searches the title and first 1500 characters. A matching phrase leaves the cause uncertain until the response is inspected.