A soft 404 is a page that looks missing to a person but returns a successful response to a crawler. Fix it by returning a real 404 or 410, or by serving a genuinely relevant replacement with a permanent redirect. Audit the body, title, status, and links together. Empty templates, search results with no match, deleted posts, and error pages that return 200 are common causes. Remove dead URLs from the sitemap and add a useful navigation path for people who arrive there.
The short answer
A soft 404 is a page that looks missing to a person but returns a successful response to a crawler. Fix it by returning a real 404 or 410, or by serving a genuinely relevant replacement with a permanent redirect. The useful version of this work is answer-first: state what the reader should do, explain the evidence that supports it, and show the limit before the reader mistakes a first layer for a guarantee.
For a live SaaS, the important question is rarely whether one URL returns 200. It is whether the route, browser, provider, and data side effect agree with the promise the customer was given. That is why a durable audit keeps the scope, expected behavior, observation, and next action together.
How the workflow works
Audit the body, title, status, and links together. Empty templates, search results with no match, deleted posts, and error pages that return 200 are common causes. Remove dead URLs from the sitemap and add a useful navigation path for people who arrive there. Begin with a representative surface and only widen the scan when the first result is understood. This reduces false confidence and makes the output easier to hand to an engineer, founder, client, or reviewer.
A passing result should be dated and reproducible. A failing result should explain impact, identify the broken boundary, and preserve enough safe detail for a second person to verify the diagnosis. If a question requires credentials, source access, or adversarial judgment, say so and route it to the deeper review it needs.
Practical checklist
- Identify pages with missing or empty primary content.
- Return 404 or 410 when there is no replacement.
- Use a 301 only when the replacement satisfies the same intent.
- Remove dead routes from internal links and sitemaps.
- Verify the final status and visible message from the live host.
Work through the list in customer-impact order. Fixing a low-risk metadata warning while a payment webhook silently drops fulfillment events creates a prettier dashboard, not a safer release. The owner should be able to point to the exact result that moved from failed to verified.
Mistakes to avoid
- Redirecting every missing page to the homepage.
- Returning 200 for a framework not-found component.
- Keeping retired URLs in the sitemap for months.
- Replacing a useful page with a generic error that offers no next action.
Do not use word count, schema volume, or check count as a substitute for usefulness. The page, scan, or report should help a real person make a decision. Preserve the limitations, cite external standards when a claim depends on them, and update the visible date when the workflow changes.
How to verify the next release
Run the same scope after the change against the canonical production surface. Compare the before and after observations, inspect the route or provider that changed, and keep the follow-up monitor or release gate that will catch a regression. That is how a one-time article checklist becomes an operating habit.