First, decide whether the page should exist
Open the page indexing report in Search Console (GSCGoogle's search performance and indexing tool.) and export the URLs listed under the soft error row. Before fixing anything, sort them into two piles. The first pile is pages that genuinely have nothing to offer: deleted items, expired listings, empty searches, error templates. For these, Google's reading is correct and only the response code is wrong. The second pile is pages that should rank: live categories, in-stock products, articles. For these, something makes a good page look hollow.
For the second pile, run URL inspection with the live test and look at the screenshot and the rendered HTML. Ask what a visitor would see if they arrived from search. A page that shows a heading, a site-wide header and footer and almost nothing else may be complete for a logged-in user but empty for a crawler. Also check the response chain: a URL that redirects somewhere unrelated can be reported against the original address. This split between "should go" and "should stay" decides every fix below.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Pages that say "not found" or "no longer available" are flagged | Request the URL and read the status code it returns | Error or not-found pages returning a success status | Return a real not-found or gone status from the error template |
| Category, tag or on-site search pages with no items are flagged | Visit the page; count listings; check how it is linked | Empty category or search results pages | Fill, merge or remove empty listings; keep internal search out of the index |
| Product pages marked unavailable are flagged | Compare the flagged products with stock and lifecycle data | Out-of-stock or discontinued product pages with little content | Keep useful content and alternatives, or retire the page properly |
| Removed URLs are flagged although they now redirect | Follow the redirect and compare the destination with the original | Redirects of removed pages to an irrelevant page such as the home page | Redirect to the closest equivalent, or return a gone status |
| Pages that look complete in a browser are flagged | Live test screenshot and rendered HTML | Content that fails to render, leaving a near-empty page | Serve the main content in HTML or fix the failing script or request |
Fix, verify and prevent each cause
Error pages answering with success
Many content systems show a friendly "page not found" template but send it with a success code. Google's explanation of how search works notes that it reads the status code the server returns, and a success code on an error page tells it the page is real.
Fix: configure the error template to send a not-found status, or a gone status for content you have removed on purpose, while keeping the helpful message and navigation for people.
Verify: request a deliberately wrong URL and a removed one and confirm each returns the correct code; the examples then move from the soft row to the not-found row of the report.
Prevent: include status checks on error templates in release testing, using the control table in crawling and indexing. The same template can also send an analytics event, so that not-found visits and the links that cause them show up in reports, as the tutorial on tracking site errors and failed actions sets out. Once those URLs move to the not-found row, the page on not-found errors in Search Console sorts that list into the entries to leave alone and the ones that still need work.
Empty listing pages
A category with no products, a tag with no posts, or an internal search page with no results is a page about nothing. Large catalogues create these whenever stock moves or filters combine.
Fix: for categories that will refill, keep the page but add useful category copy, related categories and links to similar items; for categories that will stay empty, merge them into a parent and redirect; keep internal search result pages out of the index altogether.
Verify: the flagged listings either show content again or have left the report.
Prevent: let the catalogue system hide empty categories from navigation and sitemaps automatically, as the category and product page guide recommends for catalogue templates.
Unavailable products with nothing left on the page
When a product sells out, some templates strip the description, price and images, leaving a bare "out of stock" notice. That reads as a missing page.
Fix: for temporarily unavailable items, keep the full product content, state availability clearly and offer alternatives or a restock alert. For discontinued items with a clear successor, redirect to it; for discontinued items with search value of their own, keep an informative page that links to replacements; otherwise return a gone status.
Verify: the product URLs leave the soft row, and retained ones are indexed.
Prevent: write a lifecycle rule for availability, following the section on out-of-stock and discontinued items in the category and product page guide.
Redirects that land somewhere unrelated
Sending every removed URL to the home page or a broad category feels tidy but gives a searcher nothing they asked for. Google's redirects documentation explains that redirects should point to an equivalent page; a redirect to an irrelevant destination can be treated like a not-found.
Fix: map each removed URL to its closest real equivalent, and where none exists, let it return a gone or not-found status instead.
Verify: follow a sample of redirects and confirm each lands on a page a visitor would accept as a substitute.
Prevent: require a one-to-one redirect map for every removal and every site migration, rather than a catch-all rule.
Content that never renders
If the main content is injected by JavaScript and the request behind it fails, times out or is blocked from the crawler, Google renders a frame with no substance. Google's guide to fixing search-related JavaScript problems covers exactly this pattern, including blocked resources and errors visible only to the crawler.
Fix: move the essential content into the initial HTML, or repair the failing call: allow the resource in robots.txt, handle errors gracefully, remove dependencies on user interaction.
Verify: the live test screenshot now shows the full page and the URL can be indexed.
Prevent: test rendered output in releases that touch shared components, using the checklist in JavaScript SEO.
Prevention: status codes that tell the truth
The underlying rule is simple to state and easy to drift from: a URL should return success only when it shows something worth finding. Write down, for each template, what happens when its data is missing: an empty category, a product removed from the feed, an article unpublished, an API down. Each case needs a deliberate answer, and the answer should be visible in the status code as well as the design. Keep a redirect policy that names acceptable destinations. Recheck the soft error row after every catalogue import and template release. Avoid using a Canonical TagTells search engines the preferred URL. to paper over empty pages; consolidation does not repair a page with no content, and duplicate clusters belong on the sibling page about duplicates without a user-selected canonical. Removing truly empty URLs also stops Google spending Crawl BudgetHow much a search engine crawls a site. on them.
For listings, job postings, events and offers that end by design, the guide to expired listings, events and offers writes the rule for each state before the pages expire.
An outdoor retailer after the winter range ended
An outdoor clothing shop sees 640 soft errors in April. The export splits into three groups: 410 winter products whose template hid everything except an "unavailable" banner, 160 empty seasonal categories, and 70 old blog posts redirected to the home page during a redesign. The team restores product content with a "back in autumn" note and links to current lines, merges the empty categories into their parents with redirects, and remaps the 70 posts to the closest current articles, returning a gone status for 18 with no equivalent. By June the soft error count is under 40.
Common mistakes
Redirecting every flagged URL to the home page. Adding noindex to the soft errors so they disappear from the report while the template stays broken. Returning a not-found status for products that are only briefly out of stock. Testing the page in a logged-in browser and concluding Google is wrong.
