Old links remain valuable when they still help readers find the corresponding content. This article reviews the implementation of 11 permanent redirects and the restoration of 2 articles at their original URLs, providing HTTP check commands, database safeguards, and the limits of what crawling and indexing can establish.
After Article Deduplication, People Still Visit Old Links
Removing duplicate content may reduce the number of articles, but old links in search engines, bookmarks, and articles on other sites do not disappear along with them. On September 14, 2026, Aperture addressed a batch of previously identified old URLs: of 13 URLs, 11 received permanent redirects, and 2 distinct articles were restored at their original URLs. All 22 existing public URLs and their original publication timestamps were retained, bringing the number of public articles to 24 after the repair.[S1]
The goal of this case was to preserve the content relevance of old entry points: duplicate articles were redirected to relevant articles with sufficient coverage, while distinct content was restored after additional review. It did not demonstrate an improvement in search rankings or promise that search engines would complete indexing by a particular date. This article reviews the batch handled on September 14; any old links discovered later require separate review. A one-time repair should not be treated as permanently eliminating the problem.
Determine the Content Relationship Before Deciding What to Do with the URL
A URL returning 404 only indicates that no resource is currently available; it does not, by itself, establish which article should be restored. Before taking action, check the old article’s full text, publication history, supporting sources, and the current article catalog. Permanently redirecting the old URL to an existing page is appropriate only after confirming that the page covers its central questions and substantive content.
Google’s site migration guide calls for preparing a mapping between old and new URLs and warns against redirecting many old URLs to an irrelevant destination, such as the homepage, because this may be treated as a soft 404. The same principle can guide article consolidation: when readers arrive through a link to a specific article, the destination page should answer the original question.[S3]
In this repair, the site first compared the full texts, then incorporated substantive information that needed to be retained into the destination articles and reviewed them again. The articles on historical European Central Bank interest rates and historical Nasdaq observations could not be replaced by existing content, so they were restored at their original URLs after revisions to their sources and citations. These actions were recorded as review outcomes, rather than merely implemented as temporary rules at the edge layer.[S1]
What 301 Redirects, Canonicals, and Sitemaps Each Address
A 301 is an HTTP response in which the server tells the client that a resource has moved permanently, with the destination specified in the Location header. MDN also notes that, depending on the request method, following a 301 may change the method to GET. This case therefore concerns GET and HEAD entry points for public content; a redirect rule for an ordinary web page cannot simply be applied to payments or write operations.[S5]
A canonical is a signal through which the current page identifies its preferred canonical URL; it does not make an old URL that already returns 404 work in the browser. Google classifies permanent redirects and rel=canonical as strong canonicalization signals, and inclusion in a sitemap as a weaker signal. These methods can work together, but a search engine may still choose a canonical page based on other signals.[S4]
The site redirects old web pages to specific destination articles using 301 responses. The destination articles return 200 and have self-referencing canonicals, while the sitemap lists only the retained public canonical pages. Google’s sitemap guide recommends listing the canonical URLs you want to appear in search results. Submitting a sitemap is a discovery mechanism, not a guarantee that every entry will be crawled or indexed.[S1][S6]
| Component | Role in This Case | Work It Cannot Replace |
|---|---|---|
| 301 + Location | Directs visits to old URLs to relevant destination articles | Cannot replace reviews of content coverage and quality |
| Self-referencing canonical | Identifies the destination page’s preferred URL | Cannot repair an inaccessible old entry point |
| sitemap | Lists the current discoverable public canonical pages | Cannot guarantee indexing, rankings, or clicks |
Mappings Need Persistent Storage and Protection Against Later Tasks
The site stores source_slug, the original article ID, the destination article ID, content hashes for both articles, and review records in the article_redirects table. Web pages, article JSON, and Agent entry points read the same mapping, reducing the chance that different entry points will lead to different content.[S1][S2]
Database safeguards also constrain subsequent changes: old articles that have been mapped cannot be arbitrarily deleted, renamed, or republished, and destination articles cannot be deleted or unpublished. This prevents scheduled collection jobs and editorial operations from turning old entry points back into 404s on the next run or republishing duplicate content that has already been consolidated.[S1][S2]
Two separate checks also precede the creation of a mapping: first, that the destination is currently public and readable; second, that the article text and sources compared during review have not changed by the time the mapping is committed. The site locks the relevant tables within a transaction, rechecks the catalog and content hashes, and then writes the mapping. These checks prevent outdated reviews from being committed.[S1] When implementing a similar process, it is advisable to complete model-based review outside the transaction so that database locks are not held during lengthy network calls.
Verify Direct Access, Machine Interfaces, and On-Site Navigation Separately
Seeing a normal title in the browser does not prove that the server actually issued a 301. A front-end application may change the content without updating the URL, and a CDN may continue serving an old cached response. Check the HTTP status and Location first, then the final page’s content and canonical, and finally on-site navigation with JavaScript enabled.[S1]
The following is an example from the actual mappings in this repair. The first command inspects the old URL’s status and Location; the second follows the redirect to confirm the final destination; the third checks the canonical URL in the destination HTML. These commands target public, read-only pages and contain no administrative credentials.[S1]
curl -I 'https://aperture.zhangwangshu.com/article/agent-research-20260903-2103-396'
curl -L -o /dev/null -w '%{http_code} %{url_effective}\n' \
'https://aperture.zhangwangshu.com/article/agent-research-20260903-2103-396'
curl -s 'https://aperture.zhangwangshu.com/article/agent-research-20260905-0003-599' \
| rg 'rel="canonical"'Do Not Let Campaign Parameters or Caching Hide Problems
For ordinary GET requests to articles, old links with UTM parameters should still reach the corresponding content. Campaign parameters support traffic-source attribution, while the canonical URL generally retains the article path without those parameters. During implementation, separately check URLs with and without parameters, trailing-slash variants, and direct access to the destination to avoid redirect loops or irrelevant intermediate pages. These are checks that should be performed; the result of any single test should not be generalized into a claim that all variants will remain correct forever.
Refreshing caches only helps clients see new responses sooner; it does not make search engines update their indexes immediately. Google’s migration guide warns that fluctuations may occur during recrawling and reindexing and recommends retaining redirects for a sufficiently long time. The site’s page safeguards aim to keep valid old entry points available over time, rather than treating redirects as temporary patches that can be removed after a few days.[S3]
Machine interfaces should also retain their own semantics. For example, open JSON interfaces and paid interfaces must not be conflated. Even when they reference the same article, their paths, access conditions, and final responses must be verified. The unified mapping in this case unifies content identity; it does not remove the access requirements of individual entry points.[S1][S2]
Acceptance Results for This Repair—and What They Did Not Prove
The September 14 verification records show that GET and HEAD requests to 11 old URLs returned 301, while all 24 public articles returned 200 and had a single H1 and the correct canonical. The article sitemap contained 24 articles and excluded redirect source URLs. Browser testing also covered scripts enabled and disabled, on-site navigation through old links, URL updates, and back/forward navigation.[S1]
These results establish that the paths and rendering behavior met the repair goals at that time; they do not establish that all articles had been indexed. Page indexing status, search query impressions, and clicks in Google Search Console require separate inspection. Googlebot or Bingbot visits in logs are likewise only evidence of crawling.
For ongoing maintenance, “old URLs returning 404 that have still received requests recently” can become a standing checklist. For each newly discovered entry point, first revisit content coverage and publication history, then decide whether to restore it, apply a relevant 301, or retain a genuine 404. In this way, page counts, search availability, and content deduplication no longer have to conflict: what is consolidated is duplicate expression, while what is preserved is the entry point readers still need.
Keep following engineering practices
Subscribe to new articles with an RSS reader, or share this article with someone working on similar problems.
Subscribe to RSS updates