Skip to content

How to remove AMP and keep your article URLs working

Retire paired AMP pages while keeping the original articles online. Test old permalinks, remove the amphtml reference, and verify the final destination.

By Crawlito · 6 min read

In short

  • Determine whether AMP is a secondary format or the only version of the article before removing it.
  • Remove the amphtml reference from retained canonical articles.
  • A redirect from each former AMP permalink can take existing readers directly to its full article.
On this page

To retire AMP while keeping a publication online, identify each AMP page's regular article, remove the rel="amphtml" reference from that surviving page, and route old AMP permalinks to the corresponding article. Check the old link from a fresh session. A site whose only copy of an article is AMP needs a working replacement before it removes that version. Google’s instructions for removing AMP pages

Inventory the pairs that visitors can actually open

Start with a few representative articles, including one old enough to have appeared in newsletters and one with a newer URL scheme. Save both the published non-AMP URL and the AMP URL. A plugin might generate /amp/ paths, a query parameter, or a separate mobile hostname. The output of the CMS is a better source for this mapping than guessing a pattern from one story. Watch for a legacy redirect that the current CMS has never generated.

Open each version and look for its identity. A paired AMP document typically identifies the regular article as canonical while the regular article links to the AMP alternate. An AMP-only publication may have a self-canonical article with no conventional counterpart. Record this explicitly. If an editor says the page can be removed because “we already have the regular article,” the map should identify the actual working URL and show the text readers are supposed to keep.

Include linked assets and actions that matter to the article. A correction notice, a downloadable map, a recipe ingredient list, or a chart can vanish during a format change even when the headline and lead paragraph survive. Decide what counts as the same content for the people who bookmarked it. The site migration guide can help with URL mapping when a broader redesign coincides with AMP retirement.

Keep one permanent destination for each article

Choose a canonical destination that serves the full article to anonymous visitors. A permanent move from the AMP route to this final URL is a practical way to preserve an old permalink. Google's paired-page removal instructions cover redirects and preserving permalinks. Test the actual HTTP response rather than relying on a mobile redirect implemented only after a script runs. The request should arrive at the article even from a client that does not execute that script.

Suppose /features/river-walk/amp/ used to be paired with /features/river-walk/. The publication also changed the section name to /guides/river-walk/. Point the retired format directly at the final guide when possible and check any existing redirect from the former regular path separately. Keep the title or a distinctive paragraph in your test record, since a generic homepage also returns a successful final response.

bash
curl -sS -D old-amp-headers.txt -o /dev/null https://example.com/features/river-walk/amp/
curl -sSL -D amp-chain.txt -o destination.html https://example.com/features/river-walk/amp/

The first request preserves the immediate response. The second follows the journey a visitor takes. Read the final content and its canonical address. For a query-based AMP variant, explicitly test that your redirect drops the obsolete AMP selector while preserving legitimate product, language, or article identifiers that determine the destination. A redirect that reconstructs the old query unchanged can send the request back into the same rule.

Remove alternate discovery at the publishing source

Remove the AMP alternate from the source template of retained regular articles. Check for a second publisher that can reintroduce it, such as an SEO plugin or a head injection shared across themes. Search the rendered production source for amphtml after the change. Removing the alternate on one manually edited article leaves the next story the CMS generates vulnerable to the same stale reference.

Inspect internal links, feeds, navigation, and promotional links that your publication can update. Link directly to the retained canonical URL when editing those systems. For old addresses you cannot edit, such as links in an email already sent, the redirect remains useful. The guide to internal links covers how to find an alternate URL that is still being distributed. Include old AMP paths in an access-log sample to see which ones readers actually continue to use.

Avoid blanking out an AMP template and returning a nominally successful empty document. Google's removal guidance specifically warns about empty invalid AMP documents and continued use of a previously valid copy. Keep the intended state explicit: the original content lives at a working public URL and the former format's address has a clear response. The canonical URL guide explains how alternate signals can conflict when the old template remains active.

Verify the transition without losing valid mobile pages

Check at least one mobile and one desktop session for each URL pattern. Verify the title, a distinctive part of the body, the canonical, and article-specific media or links. Test one path that contains the letters “amp” as part of an ordinary slug. A broad string replacement can damage unrelated words or non-AMP article paths. A negative test should leave the ordinary article untouched.

Choose a genuinely missing path and confirm it stays missing. If all unrecognized AMP-like URLs redirect to the publication homepage, broken links may appear to work in a status-only test. Someone following an old bookmark will have no explanation for landing at a generic destination. Where no replacement exists, handle the retired content according to your normal deletion policy and document that exception in the migration map.

Track the public URL response separately from the search engine's current copy. A successful transfer can be visible immediately to a person following the permalink while reports still describe an earlier crawl. Preserve the release time, the last observation, and one clear example of each verified pattern. The next time a reader reports an old cached result, you can check the actual address and determine whether it is a known pair or an unhandled legacy format.

Find an old article link in a public feed archive, a newsletter you own, or a past social post. Follow the URL verbatim without first visiting the publisher homepage. It may contain a different AMP path convention from the current plugin output, particularly if the site used more than one CMS. Add a row to the migration map for each distinct pattern you actually find. A hand-curated sample of the newest stories can easily miss this historical traffic.

A route served by a third-party cache may involve infrastructure the publisher cannot change directly. Capture the final publisher URL reached from the old link and compare that address with the mapping you control. When checking a report, distinguish a cached representation from a live request to your own old permalink. The former can have its own refresh behavior while the latter should follow the server policy you just deployed.

Frequently asked questions

Should we remove the regular article from the sitemap?

Keep the canonical URL for an article that remains public and intended for discovery. Check that any sitemap generator now emits the maintained version.

What if the AMP page was our only article?

Publish and verify the replacement first. Decide whether to keep the existing public address or move it, then test the resulting route before retiring the old format.

Can a redirect to the section hub stand in for an article?

A genuine matching article is the most useful destination. Use the hub only when it actually satisfies the reader's request, and handle unrelated deleted material according to its real state.

Check a surviving article

Use the free page check on a canonical article after retirement, alongside a redirect test of its former AMP permalink. Together they show where a reader lands and what the destination serves.

Filed under Access and robots.txt. Topics: redirects, canonical, google, migration