---
name: link-and-redirect-integrity-check
description: Use when asked to find broken links, broken images or assets, redirect chains and loops, or missing redirects on a website - routinely, or before and after a migration, redesign, CMS change, or URL restructure. Crawls politely, compares old and new URL inventories, and proposes redirect-map and link fixes as a review PR. Not for SEO content work (see seo-technical-audit).
---

# Link and redirect integrity check

Last reviewed: 2026-11-10

## Quick prompts

For anyone improving a site or app, whether a small-business owner or someone at a larger company practicing better prompts. Broken links frustrate readers and waste search traffic, and they pile up quietly over time. Paste a prompt into any AI assistant that can see the site or repo.

1. "Find broken links and missing images on the site I work on. List each with the page it's on and what it should probably point to. Report first; don't edit anything."
   *Why this prompt works:* Asking for a suggested target makes each finding easy to fix.

2. "Check whether links on the site I work on go through more than one redirect before landing. List the worst ones in plain language before changing anything."
   *Why this prompt works:* A narrow, named problem gives a short, useful answer.

3. "Here is a list of old page addresses: [paste]. Check where each now lands on the site I work on and flag any that hit an error or the home page. Report only; don't change anything."
   *Why this prompt works:* Giving the exact list removes guesswork about what to check.

## Hard limits

- Crawl only sites the owner controls; external links are checked with a single HEAD/GET each, rate limited, and never followed deeper.
- Do not change URLs, content, or navigation. Link fixes are limited to pointing a broken link at its correct existing target; anything ambiguous goes to the owner as a list.
- Redirect rules live in code (`_redirects`, server config, framework config) and ship as a review PR. CDN-level rules, DNS, and production deploys need the owner's approval.
- Use a real browser UA and pauses; some hosts block script user agents or rate-limit crawlers, which shows up as false 403/429s.

## Inputs to gather

1. Start URLs and environments (production, preview).
2. Old URL inventory for migrations: old sitemap, CMS export, analytics top pages, Search Console top pages, backlinks report if the owner has one.
3. Where redirects are defined today, and the host's limits (rule count, wildcard/splat support, status codes allowed).
4. Content types with special URLs: pagination, tags, categories, feeds, attachments, uploads/media paths, query-string URLs.

## Interface-type notes

- **Articles/blogs**: old date-based or `?p=123` URLs, category/tag archives, feed URLs, uploaded media paths, in-body links written years ago.
- **Docs**: versioned paths, anchors (`#section`) that must still exist after heading changes.
- **E-commerce**: discontinued products (redirect to category or keep with "unavailable", not to home), faceted URLs.
- **SPAs**: client routes need server fallbacks and real 404 status for unknown routes; check deep links on hard refresh.
- **Apps behind auth**: check public pages only unless given a test account; never crawl logout or destructive links.

## Steps

1. **Inventory**: crawl internal pages (HTML), collecting every `href`, `src`, `srcset`, CSS `url()`, canonical, `og:image`, and sitemap entry. Record status, final URL, and hop count.
2. **Classify problems**:
   - Internal 4xx/5xx links and missing assets (images, PDFs, fonts).
   - Redirect chains (more than one hop), loops, 302s that should be 301/308, redirects to the home page used as a catch-all (treated as soft 404s).
   - Internal links pointing at redirects (update them to the final URL).
   - Broken in-page anchors.
   - External links that are dead (404/410/DNS failure); treat 403/429 as "unverified", not broken.
   - Mixed content (http assets on https pages).
3. **Migration diff**: every old URL → expected new URL → actual result. Unmatched old URLs with traffic or backlinks need a redirect; truly removed content returns 410 or 404.
4. **Fix**: redirect map entries (most specific first, single hop, permanent status), internal link updates to final URLs, asset path fixes. Keep the redirect file sorted and commented by source.

## Validation

- Re-crawl the preview: zero internal broken links, zero chains over one hop, every old URL in the inventory resolves as planned.
- Test a sample of redirects in a real browser including trailing-slash and case variants and query strings.
- Confirm the host's redirect rule limit is not exceeded and rule order is correct.

## Report format

1. Counts: pages crawled, links checked, broken internal, broken external, chains, loops, anchors.
2. Tables: broken links (source page, link, status, proposed fix); redirect map changes (from, to, status, reason); unverified externals.
3. Migration coverage: old URLs total, mapped, intentionally removed, still unresolved.
4. PR link, "not merged"; owner decisions needed.

## References

- Google: Redirects and Google Search: https://developers.google.com/search/docs/crawling-indexing/301-redirects
- Google: Site moves with URL changes: https://developers.google.com/search/docs/crawling-indexing/site-move-with-url-changes
- Soft 404 errors: https://developers.google.com/search/docs/crawling-indexing/http-network-errors#soft-404-errors
- MDN HTTP redirections: https://developer.mozilla.org/en-US/docs/Web/HTTP/Guides/Redirections
- MDN 410 Gone: https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Status/410
- Tools: lychee (https://github.com/lycheeverse/lychee), linkinator (https://github.com/JustinBeckwith/linkinator), Playwright for rendered pages
