Tags: web-dev concept

Redirects and Link Equity

Date: 2026-08-16

---

A redirect tells everyone where something went. Get the type wrong and signals don’t transfer; chain them and each hop costs a round trip and some crawl budget; forget them and years of accumulated links point at 404s.

---

What it is

A redirect is a response instructing the client to request a different URL. Link equity is the accumulated ranking value from links pointing at a URL, which a redirect is supposed to pass along.

The types

Code	Meaning	Passes signals	Method preserved
301	Moved permanently	Yes	No — POST becomes GET
302	Found, temporary	Treated as temporary; original stays indexed	No
307	Temporary, strict	As 302	Yes
308	Permanent, strict	As 301	Yes
meta refresh	Client-side	Poorly, if at all	—
JavaScript	Client-side	Only if rendered; not by AI crawlers	—

Use 301 for anything permanent. The distinction that catches people is that 301 and 302 rewrite POST to GET, which breaks form submissions — 307 and 308 preserve the method, and are what you want on API routes.

Client-side redirects are a last resort. They’re slow, they cost a full page load, and non-rendering crawlers never see them — Rendering and SEO.

Chains and loops

BAD                                          COST

http://example.com/old-product               each hop is a full
  → https://example.com/old-product          round trip: ~150ms
  → https://www.example.com/old-product      on mobile
  → https://www.example.com/products/new
                                             450ms before the real
  4 requests, 3 redirects                    page starts loading

Chains cost latency directly — they’re the largest avoidable component of Time to First Byte — and each hop consumes crawl budget. Some engines stop following after a handful of hops.

Fix by flattening: every old URL should redirect to the final destination in one hop, not to the next link in a historical chain. After several migrations this requires actually auditing the redirect table rather than appending to it.

Loops are worse and are usually a conflict between two rules — a canonical host rule and a trailing-slash rule disagreeing.

Where equity leaks

• Redirecting everything to the homepage. A blanket rule for a discontinued range is treated as a soft 404, and passes nothing. Redirect to the nearest genuine equivalent, or return 404 honestly
• Redirecting to a noindex page, which is a contradiction
• 302 for a permanent move, so the old URL stays indexed and the new one never accumulates
• Chains, which dilute at each hop
• Removing redirects after “enough time”. External links don’t expire. Keep them indefinitely — they cost almost nothing

Out-of-stock and discontinued products

The ecommerce-specific case, and it’s a genuine judgement call:

Situation	Response
Temporarily out of stock	Keep the page, 200. Show alternatives and a back-in-stock signup
Permanently discontinued, close equivalent exists	301 to the equivalent
Permanently discontinued, no equivalent	301 to the parent category, or 404 if the category is a poor match
Never existed	404 or 410

Deleting a product URL throws away every link it ever earned. For a product that’s been listed for years, that’s real accumulated value, and the cost of redirecting is one row in a table.

Returning a 200 with “product not found” is the worst option — it’s a soft 404, it wastes crawl budget, and it leaves a dead page indexed.

Implementation

• At the edge or CDN for large redirect tables, so they don’t touch the origin — Edge Computing
• In server config for pattern rules
• In the application only for logic that genuinely needs application state
• Never in a tag manager. A client-side redirect from a container is slow, breaks for non-rendering crawlers, and is invisible to whoever debugs it later

For a migration, generate the map from a crawl of the old site rather than from the sitemap — the sitemap won’t contain the URLs that accumulated links years ago. See Site Migrations and SEO.

Auditing

curl -sIL https://example.com/old-url | grep -E "^HTTP|^location"

Shows every hop and its status. A site crawler will do this in bulk and flag chains, loops and redirects to error pages — worth running before any launch and periodically afterwards, since redirect tables accumulate silently.