Do Redirects Lose Link Equity? A Test

Google says 3xx redirects lose no PageRank; a generation of SEOs remembers the era when they did and hedges accordingly. Since we run migrations professionally, we've had repeated natural experiments in whether redirected equity actually arrives — and one deliberate test: a set of linked, ranking URLs redirected under controlled conditions, tracked against untouched siblings. Here's what the data shows about clean redirects, chained ones, and irrelevant ones — because those are three different answers.

The deliberate test

During a planned URL restructure on one of our network sites, we split comparable linked pages into groups: clean one-hop 301s to equivalent content (URL changed, page substantially identical), a small set routed through a deliberate two-hop chain (old → intermediate → final; the kind migrations accumulate accidentally), and untouched controls. All pages had rankings and external links; internal links were updated to final URLs everywhere (isolating the redirect as the external-equity path, per the hygiene rules).

What happened

1. Clean 301s: equity effectively transferred. The pattern across the group: a dip of a few positions in weeks 1–3 while the recrawl processed the moves, then recovery to baseline (±1 position) by weeks 4–8, holding through the six-month window. Traffic followed the same U-shape. Consistent with Google's no-loss claim as a settled end-state, with a real transition cost the claim never mentions — the migration wobble is the price, not a leak.
2. The chains: same destination, worse journey. Chained pages dipped deeper and recovered slower (several still a position or two under baseline at month three; recovered by month six after we flattened the chains mid-test — which itself produced a visible second recovery step). Whether the mechanism is crawl friction, partial per-hop discounting, or evaluation lag, the operational conclusion is identical: chains cost time and risk, flatten them always — exactly what the audit's plumbing pass exists for.
3. The natural experiments on relevance: across client migrations we've also watched the third case repeatedly — 301s to non-equivalent content (retired pages bulk-pointed at homepages or thin category stubs): rankings for the old pages' terms simply die, equity transfer unobservable. Google's soft-404 treatment of irrelevant redirects is the practical reality: the 301 passes equity to the degree the destination deserves it — relevance is the exchange rate, as the expired-domain traders learn expensively.

The operating rules this confirms

Everything the migration checklist already says, now with receipts: exact-counterpart mapping (relevance is what transfers), single hops (chains tax the transition), internal links to final URLs (don't spend the redirect where you control the link), redirects kept forever (our test's old URLs still receive crawler and human hits months later), and the settling window budgeted honestly — weeks of wobble is the clean case, and panic edits during it are how clean cases become messy ones. And the reassurance the test genuinely provides: URL changes done right are safe — the dip is temporary, the equity arrives, and fear of redirects shouldn't freeze a needed restructure, per the URL guide's counsel.

Frequently asked questions

Do 302s behave differently in practice?

Long-lived 302s in our observations eventually behave like 301s (Google treats persistent temporaries as permanent), with slower, murkier settling — ambiguity you never needed. Say what you mean with the right code, per the decision rule.

How much does one extra hop actually cost?

In end-state equity: unmeasurably little in our data once flattened; in transition: weeks of extra depth and delay per hop, plus fragility (every hop is a future breakage point). The cost is operational, and operational costs compound across thousands of URLs — flatten at migration, audit annually.

Can I redirect an old domain into my site for its links?

The relevance finding is the whole answer: equivalent-content, page-level mapping transfers; bulk irrelevant grafting doesn't (and flirts with the abuse patterns Google names). Audit what you're inheriting first — or skip the archaeology and build the equity fresh where it's cleanest: earned placements on live publications (this way).