Redirect chain: how many hops it takes to reach 30 homepages

FIELD TEST · 2026-08-18 · 30 sites · single snapshot

Sample and method: the same 30 sites from our robots.txt, sitemap and homepage structure surveys. Each chain started at http://<domain> and every hop was recorded separately, browser user agent, 2026-08-18, from an exit in Japan. Same crawl as four other posts in this batch.

Every site in the panel makes a crawler take at least one detour before it sees a homepage. All 30 redirect plain HTTP somewhere else — not one answers it with a page. Twenty-five finish there, four take a second hop and one takes three. The redirect chain is short almost everywhere, and the interesting part is what the extra hops are for.

How we measured it

Fixed before the crawl, unchanged after.

1. Request http://<domain> without following redirects, and record the status code and the Location header.
2. Resolve that Location against the current URL and repeat, up to ten hops.
3. Stop at the first response that is not a 3xx, and count the redirects, not the responses.

We started from the bare domain over HTTP because that is the least informed request anyone can make of a site — the shape a stray link, an old bookmark or a crawler working from a plain-text mention would produce.

How many hops it takes to reach 30 homepages

The distribution is tight. One hop is the norm, and nothing in the panel comes close to the limit crawlers enforce.

Google's documentation puts the ceiling at ten: "By default, Google's crawlers follow up to 10 redirect hops. However, specific products' crawlers may have different limits." Read 2026-08-18 on the HTTP status codes page at developers.google.com. Nobody here is near it. The guidance that matters is the softer one from the crawl budget guide: "Avoid long redirect chains, which have a negative effect on crawling."

Which status code the first hop uses

The upgrade from HTTP to HTTPS is the one redirect every site in the panel performs, and they do not agree on how to express it.

The BBC answering plain HTTP with a 302 is the odd one out. A temporary redirect says the old address may come back, and that is a weaker consolidation signal than the 301 the other 29 send for the same move.

Three sites also write an explicit port into the target: slack, webflow and arstechnica all send Location: https://host:443/. It resolves fine. It also means the canonical address a crawler first learns about carries a port number that no other link on the internet uses.

The extra hops are locale routing, not sloppiness

Only five sites needed more than one hop, and each one had a reason worth naming.

Slack's is the longest and the strangest. It goes to https://slack.com:443/, then to /?unknown_domain=1, then to /intl/ja-jp/?unknown_domain=1. The address a crawler would record as the homepage carries a parameter that reads like a diagnostic flag.

Why our own location is part of this result

Stripe sent us to /jp and Slack to /intl/ja-jp/ because our requests exit in Japan. Those hops are not defects in their setup; from Osaka they are exactly right.

They are also the point. A crawler fetches from wherever its infrastructure sits, and it does not shop around. If you route visitors to a locale by IP address before serving anything, the version that gets crawled and quoted is whichever one your redirect hands to that address block. We hit the same wall measuring canonical tags in the previous batch, and dropped that topic rather than report an artifact as a finding.

Geolocation redirects do not show a crawler your site. They show it one country's copy of your site.

What this crawl cannot tell you

We did not test anything past the front door. Internal links, old article URLs and campaign parameters are where long chains actually accumulate, and one request to a bare domain says nothing about them.

We also cannot tell how any of these sites respond to a crawler's own address block. Every hop above was served to a browser user agent from one IP address in one country, and geolocation logic can differ per range.

What to do about your own redirect chain

Two checks and two habits.

• Run curl -sIL http://yourdomain.com and count the 3xx lines. Two is normal, three deserves an explanation, four means something is redirecting to something that redirects.
• Make the HTTP to HTTPS jump land on the final host in one move. Sending http://example.com to https://example.com and only then to https://www.example.com is one avoidable hop on every entry point you own.
• Do not use 302 for a move that is permanent. It costs you the consolidation the 301 would have given.
• Do not decide a visitor's language by IP address before they have seen anything. Link the locales instead, and let the crawler see all of them.

Which locale versions a site declares to search engines is measured in the hreflang survey, and it is the alternative to the IP-based routing above. What a crawler finds when a URL resolves to nothing is in the soft 404 survey. Following every chain on a site instead of one front door is part of what QueryWin is being built to do.

Common questions

How did you measure this?

One request per hop, starting at http://<domain>, redirects resolved manually so each Location could be recorded. Browser user agent, 2026-08-18, one pass, no retries.

How many redirects are too many?

Google's crawlers follow up to ten by default, so the technical limit is generous. The practical answer from this panel is that 25 of 30 need exactly one, which makes two the number that should prompt a look.

Is 308 better than 301?

For a homepage the difference is academic — both are permanent and both consolidate. 308 preserves the request method, which matters for form submissions and not for a crawler fetching a page.

Does a redirect chain lose ranking signals?

Google has said permanent redirects pass signals, and this crawl tested none of that. What we can say is that every hop is a request a crawler spends without receiving content.