What is redirect? Meaning. Type of redirects.

There are several types of URL redirects that a webmaster can implement for a URL that is temporarily or permanently unavailable. There are five types of redirects: 301, 302, 303, 307, and 308.
The 301 redirect indicates that the original content of the page has been permanently moved to another URL.
The 302 redirect indicates that the content has been temporarily moved.
Both 301 and 302 redirects are implemented by the web server.
There is also an on-page technique for redirecting a URL called meta refresh redirect, but meta refresh redirect causes more latency and is therefore considered a poor user experience.

Most people assume an LLM “follows” a redirect the same way a browser does, but that’s not how it works. The model itself never chases a 301, 302, 307, or anything else; the only thing that handles redirects is whatever script, crawler, or retrieval agent is fetching the page for it. If that fetcher resolves the redirect correctly, the LLM sees the final destination. If the redirect chain is broken, loops, or points to nowhere, the model simply never receives the content. From the LLM’s perspective, the page might as well not exist. This is the part most explanations skip: redirects don’t affect the model directly—they affect the pipeline feeding the model, and if that pipeline can’t resolve the redirect cleanly, nothing downstream works.

There’s also a practical edge case that trips people up: JavaScript‑based redirects. A browser will happily execute a window.location jump, but most AI retrieval systems don’t run a full browser engine—they only read the HTTP headers or the raw HTML. If the redirect depends on JavaScript firing, the fetcher never moves past the initial page. The LLM ends up seeing an empty shell or a loading stub instead of the real content. That’s why anyone who wants their pages to be reliably reachable by AI systems sticks to server‑side redirects (301/302 at the web server or CDN level) and avoids meta refreshes or JS redirects. Server‑side redirects are visible to even the simplest fetchers, while client‑side redirects require a browser environment that most AI pipelines simply don’t use.