In short. The HTTP Referer header carries a client-reported referring address, subject to omission and limits on how much information is sent.

Definition: The HTTP Referer header is a request field that identifies the resource from which the requested address was obtained. Its value can contain a referring page address or a shortened form, and some requests omit the field.

Key facts

The IETF's RFC 9110, section 10.1.3 defines Referer and explicitly identifies its spelling as a misspelling of “referrer”.
The W3C's Referrer Policy specification defines no-referrer, which causes the header to be omitted entirely.
As of 17 October 2026, Apache's Log Files documentation includes Referer in its Combined Log Format, recording the referring site reported by the client.

What information does the header carry?

A user agent is the client software that sends a request. RFC 9110 permits it to supply an absolute address or an address relative to the request target. The RFC gives this header example:

Referer: http://www.example.org/hypertext/Overview.html

Here, the value identifies the referring resource. The RFC names analytics, logging and finding obsolete or mistyped links as uses for the field.

The user agent must exclude any fragment identifier for a secondary resource and any user information component containing credentials. The value may also be shortened beyond those required removals. An origin is the combination of a URL's scheme, host and port; the RFC permits truncating parts other than the referring origin.

How does Referrer-Policy change the value?

A referrer policy controls how referral information is included in outgoing requests. The W3C's Referrer Policy Candidate Recommendation, dated 26 January 2017, defines a response header for communicating that policy:

Referrer-Policy: no-referrer

The policy header uses the correctly spelled word Referrer. The request header retains Referer.

The specification's strict-origin-when-cross-origin policy distinguishes requests within an origin from requests to another origin. Its example starts at https://example.com/page.html: a navigation to https://example.com/not-page.html sends the referring page address, while a navigation to https://not.example.com/ sends only https://example.com/. A navigation from that page to http://not.example.com/ sends no header.

For referral measurement, that origin-only value identifies the referring site while leaving the page path unavailable. This distinction matters when interpreting visits from links in answer engines: a recorded referring origin offers less detail than a full referring page address.

How can a website record it?

Apache's Log Files documentation shows a separate referral log using this configuration:

CustomLog "logs/referer_log" "%{Referer}i -> %U"

The %{Referer}i expression records the incoming request header. Apache also includes it alongside the client identification header in its Combined Log Format. Its documentation describes the referral field as information reported by the client, which is the appropriate interpretation when examining a server log.

FAQ

Why can a request have no Referer header?

RFC 9110 requires omission or about:blank when the address comes from a source without its own address, such as keyboard input or a bookmark. A referrer policy can also suppress the field.

Can you recover the referring page from an origin-only value?

The W3C's strict-origin-when-cross-origin example reduces https://example.com/page.html to https://example.com/ for a cross-origin HTTPS navigation. The resulting header contains no page path to recover.

Sources

Check how a page reads to answer engines. Open the GEO score analyzer