Why does my own website show up as a referrer in my analytics?
Because a visit restarted while the person was still on your site, or crossed onto an address your analytics treats as a different website, so the next page they loaded carried your own address as its referrer. Most analytics tools close a session after about 30 minutes of inactivity, so someone who leaves a long page open and then clicks an internal link starts a new session whose source is you. It is a labelling problem, not a traffic source: list your own domains as unwanted referrals, tag the links in your own emails, and expect those visits to move into Direct rather than disappear.
Four ways you end up referring yourself
A referrer is only the address of the page someone was on when they clicked. Normally that is another website, which is why analytics reports it as a traffic source. Your own address gets into that column through one of four ordinary routes, and each leaves a slightly different mark.
- The session timed out mid-visit. Analytics tools split activity into sessions and close one after a stretch of inactivity — 30 minutes is the usual default. A visitor who leaves a tab open over lunch and then clicks a link on your page starts a fresh session, and the only source that session has is the page they clicked from: yours. The mark is your own domain as referrer, landing on a page you link to internally.
- They crossed between two of your own addresses. A marketing site on one domain and the app, blog, docs or checkout on another looks like two unrelated websites to a tool that has not been told otherwise. Every click from one to the other is recorded as a referral from yourself, usually in steady numbers that rise and fall with your signups.
- They took a round trip through someone else's page. Payment processors, card-verification pages and sign-in or password-reset flows send people away and bring them back. On the way back the referrer is that third party, so strictly it is not your own name — but it is the same fault: a step inside your own flow reported as the place the visitor came from. Google's help page for its exclusion setting names exactly these cases: payment processors, 3-D Secure pages and password recovery.
- They clicked a link in an email you sent. An email is not a web page, so what shows up depends entirely on how it was read. A webmail client sends its own address as the referrer, a desktop or phone mail app usually sends nothing, and an email service with click tracking routes the click through its own redirect domain first. The same message can arrive under three different labels.
What it looked like in our own report
We found every one of these on our own site. Auditing one day's channel breakdown on 30 August 2026, the Referral bucket held five visitors, and exactly one was a genuine referral — from GitHub. Two came from our own domain: sessions that had timed out while someone was still reading and restarted on an internal click. The other two were our own emails, one arriving through our email provider's click-tracking redirect and one from a webmail client. Four of five referrals were us, and nothing in the report said so.
| Referrer you see | Most likely cause | What to do |
|---|---|---|
| Your own domain | A session timed out mid-visit | List it as an unwanted referral |
| Another domain you own (app, blog, docs) | Two properties measured as separate sites | Set up cross-domain measurement, or list it as unwanted |
| A payment or sign-in provider | A round trip inside your own flow | List it as an unwanted referral |
| Your email provider or a webmail address | A click in an email you sent | Put UTM parameters on every link you send |
Email needs a different fix from the rest, because excluding a referrer only removes a wrong label; it does not put the right one back. Tag every link you send with UTM parameters and the visit is recorded as email whichever client the reader used. We now add them automatically to every link in every email we send, because leaving it to each template is how the first untagged one slips out.
Fixing it makes Direct bigger, and that is correct
In Google Analytics 4 the setting is called List unwanted referrals, in the tag settings of your web data stream; a matching referrer is ignored for attribution instead of being shown as a source. Other tools have an equivalent exclusion list, or let you rewrite the referrer before an event is sent — which is what we did: any referrer on our own host is now recorded as no referrer at all.
Here is the part we predicted wrongly, so you do not have to. We expected the clean-up to shrink our Direct traffic, and it grew it. An excluded referrer does not delete the visit; the visit still happened, it simply no longer has a named source, and a visit with no named source is by definition Direct. Judge a fix like this by what each bucket now contains, not by whether a total went down.
That also means a larger Direct number after the change is not automatically real people. Direct is everything that arrives without a referrer, and automated clients arrive without one too — how much of your traffic is actually bots covers how to tell them apart. Some of what lands there is the most valuable traffic you get, though: dark social is the name for the word-of-mouth share of it.
常见问题
- Does self-referral traffic hurt my SEO?
- No. It is a labelling issue inside your analytics, and search engines never see it. The damage is to your own decisions: a channel report that credits your own domain hides where people really came from, and attribution models hand credit to a source that is actually the middle of a visit.
- Should I just filter these visits out of my reports?
- No — exclude the referrer, not the visit. These are real people, and often your most engaged ones, since they were still on your site after half an hour. Filtering them out removes good visitors to fix a label. The exception is your own team: their browsing tells you nothing about customers and is worth excluding completely.