Referral exclusion is a setting in web analytics software that tells the tool to ignore the referring domain on certain visits, so that those domains are never recorded as the source of traffic or credited with a sale. It exists because the browser's referrer header records the last page a person left, not the marketing activity that brought them. A shopper who clicks a paid search ad, pays through PayPal and returns to an order confirmation page arrives back carrying paypal.com as the referrer. Without an exclusion, the analytics tool can credit the purchase to PayPal rather than to the ad.

The domains typically excluded fall into four groups: payment processors and buy now, pay later providers (PayPal, Stripe Checkout, Klarna, Clearpay), bank pages used for card authentication under 3-D Secure, sign-in providers such as accounts.google.com, and the site's own domains and subdomains, whose appearance as a referrer is known as a self-referral. In Google Analytics 4 (GA4), the feature is called List unwanted referrals. In its predecessor, Universal Analytics (UA), it was the Referral Exclusion List.

How it works in Google Analytics 4

Every page view sent by the GA4 tag carries a page_referrer value, read from the browser's document.referrer property. When a session begins, GA4 uses that value, together with any campaign parameters such as utm_source and the Google Ads click identifier (gclid), to assign a source, a medium and a channel. A visit from an unknown domain with no campaign tags becomes source "paypal.com", medium "referral".

The unwanted referrals list sits at the level of a web data stream, under Admin, Data streams, Configure tag settings. Each rule matches the referring domain using a condition such as "contains", "exactly matches", "begins with", "ends with" or a regular expression. According to Google's help page Identify unwanted referrals, when an incoming referrer matches a rule, the tag adds a parameter called ignore_referrer with the value true to the event. GA4 then processes that event as if no referrer existed.

The consequence depends on what came before. A visit with no referrer and no campaign tags is classed as direct. GA4's session-level dimensions use a last non-direct click rule, so a direct-looking return from PayPal does not displace the paid search click that started the journey. The purchase event is then credited by the property's attribution model, which by default also discounts direct visits when an earlier channel exists. The exclusion does not delete the visit, change the number of users, or stop a session from being counted.

The same parameter can be set by hand. Developers can send ignore_referrer as true from the gtag.js configuration or from a Google Tag Manager tag, for example only on a checkout return URL. That conditional approach is the workaround for gateways that use many domains, and open-source Tag Manager templates now package lists of known payment and 3-D Secure hosts for the purpose.

GA4 also excludes some referrers automatically. Referrals from the domain on which the tag fires, and from any domains listed under the stream's cross-domain settings, are not treated as new sources. That automatic step only works when the domains are configured, which is why misconfigured cross-domain set-ups produce self-referrals.

Two timing rules matter. Changes apply from the moment they are saved; GA4 does not reprocess historical data, so earlier PayPal-credited sales stay in the reports. And GA4, unlike UA, does not start a new session when the traffic source changes mid-visit. A session expires after 30 minutes of inactivity by default, adjustable by the property owner. A shopper who spends 35 minutes inside a bank's authentication flow comes back in a new session, and that new session takes its source from the referrer unless it is excluded.

Origin and evolution

The idea predates GA4 by more than a decade. The ga.js library of classic Google Analytics, released in December 2007, offered a method called _addIgnoredRef, which made the tag treat listed referrers as direct traffic. It required a code change on every page.

Universal Analytics, launched in public beta in October 2012 and released generally in April 2014, moved the setting into the administration interface as the Referral Exclusion List under property tracking settings. UA also added a property's own domain to the list automatically when the property was created. Because UA started a new session whenever the source changed, an un-excluded payment gateway in UA both split the visit in two and claimed the transaction.

Google announced on March 16, 2022 that Universal Analytics would stop processing data on July 1, 2023, with Analytics 360 properties shut on July 1, 2024. UA referral exclusion lists did not migrate. GA4, which launched on October 14, 2020, rebuilt the function at data stream level and exposed the underlying ignore_referrer parameter, a change that gave developers per-page control UA never offered.

Other analytics vendors handle the same problem under different names. Adobe Analytics uses internal URL filters in report suite settings to decide which referrers count as internal, and its marketing channel processing rules then decide how those visits are classified, according to Adobe's Experience League documentation. Matomo lets administrators set excluded referrers per site and globally, according to its FAQ.

Why it matters for marketers

Referral exclusion is a small administrative list with direct effects on budgets. Channel reports and attribution models feed decisions about where spend goes, and Google Ads can import GA4 key events for bidding. If a share of purchases is credited to a payment processor, every paid channel looks less productive than it is, and the error is uneven: it hits the sites where customers most often choose third-party payment.

The setting also sits inside a system many practitioners misread. In a December 2025 poll of 624 analytics professionals, 59% answered a question on GA4 session attribution incorrectly. Knowing that an excluded referral becomes direct, and that direct is skipped by last non-direct click, is part of the same knowledge. PPC Land's own guide to excluding referrals in GA4 and its GA4 audit checklist treat the list as a routine configuration step.

Ecommerce platforms add another layer. Shopify stores whose checkout runs on separate infrastructure have faced attribution gaps even after Google began routing Shopify purchases into GA4, a reminder that the referral list addresses only one of several ways a purchase can lose its source.

Limitations and disputes

The most common misuse is treating the list as a spam filter. Adding a spam domain does not remove its hits; it relabels them as direct traffic, which inflates the direct channel and hides the problem. GA4 offers data filters for that job, and a hostname filter that blocks data from domains not on an approved list for spam sent to a measurement ID from unrelated sites.

Exhaustive lists are not achievable for card authentication. 3-D Secure pages are hosted by card issuers and their processors, so the referring domain varies by the shopper's bank and changes over time. Analysts disagree on the remedy: some maintain long regular expressions of bank domains, while others prefer to set ignore_referrer on the order confirmation page, which handles any gateway but also suppresses a genuine referrer if a customer arrives there directly from an email or another site.

Excluding broad domains carries its own cost. A rule for paypal.com reclassifies every paypal.com visit, including real referral clicks from PayPal's own marketing surfaces. And practitioners dispute whether owned domains, such as a separate booking engine or a subdomain shop, belong on the list at all. Google's documentation directs owned domains to cross-domain measurement instead, because exclusion only hides the referrer; it does not carry the user's client identifier across, so GA4 still counts a second user.

Finally, the setting can only act on referrers the browser sends. Apps that open links in their own browsers, links marked "noreferrer" and strict referrer policies all produce visits with no referrer to exclude, and those arrive as direct regardless.

Cross-domain measurement links sessions across domains a company controls, appending a _gl parameter to links so the same client identifier persists. It solves user and session continuity; referral exclusion solves only source attribution.

Data filters in GA4 remove or flag events, for example internal or developer traffic, so they never reach reports. Referral exclusion keeps the event and changes its source.

Referral spam filtering targets fake hits; a hostname include filter or bot filtering is the tool, not the referral list.

Affiliate self-referral exclusion is a separate meaning. Affiliate platforms such as Rewardful and UpPromote use "exclude self-referrals" to stop affiliates earning commission on their own purchases, a fraud control unrelated to analytics attribution.

Recent developments

Changes in how referrers are sent have widened the gap between what the list can fix and what arrives as direct. When Google's AI Mode launched in the US in May 2025, its links stripped the referrer, and Google acknowledged the behaviour as a bug on May 23, 2025. On May 13, 2026, GA4 added an AI Assistant channel to its default channel group, creating a new referral category that analysts must take care not to exclude by accident with loose regular expressions. Google Gemini, meanwhile, stopped adding UTM tags to cited links after three weeks, leaving those clicks dependent on the referrer alone. As of November 2026, the unwanted referrals list remains a per-stream, forward-only setting, and Google has not announced retroactive reprocessing for it.


Timeline

  • December 2007 - Google releases ga.js for classic Google Analytics, including the _addIgnoredRef method for ignoring referrers in code.
  • October 2012 - Universal Analytics enters public beta.
  • April 2014 - Universal Analytics leaves beta; the Referral Exclusion List becomes an interface setting at property level.
  • October 14, 2020 - Google launches Google Analytics 4, with referral handling rebuilt at data stream level and the ignore_referrer parameter.
  • March 16, 2022 - Google announces the end of Universal Analytics.
  • July 1, 2023 - Standard UA properties stop processing data; UA referral exclusion lists do not carry over to GA4.
  • July 1, 2024 - Universal Analytics 360 properties are shut down.
  • May 23, 2025 - Google calls AI Mode's missing referrer data a bug.
  • December 2025 - A poll of 624 practitioners finds 59% misunderstand GA4 session attribution.
  • May 13, 2026 - GA4 adds the AI Assistant default channel.

Summary

Who: Site owners, analysts and agencies configure referral exclusions in tools such as Google Analytics 4, Adobe Analytics and Matomo; payment providers, banks and sign-in services are the domains most often excluded.

What: A list of referring domains that the analytics tool ignores, so visits from them are treated as having no referrer and cannot claim credit for sessions or conversions.

When: The idea dates to classic Google Analytics' _addIgnoredRef method from 2007, became an interface setting in Universal Analytics by 2014, and was rebuilt in GA4 after its October 2020 launch.

Where: In GA4, under each web data stream's tag settings, or through the ignore_referrer parameter in gtag.js or Google Tag Manager.

Why: Because the referrer header records the last site visited, payment and authentication redirects would otherwise take credit for sales generated by advertising and other channels, distorting attribution and budget decisions.