GA4 Source Group: fixing fragmented social attribution

GA4 Source Group is a new traffic-source dimension Google rolled out on June 11, 2026 that automatically consolidates the fragmented source strings a single social campaign scatters across your reports. One Facebook campaign could previously surface as facebook, fb, m.facebook.com, and more — each splitting your sessions, conversions, and CPA into separate, uncomparable rows.

For any team comparing channel performance in GA4, that fragmentation quietly corrupted the math. If half of a Facebook campaign’s sessions sat under l.facebook.com and the rest under facebook, neither row told you what the campaign actually delivered — and budget decisions built on those rows inherited the error.

This guide covers what Source Group actually does, the platforms it covers, how it differs from the existing Source Platform dimension, why its retroactive behavior matters, and a concrete audit checklist to verify it in your own property. Every fact below is sourced from Google’s announcement and the GA4 analyst community that has been pressure-testing it since launch.

Source Group is a traffic-source dimension that gives a consistent, single name to traffic from a given platform — even when the raw referral string varies. Google launched it on June 11, 2026 through the Google Analytics Help Center, where the announcement frames the rationale plainly: a single value for a given source is vital for performance and attribution. It is populated automatically, so most property owners do not need to touch their tagging at all.

The dimension surfaces natively in GA4 standard reports and in Explorations via the dimensions dropdown — no custom metric, no BigQuery export required for basic use. If you want to see where it fits among GA4’s wider taxonomy, our full GA4 dimensions reference maps every source-related dimension and its scope. One important nuance: as of this writing the rollout is incremental. Not every property had access at announcement, and Google’s typical global deployment window for incremental rollouts spans several weeks. The reliable check is to open the Explore dimensions picker in your own property and look for Source Group there.

Social platforms hand off traffic through a tangle of redirect and link-shim domains. Facebook alone previously generated at least five distinct source values in GA4 — facebook, fb, m.facebook.com, mobile.facebook.com, l.facebook.com, and Meta-facebook — depending on the device, app, and link wrapper involved. Each of those landed as its own row in your source/medium report.

The downstream damage is subtle but real. Sessions, conversions, and CPA for one campaign split across multiple rows, so no single row reflects the campaign’s true performance. Until June 11, analysts patched this by hand — building custom channel groups with regex patterns such as “Source contains facebook.com” to stitch the variants back together. Source Group eliminates that manual overhead as a system-level default.

The same pattern shows up across social and increasingly across AI referrers. ChatGPT traffic, for instance, can arrive as openai.com, chatgpt.com, or chat.openai.com — three strings, one source. That Google chose to fold ChatGPT and Perplexity into Source Group at launch is a quiet signal about where it expects referral traffic to grow, and it future-proofs your attribution for a channel that is only getting noisier.

The table below assembles the raw-to-grouped mapping for the major platforms Source Group covers, alongside the corresponding Source Platform label and the dimension scope. Facebook’s variant list is verified from analyst documentation; for the other platforms the raw strings are illustrative of the kind of fragmentation Source Group resolves rather than an exhaustive enumeration.

These two dimensions are routinely conflated, and the distinction is the most useful thing to internalize. Source Group identifies the originating platform — paid and organic combined — so all Instagram traffic reads as “Instagram.” Source Platform identifies the ad-buying ecosystem, so Instagram ad traffic reads as “Meta Ads.”

Put together, they hand you a paid-versus-organic split inside a single property with no custom regex. Instagram ad traffic maps to Source Group = “Instagram” but Source Platform = “Meta Ads,” while organic Instagram posts map to Source Group = “Instagram” with no Meta Ads platform value. Filter on the two together and you have clean paid and organic Instagram cohorts.

Google paired the Source Group launch with an update to Source Platform classifications, aligning the two and giving third-party platforms the same granularity that Google’s own inventory — YouTube and Search — already enjoyed. The two dimensions are designed to be read together, not as substitutes.

The narrative hook of this release is retroactivity. Source Group is applied to historical data, so your past Facebook traffic — once split five ways — consolidates backward into a single clean value. Year-over-year analysis has no date-break: the dimension reads consistently across the period before and after June 11.

That stands in deliberate contrast to GA4’s AI Assistant channel update earlier in 2026, which was forward-facing only and left the analytics community frustrated that historical data could not be relabeled. If you compared this quarter to last quarter on that earlier change, you hit a wall at the launch date. Source Group does not have that problem.

Source Group landed on June 11. Four days later, on June 15, 2026, the June 15 consent-split change made ad_storage the single gate between GA4 and Google Ads. Previously, Google Signals acted as a fallback for remarketing and Smart Bidding signals; after June 15 that fallback was removed. These are two distinct systems — Source Group is a reporting dimension, the consent split affects data collection for Google Ads — so they are complementary, not causally linked. But for paid-social teams, they interact in a way worth planning around.

The order of operations is the practical takeaway: clean up your source data so campaign performance reads accurately, then make sure your consent banner fires correctly so that data still reaches Google Ads. Cleaner source grouping means nothing if a misfiring consent banner drops your remarketing pool. The two changes belong on the same audit.

There was a third companion launch worth noting on the same day as Source Group. GA4 shipped a hostname filter on June 11 — found under Admin then Data Filters then Create Filter then “Web hostname traffic” — that excludes events from unapproved domains such as staging sites, ghost spam, and scrapers before they reach your reports. It is exclude-only at launch: you specify individual rogue domains, and no allowlist mode exists yet. Pair it with Source Group and you remove a second source of attribution noise: junk hostnames never enter the data in the first place.

Because Source Group needs no setup, the work is verification, not implementation. Run these five checks to confirm it is live in your property and to capture the attribution shift it surfaces.

If your team wants this verification run rigorously — confirming Source Group availability, rebuilding clean social reporting, and tying it to the consent-split audit — our analytics and measurement engagements start with exactly this kind of GA4 attribution health check. Cleaner source data also makes our paid media work and social media campaigns measurable against a budget that finally compares like with like.

The first-order effect is that channel comparisons become honest. For years, GA4’s raw source/medium report quietly understated social platforms by scattering them across redirect domains; budget decisions inherited that understatement. With Source Group as the default grouping, a Facebook line item finally reflects the whole Facebook campaign — and the relative ranking of your channels can shift once the fragments reassemble.

Looking forward, the inclusion of ChatGPT and Perplexity is the part to watch. AI-assistant referral traffic is still small for most properties, but it is growing and it is fragmented by nature — three domains for one assistant is already common. By folding these sources into the same grouping framework as established social platforms, Google is positioning GA4 to measure a channel that most attribution stacks cannot yet see clearly. Teams that build their reporting on Source Group now will have a continuous, retroactive record of AI referral growth as it accelerates, rather than scrambling to reconstruct it later.