Meet an Expert

Gemini vs Perplexity Referrals, Why GA4 Hides the Difference

Gemini is closing in on Perplexity for referral share. Here is why the mechanism differs and how to track Gemini clicks separately in GA4.

Lior Eldan Lior Eldan
Oct 30, 2026 7 min read Engines

A travel brand we looked at recently saw its “AI referral” channel in GA4 climb 40% quarter over quarter and assumed Perplexity was driving it, since Perplexity had been the loudest AI referrer for most of 2024. The breakdown said otherwise. Gemini had overtaken Perplexity in referral share, and the two were sending traffic for completely different reasons. Lump them into one line item and you lose the ability to act on either.

Key Takeaways

  • Gemini and Perplexity send referral traffic through different mechanisms, so a rising “AI referral” total can mask which engine is actually responsible.
  • Gemini draws on Search, YouTube and Scholar together, meaning a referral increase there often reflects broader web and video presence rather than inline citation quality.
  • Default GA4 channel groupings treat most AI referrers as generic traffic or search, so separating Gemini requires a custom channel group built on referrer domain.

Why Gemini and Perplexity Are Not the Same Referral Source

Perplexity answers with inline citations by design. Every response carries numbered source links, and clicking one is the main way a user leaves the page at all. Referral traffic from Perplexity is, in effect, baked into the product.

Gemini works differently. It cross-references Google Search, YouTube and Scholar, and its referral clicks come from a mix of surfaces: answer boxes, “sources and related content” panels, and Search integration points that do not always look like a citation in the way Perplexity’s do. A brand gaining Gemini referral share is not necessarily winning on answer quality alone. It may be winning because its YouTube presence, structured data or Scholar-indexed content gave Gemini more to pull from across formats Perplexity does not touch.

That distinction matters for where you invest. If Perplexity sent the clicks, the lever is citation-worthy, well-sourced content. If Gemini sent them, the lever might be video, structured markup or Search-adjacent authority signals that have nothing to do with how a Perplexity answer gets written.

What the Overtake Actually Looks Like in the Data

The pattern we have seen across clients is not that Perplexity referrals are falling. It is that Gemini referrals are growing faster, often because Gemini’s reach extends further into a user’s existing Google habits. A Gemini answer surfaced inside Search has a much larger audience than a standalone Perplexity session, even if the per-query citation rate looks similar.

This is a Share of Citation question, not a Share of Voice question. A brand can be mentioned constantly across the web and still have near-zero Share of Citation if no model is naming it as a source. Referral share tracks the sharper of the two: it only counts when a model trusted the brand enough to link out.

42,435
AI citations tracked for Moburst’s own brand, with 309 unique pages cited and an average position of 1.77 when cited.See the case study

That volume only shows up as useful if you can see which engine is responsible for which pages. A single “AI mentions” total tells you citations are compounding. It does not tell you whether Gemini or Perplexity is the one compounding them, and the two require different content to keep the streak going.

An aggregate AI referral number tells you traffic is coming. It does not tell you which engine trusts you enough to keep sending it.

Why a Single “AI Referral” Bucket Hides the Decision You Need to Make

Most GA4 setups still route AI referrers one of three ways: as unassigned referral traffic, as organic search if the domain happens to match a pattern, or into a generic “Other” channel nobody checks. None of these separate Gemini from Perplexity from ChatGPT, which means the monthly report shows one number moving without showing which lever moved it.

This becomes a real problem once you are trying to brief content or PR work. Perplexity referral growth points toward citation-worthy source material and structured, current content. Gemini referral growth points toward a wider mix: YouTube, Scholar-indexed research, and Search-adjacent authority. If your dashboard cannot tell these apart, you end up briefing the wrong team for the wrong result.

Setting Up Tracking That Actually Separates Them

The fix is a custom channel group in GA4 built on referrer hostname, not on GA4’s default channel logic, which was not built with AI assistants in mind.

  • Build a referrer-based channel group. Under Admin, create a custom channel group with rules matching source contains gemini.google.com for Gemini, source contains perplexity.ai for Perplexity, and separate rules for chat.openai.com, claude.ai and x.com or grok referrers. Order the rules so the most specific match wins.
  • Add a secondary dimension for landing page. Pair the new channel group with landing page in your reports so you can see which pages each engine is actually sending clicks to, not just how many clicks arrive.
  • Tag outbound links where possible. If you control UTM parameters on content you expect models to cite (pricing pages, comparison pages, documentation), a consistent utm_source value gives you a second, more reliable signal than referrer header alone, since some AI clients strip or rewrite referrer data.
  • Set up an exploration report, not just a standard report. A free-form exploration with channel group as the row dimension and sessions, engaged sessions and conversions as metrics lets you compare Gemini against Perplexity side by side, rather than reading them off separate default reports.
  • Re-check the rules quarterly. Referrer domains and query parameters for AI assistants change as the products iterate. A rule built in January may stop matching by summer if a vendor changes its link format.

None of this requires a new tool. It requires GA4’s existing channel grouping feature pointed at the right referrer strings, something most default implementations never set up because AI referrers did not exist when the property was first configured.

What To Do Next

Pull your GA4 referral report for the last 90 days and check whether AI traffic is still sitting inside “Unassigned” or a generic “Other” channel. If it is, build the custom channel group described above before you draw any conclusion about which engine is winning.

Once Gemini and Perplexity are separated, look at landing pages next to each one. If Gemini is sending clicks to pages with video embeds or Scholar-style citations and Perplexity is sending clicks to pages with tight, current sourcing, you have your answer on where each engine’s next click is coming from, and which content type to produce more of.

FAQs

Does GA4 Track Gemini Referral Traffic by Default?

Not as a distinct channel. GA4’s default channel grouping was not built to recognize AI assistant referrers, so Gemini traffic typically lands in Unassigned, Referral, or occasionally Organic Search depending on how the link was formatted. A custom channel group based on referrer hostname is required to isolate it.

Why Would Gemini Referral Share Grow While Perplexity’s Mention Count Stays Flat?

Gemini draws on Search, YouTube and Scholar together, so referral growth there can come from video or Scholar-indexed presence rather than from answer-text citation quality. Perplexity’s referral behavior is tied almost entirely to inline citations, so its growth depends on being named as a direct source inside the answer itself.

Is Referral Share the Same as Share of Citation?

They are related but not identical. Share of Citation measures how often a model names a brand as a source at all. Referral share measures how often that citation turns into an actual click. A brand can have strong Share of Citation and still see weak referral share if the citation does not carry a compelling link or context.

Should UTM Tagging Replace Referrer-Based Tracking?

No, use both. Referrer header data catches traffic you have no control over, including links a model generates on its own. UTM tagging is more reliable when you can influence the link an AI assistant surfaces, such as on pricing or documentation pages, but it will never cover every citation path.

Lior Eldan

About the author

Lior Eldan

Lior Eldan is the Co-Founder of Moburst and serves as its COO. He works at the intersection of marketing, AI and growth, helping brands' teams adapt to AI-driven discovery and decision-making through data-informed strategy and systems thinking.

Meet the whole team

Ready to make your brand the answer?

Our growth strategists take brands from overlooked to recommended. In one call you will see how the AI engines read your brand today, and leave with your own roadmap to dominate the new era of search.