Meet an Expert

Gemini Overtakes Perplexity in Referral Traffic Share

Gemini overtaking Perplexity in referral share changes attribution. Here is how to separate Gemini clicks from generic AI referral totals.

Lior Eldan Lior Eldan
Oct 29, 2026 6 min read Engines

A SaaS marketing lead opens her GA4 referral report one morning in March. Gemini is now outsending Perplexity. Six months back, Perplexity owned the entire AI referral conversation on her dashboard, no contest. The reversal isn’t a footnote for a monthly recap. It changes who gets credit for pipeline, and most referral setups still dump both engines into one bucket called “AI traffic.” The shift is invisible unless someone goes looking for it.

Key Takeaways

  • Gemini and Perplexity drive clicks through different mechanics, so a rising Gemini share does not mean the same thing as a rising Perplexity share.
  • Default GA4 channel groupings often misclassify Gemini-driven sessions as organic Google search instead of AI referral, because Gemini surfaces inside Google’s own properties.
  • Separating engines in analytics requires isolating referrer domains by regex, not relying on a single blended “AI referral” segment.

What a Referral Share Shift Actually Tells You

Referral share counts clicks. Not mentions. Not citations. Not how often a model name-drops a brand mid-answer. A company can get cited in a Gemini response a thousand times a day and still see zero sessions if the answer already gave the user what they needed. Referral share is the blunt, unglamorous number underneath all of that: someone left the AI interface and landed on an actual page.

So when that number tilts from Perplexity toward Gemini, what’s really happening? It’s mostly about where people start their searches, and how willing each engine is to hand them off. Perplexity’s whole interface is built around visible, numbered citations. Click one, you’re gone. Gemini has no equivalent habit baked in. A rising Gemini share usually just means Gemini’s query volume is outgrowing Perplexity’s, not that Gemini suddenly got generous with outbound links.

Why the Two Engines Send Traffic So Differently

Perplexity puts its sources in a numbered list, right there, clickable without digging. Users have learned to click it. That behavior is baked into the product. Gemini doesn’t work that way. It pulls from Search, YouTube, and Scholar, and a lot of its links surface inside Google’s AI Overviews rather than on a dedicated answer page with its own citation trail.

Here’s why that matters for tracking: AI Overviews live on google.com. A click from one can look, in a referral report, nearly identical to a click from a plain organic result. Same referrer domain. Often the same medium bucket. Unless someone deliberately pulls AI Overview sessions out of standard organic sessions, Gemini’s real contribution gets swallowed by “Organic Search” and vanishes from the AI conversation entirely.

1.59
average rank NewDay USA achieved on Google AI Overviews, tracked weekly across five engines over six months.See the case study

A 1.59 average rank on AI Overviews is a visibility number, nothing more. It doesn’t tell you how many of those impressions turned into sessions. You only find that out once the referral data gets pulled apart from standard organic traffic first. Visibility and clicks are not the same currency, and treating them as interchangeable is where most tracking breaks down.

The Blended Metric Problem

Most analytics setups treat “AI referral” as one channel. One segment, summing sessions from chatgpt.com, perplexity.ai, gemini.google.com, maybe Copilot or Claude’s web interface thrown in too. It looks clean on a dashboard. It also hides exactly the shift this piece opened with.

Say Gemini referrals climb while Perplexity referrals drop by roughly the same amount. The blended total barely moves. Anyone looking only at that combined line concludes nothing changed. Meanwhile the content strategy built for Perplexity’s citation-driven clicks might be doing nothing for Gemini’s Search-and-YouTube discovery path. The blended number never shows that gap. It can’t.

Referral share tells you who clicked. A blended AI total tells you nothing about why.

Building Gemini-Specific Tracking in GA4

Skip the pre-built channel groupings. Start with the session source dimension instead. In a GA4 Exploration report, build a free-form table: Session source as the primary dimension, Sessions and Engaged sessions and Conversions as metrics. Filter with regex that isolates each AI referrer on its own row: gemini.google.com, perplexity.ai, chatgpt.com, claude.ai. No summing them together.

Gemini is the hard one. Sessions arriving through an AI Overview often carry source=google and medium=organic, identical on paper to a plain search click. To pull them apart, watch landing page behavior alongside the referral data. AI Overview traffic tends to cluster around specific query patterns and spike in short bursts tied to when a page gets cited inside an overview. Cross-checking server logs against GA4 is really the only reliable way to confirm a spike came from Gemini and not from an ordinary ranking fluctuation that week.

Save a segment named explicitly for Gemini. Keep it separate from a segment labeled general Google organic. Repeat for Perplexity, ChatGPT, whatever else is worth tracking. Review them side by side on a set cadence. Don’t merge them back the moment the report gets built, that defeats the whole point.

What the Separated Data Actually Buys You

Once the engines are split apart, the useful question is what each one rewards. If Gemini pulls heavily from YouTube and Scholar, then video content and well-documented research matter more there than they do for Perplexity, which leans on open-web pages it can cite inline in real time.

Look at conversion rate and engaged session rate per engine, not just raw session counts. A smaller batch of Perplexity clicks converting at twice Gemini’s rate should change where content effort goes next quarter. This is also where Share of Citation and Share of Voice stay distinct from referral share. A brand can get cited constantly by Gemini and still see almost no resulting clicks, because Gemini’s answer already satisfied the question on the spot. Separated referral tracking is what exposes that gap. Assuming citation equals traffic is the mistake that keeps teams chasing the wrong number.

What to Do This Week

Pull the current GA4 channel grouping. Check whether Gemini, Perplexity, ChatGPT, and Claude show up separately or get blended into one AI referral line. If they’re blended, rebuild it using session source regex instead of the default grouping. This takes an afternoon, not a sprint.

Set a four-week baseline per engine: sessions, engaged sessions, conversions. Revisit it monthly. Watch for the kind of share shift described above instead of waiting for the combined total to budge, because a blended number moves slowly and by the time it does, the window to act on it has mostly closed.

FAQs

Why does Gemini referral traffic sometimes show up as organic Google traffic?

Because AI Overviews live inside google.com, clicks from them often carry the same source and medium values as a standard organic search result. Without separating landing page and timing patterns, GA4 cannot distinguish the two automatically.

Is Share of Citation the same as referral traffic?

No. Share of Citation measures how often a model names a brand as a source inside its answer. Referral traffic measures how many users actually clicked through to the site afterward. A brand can score high on one and near zero on the other.

Do I need separate UTM parameters for each AI engine?

UTM parameters only work if the linking party adds them, and AI engines generally do not. Isolating engines requires filtering the session source dimension by referrer domain instead, since UTM tagging is not under the brand’s control here.

How often should referral share by engine be reviewed?

Monthly is a reasonable baseline for most mid-market teams, with a closer weekly check during any period where one engine’s share appears to be moving quickly, as happened with Gemini and Perplexity.

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.