Skip to main content
METRIC VAULT
← All posts

Which AI Engine Actually Sends You Paying Visitors

ChatGPT, Perplexity, and Gemini all drive referrals now — but they don't convert equally. Here's what GA4 reveals about which AI engine sends buyers, not browsers.

·

A marketer intensely studying an upward-trending line chart on a monitor at her late-night desk, lit by blue screen glow amid cold tea and crumpled notes.

Perplexity usually sends the highest-converting AI referral traffic, because its citation-driven visitors are mid-funnel researchers. ChatGPT drives the most total conversions on volume alone, while Gemini is mainly a top-of-funnel reach channel that GA4 often undercounts. Set up a custom AI channel group and track conversion value to confirm it for your own site.

Every performance marketer has watched the same chart shift over the past year. A referral source called chatgpt.com showed up in GA4. Then perplexity.ai. Then Gemini began leaking clicks through Google properties. The question is no longer whether AI engines send traffic — it's which ones send visitors who open their wallets.

This post compares ai referral traffic from the three engines that matter most to Western marketers right now, using GA4 evidence rather than vibes. We'll look at volume, engagement, and the part most reports skip: actual conversion behavior.

Why AI referral traffic needs its own analysis

Treating AI referrals as one lump is the first mistake. Someone who clicks a citation inside Perplexity behaves nothing like someone who copies a ChatGPT answer and never clicks at all. The intent, the context, and the point in the funnel are all different.

Three things set AI referral traffic apart from classic organic search:

  • The answer came first. Users arrive already informed, often further down the funnel than a cold Google searcher.
  • Click-through is rarer. Most AI answers resolve the query in-engine, so the people who do click are self-selected and high-intent.
  • Attribution is messy. Some engines pass clean referrers, others strip them, and some route through redirect domains.

That last point is why your raw numbers almost certainly undercount the real impact. Before you compare engines, get your tracking in order.


Setting up GA4 AI tracking that you can trust

Good ga4 ai tracking starts with pulling AI sources out of the generic "Referral" bucket. Out of the box, GA4 lumps chatgpt.com and perplexity.ai in with every other referring domain. You want them grouped and visible.

Build a custom channel group

In GA4, go to Admin → Data settings → Channel groups and create a group with an "AI Assistants" channel. Match on the referrer domains:

Source matches regex:
chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|.*\.bing\.com/chat

Add a UTM-free explore for session source

Many AI engines link without UTM parameters, so lean on the Session source / medium dimension in an Exploration. Filter to the AI domains above and pull in:

  • Sessions
  • Engaged sessions
  • Average engagement time
  • Key events (conversions)
  • Session key event rate

Watch for the referrer-stripping problem

Gemini is the worst offender here. Because it lives inside Google's ecosystem, a chunk of its clicks show up as google / organic or as direct traffic rather than a distinct AI referrer. If your Gemini numbers look suspiciously tiny, this is why — not because Gemini sends no one.

Once the plumbing is reliable, the comparison gets interesting.


The volume picture: who sends the most clicks

Raw session counts are the least useful metric, but they set the stage. Across most B2B and content-driven sites, the pecking order for click volume looks like this:

EngineRelative click volumeReferrer cleanlinessTypical user intent
ChatGPTHighClean (chatgpt.com)Mixed — research to transactional
PerplexityMediumClean (perplexity.ai)Research-heavy, citation-driven
GeminiLow (undercounted)Poor — often maskedBroad, consumer-leaning

ChatGPT wins volume largely on scale — it simply has more users. But volume is a trap. The marketer's real question is what happens after the click. That's where the engines separate.


Engagement quality by engine

Engagement time and pages-per-session are your first proxies for visitor quality. Here the ranking often flips.

Perplexity traffic tends to engage deepest

Perplexity traffic consistently shows the strongest engagement metrics in GA4 across a wide range of sites. The reason is structural: Perplexity surfaces multiple cited sources and nudges users to click through and verify claims. People who click a Perplexity citation are actively researching — comparing vendors, checking specs, validating a recommendation.

In practice that shows up as:

  • Higher average engagement time (often 1.5–2x ChatGPT referrals)
  • More pages per session
  • Lower bounce on pricing and comparison pages

ChatGPT referrals are broader and shallower

ChatGPT sends more people, but a larger share arrive casually. Many clicks come from exploratory conversations rather than purchase research. Engagement is solid but not spectacular, and it swings widely depending on the type of query that surfaced your link.

Gemini is the wildcard

Because Gemini traffic is partly misattributed, clean engagement comparisons are hard. The clicks you can isolate tend to be consumer-flavored and shorter — more "quick answer" than "deep research."


The part that matters: who converts

Engagement is a means to an end. The end is revenue. When you map ai referral traffic to key events — signups, demo requests, purchases — a clear hierarchy emerges for most commercial sites.

EngineSession key event rateRevenue per sessionBest-fit funnel stage
PerplexityHighestHighMid-to-bottom (comparison, evaluation)
ChatGPTModerateModerateTop-to-mid (discovery, research)
GeminiLow / noisyLowTop (broad awareness)

Perplexity punches far above its volume. It sends fewer visitors than ChatGPT, yet a meaningfully higher percentage of them complete a key event. If you sell a considered product — software, services, anything with a comparison phase — Perplexity referrals often convert at rates that rival or beat your branded search.

ChatGPT earns its keep on absolute numbers. Even with a lower conversion rate, its sheer volume means it frequently drives the most total conversions. For many sites, ChatGPT is the biggest contributor by raw count and Perplexity is the most efficient per visitor.

Gemini, once you correct for undercounting, usually lands as a top-of-funnel awareness channel. Valuable for reach, weak for immediate conversion.


How to verify this on your own site

Don't take benchmarks as gospel — your mix depends on your niche and content. Run this three-step check in GA4.

1. Compare engagement side by side

Build an Exploration with Session source/medium as the row dimension, filtered to your AI channel. Add engagement time and session key event rate as values. Sort by key event rate, not sessions.

2. Attach a conversion value

If you have ecommerce or a value assigned to key events, add Total revenue and Revenue per session to the same table. This instantly reveals which engine sends buyers versus browsers.

3. Segment by landing page

Create a comparison that pairs AI source with landing page. You'll usually find Perplexity clicks land on comparison and "best X for Y" pages, while ChatGPT clicks spread across educational content. That landing-page pattern tells you which content earns high-intent AI citations.

Exploration setup:
Rows: Session source/medium, Landing page + query string
Values: Sessions, Engaged sessions, Session key event rate, Revenue per session
Filter: Session source matches AI channel regex

Turning the data into strategy

Once you know which engine sends paying visitors, you can act instead of just reporting.

Double down on Perplexity-friendly content

Perplexity rewards content that is citable: clear claims, specific numbers, structured comparisons, and sources. If Perplexity is your highest-converting AI channel, build more comparison tables, specification roundups, and direct-answer sections that are easy to cite.

Capture ChatGPT's volume higher in the funnel

ChatGPT sends researchers, so give them a next step. Strong educational content with clear CTAs and email capture turns a research-stage ChatGPT visitor into a tracked lead you can nurture — rather than a one-and-done session.

Don't write off Gemini — fix your tracking first

Before you decide Gemini is worthless, confirm you're not losing its clicks to the direct and organic buckets. Improving your ga4 ai tracking may reveal that Gemini is quietly contributing more than your default reports suggest.


The honest caveats

A few things to keep in mind so you don't over-read early data:

  • Sample sizes are often small. AI referral volumes are growing fast, but they can still be too thin for statistical confidence in a given month. Look at trends, not single weeks.
  • Attribution windows matter. AI visitors often research now and buy later, sometimes via a branded search. Last-click models will undercredit AI's real influence.
  • The engines change monthly. Citation behavior, link formats, and referrer handling all shift. Re-run this analysis quarterly.

The bottom line

If you want the single most efficient source of ai referral traffic, Perplexity is usually the winner — its citation-driven, research-stage visitors convert at the highest rate. If you want the most total conversions, ChatGPT's scale typically wins on absolute numbers. Gemini is a reach play whose real contribution stays hidden until your tracking is clean.

The practical move isn't to pick one engine. It's to measure all three properly in GA4, match each to the funnel stage it serves best, and build content that earns citations where your buyers actually click through. Set up the channel group, assign conversion values, and let your own revenue data settle the argument.