ChatGPT vs Gemini vs Perplexity: Where AI Referral Traffic Is Going

AI referral traffic is overwhelmingly concentrated in ChatGPT, and the second place changed hands in 2026. Here are the current shares, the three nuances that change how you should read them, and what to actually do about it.

The current shares

Source: StatCounter Global Stats, March 2026.

Gemini’s rise is the notable movement: from 2.31% in April 2025 to 4.74% in December 2025, then 7.20% in January 2026, overtaking Perplexity for the first time. Roughly a quadrupling in under a year.

The nuance that changes how you read this

Share is not volume. The overall AI-referral pie is expanding quickly, so an engine can lose share while its absolute referral count rises. If Perplexity’s share fell but your Perplexity sessions grew, nothing is wrong. Always check absolute counts in your own analytics before reacting to a percentage.

Referrals are not visibility. Most AI answers are zero-click. A brand named in an answer nobody clicked still influenced the decision, and no referral will ever record it. Referral share measures the traffic slice of a much larger exposure picture, which is why presence-based metrics matter alongside it. See the AI visibility metrics that actually matter.

Referral share is not optimization difficulty. ChatGPT sends the most traffic, but its search leans heavily on Bing’s index, which is a completely different technical prerequisite from Perplexity’s own crawl. The engine that sends the most traffic is not automatically the easiest to win.

The crawl-to-referral gap

The most uncomfortable number in this area is how much AI companies read versus how much they send back. Cloudflare’s June 2025 snapshot measured crawl-to-referral ratios of roughly:

• 14:1 for Google
• 1,700:1 for OpenAI
• 73,000:1 for Anthropic

Those ratios have tightened since, and they vary by site and sector, but the direction is unambiguous: your content is being read far more often than it is being credited with a visit. That is the economic tension behind the entire blocking debate, which we work through in should you block AI crawlers.

Quality, not just quantity

Raw session counts undersell this channel. AI-referred visitors frequently arrive later in the buying cycle, because the assistant has already done the comparison and shortlisting work before the click.

Do not take that on faith, and do not take it from a vendor. Test it on your own data: build the AI referral segment and compare engagement rate, pages per session, and conversion rate against your organic baseline. If AI referrals convert better, that changes the value of the channel materially even at low volume. The setup is in tracking AI traffic in GA4 and Search Console.

B2B looks different from the whole market

Aggregate numbers describe the whole web, and your category may not resemble it. Professional and B2B audiences skew differently across assistants than a general consumer population, and vertical differences are large enough that a market-wide share table is a poor planning input on its own.

The practical response is to measure your own mix rather than adopting anyone’s averages, including these.

What to actually do about it

1. Track all four surfaces, not just the leader: ChatGPT, Gemini, Perplexity, and Google’s AI Mode. They retrieve from different indexes and select differently, so a single blended number hides a total absence on one.
2. Prioritize by your own data, not by market share. If your analytics show Perplexity sending qualified traffic, that matters more than its global percentage.
3. Fix the shared prerequisites first. Crawler access, rendering, and indexation serve every engine. Bing indexation disproportionately serves the largest one.
4. Measure presence separately from referrals, because the zero-click majority never becomes a session.
5. Re-check quarterly. These shares moved substantially inside twelve months and there is no reason to expect stability.

The takeaway

ChatGPT dominates AI referrals, Gemini is climbing fast and has taken second, and Perplexity punches above its referral share on citation visibility because it cites on nearly every answer. But the ranking is less useful than it looks: shares move, referrals capture only the clicked minority of exposure, and your own category may not match the aggregate at all.

Track every engine, measure presence alongside traffic, and let your own data set the priority. To see how each engine currently describes and cites your brand, request an AI visibility audit.