ChatGPT, by a wide margin. SE Ranking’s June 2026 study found it drives 74.78% of AI referral traffic to websites, against 11.56% for Gemini, 7.23% for Perplexity, 3.51% for Copilot and 2.62% for Claude. But volume is not value. The smaller engines send visitors who convert well, and Google’s AI surfaces send clicks you cannot isolate at all.
Every AI visibility conversation eventually reaches the same practical question. If there are five or six assistants worth caring about and a limited budget, which one actually puts visitors on your site?
The answer depends on which number you look at. Measured by volume, it is not close. Measured by visitor quality, the order reshuffles. Measured by audience size, the largest AI surface of all barely appears in referral data. This AI referral traffic comparison walks through all three, with the sources and the caveats, and finishes with how to see your own split. It builds on our complete guide to AI search visibility.
1. The traffic split, platform by platform
SE Ranking’s AI traffic research, published in June 2026, measured how AI referral traffic to websites divides between platforms. It is one of the most lopsided distributions in digital marketing.
Together those five account for 99.7% of the AI referral traffic in the study. ChatGPT alone sends roughly three times as much as the other four combined.
These are shares of referral traffic: visits that reach a website from an AI platform. They are not shares of usage, and they say nothing about how often each engine mentions you. A platform can be widely used and still send few referrals if its answers rarely include links, or if its clicks arrive under another label.
That second case matters a great deal for Google. A click from an AI Overview or an AI Mode answer reaches your site from google.com, so analytics counts it as Google organic search, not as an AI referral. The 11.56% attributed to Gemini refers to the Gemini assistant, not to Google’s AI features inside Search.
Keep the totals in proportion as well. By the same study, AI platforms sent 0.32% of all web traffic in 2026, up 16 times from 2024. Multiply ChatGPT’s share by that total and the clear market leader accounts for roughly a quarter of one percent of all web visits. Conductor, measuring differently across 13,770 domains, found AI referrals averaging 1.08% of a site’s sessions. The two figures use different bases, an average per site versus a share of all traffic, but they agree on the shape: real, growing fast, still small.
2. Why ChatGPT is so far ahead
Two reasons, one about audience and one about mechanics.
Audience. OpenAI reported 900 million weekly active users for ChatGPT in February 2026. A large share of what people do there is research and shortlisting, which is exactly the behavior that ends in a website visit.
Mechanics. ChatGPT cites its sources as clickable links and tags many of them with utm_source=chatgpt.com. That tag, together with the referrer, is how analytics tools recognize the visit. A platform whose clicks arrive cleanly labeled gets counted. A platform whose clicks arrive unlabeled is easy to undercount. That does not explain a lead this large, but it is worth remembering whenever the gap between engines looks enormous.
Google’s AI surfaces have the largest audiences in AI search: 2.5 billion monthly users for AI Overviews and more than 1 billion for AI Mode, per Google I/O in May 2026. ChatGPT reports weekly users rather than monthly, so those audiences are not directly comparable. The click behavior is the telling part. Semrush found 92 to 94% of AI Mode sessions end without a click to any website.
The biggest AI audience is also the one least likely to click, and the clicks it does send are credited to Google search. That is why it barely registers in a referral comparison like this one, and why ignoring it would be a mistake. We covered the traffic effect in what Google AI Mode means for your traffic.
3. Volume and value are different questions
Share tells you where visits come from. It does not tell you what those visits are worth. Seer Interactive measured conversion rates for one client between October 2024 and April 2025, and the order changes. This is a case study, not an industry-wide benchmark.
Perplexity is the clearest example. It sends a small fraction of AI referrals, but in Seer’s data its visitors converted at 10.5%, roughly six times the rate of Google organic at 1.76%. ChatGPT converted at 15.9%. Even Gemini, the lowest of the AI sources measured, converted at 3%, still well above organic search.
The mechanism is the same across all of them. A visitor who arrives from an AI answer has usually had the question answered, compared options and seen your business named before they click. They arrive closer to a decision than a cold searcher does.
The samples were small. Seer’s study dates from June 2025, when AI traffic volumes were far lower than today, and the sample sizes for AI sources were small. Treat the conversion figures as evidence of direction, not as rates to forecast with.
Do not combine the two studies. The share figures come from SE Ranking in 2026 and the conversion figures from Seer a year earlier, measured on different sites. Multiplying one by the other to produce a “value per platform” would look precise and mean nothing.
4. What each platform means for you
| Platform | Share of AI referrals | What it means for you |
|---|---|---|
| ChatGPT | 74.78% | The volume leader and the engine to test first. Its links are usually tagged, so its traffic is the easiest to see. Highest conversion rate in Seer’s data, at 15.9%. |
| Gemini | 11.56% | The Gemini assistant, not Google’s AI features in Search. A distant second for referrals, with the lowest conversion of the AI sources Seer measured, at 3%. |
| Perplexity | 7.23% | Small volume, high intent. Its answers are built around visible, numbered sources, and Seer measured its referrals converting at 10.5%. Worth watching even when the session counts look trivial. |
| Copilot | 3.51% | Microsoft’s assistant, drawing on Bing. Low share, and included among the examples in Google Analytics’ current AI Assistant channel documentation; verify its source and medium in your property. |
| Claude | 2.62% | The smallest of the five by referrals. Seer measured 5% conversion. One of the three assistants Google named when it launched GA4’s AI Assistant channel. |
| Google AI Overviews and AI Mode | Not counted as AI referral | The largest audiences in AI search, but their clicks arrive as Google organic. Use Google Search performance reporting alongside analytics; these clicks are not standalone AI referrals. |
5. The AI traffic you will never see labeled
Referral reports show the clicks that arrive with a recognizable source. A large share of AI influence never does.
Google Analytics introduced an AI Assistant channel in May 2026. Its current default-channel documentation includes sources such as ChatGPT, Gemini, Deepseek, Copilot and Grok, while excluding Google AI Overviews and AI Mode. Review the current channel definition and your Session source / medium report rather than relying on a fixed list of assistant domains.
The channel has two hard limits. It only catches clicks that carry a referrer, so anything arriving without one still lands in Direct. And it cannot see Google’s own AI surfaces, whose clicks arrive from google.com.
How fragile the first limit is became obvious in October 2026. Search Engine Roundtable reported on October 14 that ChatGPT links were opening without a referrer and without the utm_source tag, so visits were landing in Direct. Adobe’s Cedric Huesler, who flagged it, updated his post the next day to say it had been a bug and that tagging had resumed. For roughly a week, starting October 9 by Evertune’s count, the largest source of AI referrals partly vanished from many sites’ reports, while nothing about real visitor behavior had changed.
Then there is everything that produces no click at all. Similarweb found in June 2026 that brands appearing in ChatGPT recommendations were 2.5 times more likely to receive a site visit within seven days. Many of those visits arrive later, through a branded search or a typed address, and get credited to Organic Search or Direct. Search Console will not fill the gap for the non-Google engines either, because it reports on Google’s surfaces only, as we covered in how to read the Search Console AI visibility report.
Your AI referral line is a floor, not a total. Comparing platforms by referral share tells you which engine sends the most labeled clicks. It does not tell you which engine influences the most buyers. For that you need prompt testing alongside analytics, and branded search volume as a running proxy.
6. So which one should you optimize for?
Mostly, you should not choose. The inputs that get you cited are largely the same on every engine: independent sources that name you, content that is easy to quote, consistent business details, and a sound SEO foundation underneath. We covered those in why your competitors get named in AI answers and you don’t.
What differs is where each engine looks, and that is the case for testing them separately rather than optimizing for them separately. Ahrefs found only 6.82% of ChatGPT results overlap with Google’s top 10, and that 28.3% of ChatGPT’s most-cited pages have no organic visibility in Google at all. Even Google’s own surfaces disagree: AI Overviews and AI Mode shared only 13.7% of cited URLs in Ahrefs’ December 2025 analysis. Ranking well on Google helps everywhere, but it covers far less of ChatGPT’s source pool than it does Google’s.
A sensible order of attention for most businesses:
- ChatGPT first. The largest referral source, the cleanest attribution, and the engine most likely to cite pages Google does not rank. If you test one engine, test this one, using the method in how to check if ChatGPT recommends your business.
- Google AI Overviews and AI Mode second. Not because they send countable referrals, but because their audiences are the largest in search and their answers sit on top of the results you already depend on.
- Perplexity third. Small, but its visitors convert, and its visible citations make it a useful read on which sources your category trusts.
- Gemini, Copilot and Claude in the same test set. Include them in every prompt run. Do not build anything specific for them.
If a proposal includes separate optimization packages for individual AI engines, ask what the work does differently for each one. By SE Ranking’s 2026 figures, Perplexity accounted for 7.23% of AI referrals, and AI referrals for 0.32% of all web traffic. That is about two visits in every ten thousand. Those visitors may well be valuable, but the work that wins them is the same work that wins ChatGPT.
7. How to see your own split
You do not need a paid tool for a first read. In GA4:
- Open Reports, then Acquisition, then Traffic acquisition. Set the dimension to Session default channel group and look for the AI Assistant row. Switch the dimension to Session source to see it broken out by platform.
- Check Referral and Direct for what the channel misses. Review assistant domains in Session source / medium and check whether any visits fall outside AI Assistant. If ChatGPT traffic dipped in mid October 2026, check whether Direct rose in the same days before concluding anything.
- Keep a custom channel group as a backstop. If you built one before May 2026, leave it running. A session source match like the one below catches the main assistants regardless of what Google’s list includes.
Then read it alongside branded search volume and tracked calls, month over month. The trend matters more than the starting number, and on most sites a single month of AI referral data is too small to draw conclusions from.
See which engines name you, not just which ones click
We test 50 real buyer prompts across five AI engines, record who gets named in your market and which sources each answer cites, then set that against your analytics. You get the full picture: the clicks you can see and the recommendations you cannot.
Get an AI visibility auditFrequently asked questions
Which AI platform sends the most traffic to websites?
ChatGPT, by far. SE Ranking’s June 2026 study found it drives 74.78% of AI referral traffic, followed by Gemini at 11.56%, Perplexity at 7.23%, Copilot at 3.51% and Claude at 2.62%. Together those five account for 99.7% of AI referrals. Clicks from Google AI Overviews and AI Mode are not included, because they arrive as Google organic search.
Does Google AI Mode count as AI referral traffic?
No. A click from an AI Overview or an AI Mode answer reaches your site from google.com, so Google Analytics records it as Organic Search, and GA4’s AI Assistant channel does not include it. Search Console reports Google Search performance; it does not measure referrals from standalone AI assistants. That is why the largest AI audiences barely appear in referral comparisons.
Which AI traffic converts best?
In Seer Interactive’s June 2025 data, ChatGPT referrals converted at 15.9%, Perplexity at 10.5%, Claude at 5% and Gemini at 3%, against 1.76% for Google organic. Every AI source measured beat organic search. The sample sizes for AI sources were small at the time, so treat the figures as a direction, not a forecast.
How do I see AI referral traffic in Google Analytics?
In GA4, open Reports, then Acquisition, then Traffic acquisition, and look for the AI Assistant row in the default channel group. Google added that channel on May 13, 2026, and named ChatGPT, Gemini and Claude as covered. Use Session source / medium to check individual assistant domains, and retain a custom AI channel group as a comparison if you already use one.
Why did my ChatGPT traffic drop in October 2026?
Possibly a tracking bug, not a real decline. From October 9, 2026, ChatGPT links opened without a referrer or utm_source tag, so visits landed in Direct. Search Engine Roundtable reported it on October 14, and Adobe’s Cedric Huesler confirmed on October 15 that it was a bug and tagging had resumed. If Direct rose on the same days your ChatGPT referrals fell, that is the likely cause.
Should I optimize differently for each AI engine?
Mostly no. The inputs that earn citations are largely shared: independent sources that name you, quotable content, consistent business details and solid SEO. What differs is where each engine looks. Ahrefs found only 6.82% of ChatGPT results overlap with Google’s top 10, so test each engine separately with real buyer prompts, then fix the gaps those tests reveal.
Sources
- SE Ranking, AI traffic research study, June 2026 (platform share of AI referral traffic; 16x growth from 2024 to 2026; 0.32% of all web traffic).
- Seer Interactive, conversion rate by traffic source, June 2025 (ChatGPT 15.9%, Perplexity 10.5%, Claude 5%, Gemini 3%, Google organic 1.76%).
- Conductor, AI referral traffic share study across 13,770 domains (1.08% of sessions).
- Similarweb, ChatGPT recommendation and site visit correlation, June 2026 (2.5x within seven days).
- OpenAI, February 2026 (ChatGPT 900 million weekly active users).
- Google I/O keynote, May 2026 (AI Mode 1 billion monthly users, AI Overviews 2.5 billion).
- Semrush, AI Mode zero-click rate research (92 to 94% of sessions ending without a click).
- Ahrefs, ChatGPT and Google top 10 overlap analysis (6.82% overlap; 28.3% of most-cited pages with no organic visibility).
- Ahrefs, AI Overviews versus AI Mode URL overlap, December 2025 (13.7%).
- Google Analytics Help, “What’s new”, AI Assistant channel added to the Default Channel Group, May 13, 2026. Reported by Search Engine Journal, May 14, 2026.
- Search Engine Roundtable, “OpenAI Drops UTM Parameters From ChatGPT Results (GPT-6)”, October 14, 2026, updated October 15, 2026 with Cedric Huesler’s confirmation that the issue was a bug.
- Evertune, “The week without ChatGPT attribution”, October 2026 (missing utm_source tags from October 9, 2026).
- Google Search Central, documentation on AI features in Search.
Note on sourcing: the share and conversion figures come from separate studies with different samples and dates, and are shown side by side for orientation only. The description of GA4’s AI Assistant channel is based on Google’s announcement and contemporaneous trade reporting, with the current default-channel definition checked before scheduling. Verify the channel’s current coverage in your own property.
Verified source links: SE Ranking AI traffic study · Seer conversion case study · GA4 default channel definitions
Muhammad Asjad Khan is a seasoned digital marketer and the Chief Operating Officer at Digital Age, where he leads strategy, innovation, and client success. With years of experience in SEO, content marketing, and performance-driven digital campaigns, Asjad is passionate about helping brands grow their online presence and achieve measurable results. When he’s not optimizing websites or scaling marketing strategies, he’s exploring the latest trends in tech and digital media