How Much Traffic Comes from AI Search
Tomohiro Iida ·
How much traffic do AI search tools such as ChatGPT and Perplexity actually send to a website? Rather than guessing, this article publishes primary-source data: Google Analytics 4 (GA4) numbers for netsujo.jp's own site over the past 90 days. We do not inflate the figures — we report exactly what the data shows, with its source. GA4 classifies referral traffic from generative AI services such as ChatGPT, Perplexity, and Gemini under a channel called 'AI Assistant.' As of 25 June 2026, over the preceding 90 days that channel accounted for about 1% of all sessions (12 sessions), against roughly 59% from organic search. The number is small. We do not call it a surge or an early sign of growth: this observation does not fix a single aggregation window, so it cannot support a claim about the trend. The conclusion: organic search remains the dominant channel, and AI-driven traffic, while still marginal, is worth tracking alongside it.
Key takeaways
- netsujo.jp's own GA4 data shows AI Assistant traffic (the GA4 channel that captures referrals from generative AI tools) at about 1% of all sessions — 12 sessions over the past 90 days, against about 59% from organic search (source: GA4, past 90 days, as of June 25, 2026).
- The number is small. We do not claim a surge or an early sign of growth, because this observation does not fix a single aggregation window and so cannot support a trend claim.
- Organic search remains the dominant channel by far. Our conclusion is to keep SEO as the foundation while building out AIO/GEO — AI search visibility and optimization — in parallel.
What GA4's 'AI Assistant' channel actually measures
Before looking at the numbers, it's worth being precise about what is being measured. GA4 added 'AI Assistant' as one of its default channel groupings: referral traffic from generative AI services such as ChatGPT, Perplexity, and Gemini is classified here when a user opens a link that appeared in an AI-generated answer. The classification is based on the referrer domain at the time of the visit — if a user is only mentioned by name in an AI's answer but never clicks through, it never shows up in this channel. The classification is also likely incomplete: depending on the AI service or how a user navigates, AI-driven visits can sometimes be counted under Direct or Referral instead. The honest way to read the 'AI Assistant' figure, then, is as a lower-bound estimate of AI-driven traffic, not a precise total.
The numbers: session share by channel
| Channel | Sessions | Share |
|---|---|---|
| Organic Search | 1,208 | About 59% |
| Direct | 675 | About 33% |
| Referral | 102 | About 5% |
| Organic Social | 34 | About 2% |
| Unassigned | 21 | About 1% |
| AI Assistant | 12 | About 1% |
Source: GA4 (property 382871067), past 90 days, as of June 25, 2026. Shares are rounded and may not sum to 100%. Site-wide, the period recorded 2,052 sessions and 1,626 users. Organic search is clearly the leading channel, followed by direct traffic; AI Assistant sessions are a small fraction of the total. How those 12 sessions were distributed within the period is not covered here: showing a daily skew would require fixing one aggregation window and matching the denominator to it, which this observation does not do. We make no claim of growth or decline, because the absolute numbers are still small and a few weeks of observation isn't enough to call a trend.
How we are interpreting this data
- Organic search is still the main driver. About 59% of sessions over the past 90 days came from organic search — regardless of how much attention AI search gets, it remains by far the largest source of traffic on our site, and SEO fundamentals continue to carry the bulk of it.
- AI-driven traffic was still small in this observation. AI Assistant sessions were 12 out of roughly 2,052 over 90 days — about 1%. Because this observation does not use a single fixed aggregation window for within-period comparison, it does not support a claim about whether traffic was growing or declining. The point of publishing primary data is to report the measured scale without inflating a small number.
- This is why we treat SEO and AIO/GEO (AI search visibility and optimization) as two tracks to run in parallel, not a choice between them. SEO is still the primary channel and deserves no less investment, while AI referral traffic was observable in the measured period and is worth tracking and preparing for on the same foundation of accuracy and structure.
Related work on being cited by AI
This article focuses only on the traffic data. The concrete steps we've taken to try to grow that AI-driven traffic — structured data, factual consistency, answer-first writing, and an llms.txt file — are covered in a separate series of articles: our own practical log of SEO/AIO work, an explainer on what Generative Engine Optimization (GEO) means, a guide to writing an llms.txt file, and an article on common reasons sites are not cited by AI. This article is the primary-source data that those efforts are aimed at moving.