Observation Period: 2026-05-01 ~ 2026-05-22
Estimated Percentage of Total Accesses Referred from ChatGPT
0.00131%
| Item | Access Count |
|---|---|
| Total Accesses | 304,911 |
| GPTBot (Crawl & Training) | 11,196 |
| ChatGPT User (In-App Fetching) | 1,852 |
| OAI-SearchBot (Search Purpose) | 9,227 |
| Visit From GPT (Referrals from ChatGPT) | 4 |
The number of accesses by ChatGPT-related functions was 22,275.
*22,275 = 11,196 (GPTBot) + 1,852 (ChatGPT User) + 9,227 (OAI-SearchBot)
Share of ChatGPT-related traffic in total accesses: 7.31%
Share of human referrals from ChatGPT in total accesses: 0.0013%
Referral rate relative to ChatGPT-related functional accesses: 0.018%
*In this article, "Referrals from ChatGPT" refers to "human web traffic estimated to have arrived via ChatGPT."
During this observation period, a significant number of accesses by GPTBot, ChatGPT User, and OAI-SearchBot were recorded.
In particular, the access volume from GPTBot and ChatGPT User was high, confirming that OpenAI-based AI systems are actively crawling and referencing web page information.
On the flip side, despite this large volume of automated traffic, the actual number of human referrals coming from ChatGPT to the website was extremely small.
In short, the tangible traffic benefit to websites remains negligible, a trend that warrants close, ongoing monitoring.
Google Search is inherently structured to guide users outwards to external websites.
Conversely, ChatGPT is a conversational AI designed to resolve and complete the user's inquiry within the chat dialog itself.
Consequently, even when it references web pages to generate its answers, the instances where it actually routes users to the source website appear highly limited.
For instance, if a user asks, "I want to go camping, give me some recommendations," ChatGPT will provide specific campground names and highlights directly. However, it displays links to those campgrounds very sparingly.
As a result, users take the suggested features as a reference and then turn to Google Search or map services to do their own deep-diving, eventually making a booking.
These observation results underscore the necessity of treating "AI scraping/browsing your site" and "humans visiting your site" as two entirely separate phenomena.
This data has been aggregated and extracted from the combined access logs of multiple small-to-medium enterprise (SME) websites within Japan.
WebPita tracks and analyzes ChatGPT-related traffic by cross-referencing OpenAI's public documentation with web access logs based on the following criteria:
Our analysis focuses primarily on hits that resemble a standard "page view" by either an AI bot or an AI-referred user.
How WebPita Identifies ChatGPT-Related Traffic
Criteria for Determining ChatGPT Referrals and Post-Click Navigation
While WebPita's logs captured an abundance of backend ChatGPT activity, actual human referrals driven by the AI were trace-level at best.
In the era of Generative AI, "AI referencing your information" does not automatically translate into "users visiting your website." It is critical for webmasters to adapt to this shift.
