Standard web analytics tools like Yandex Metrica and GA4 often misclassify traffic from AI services as direct visits if the referral header or UTM tags are missing. Tenpoint Labs estimates that AI-driven traffic accounts for a maximum of 6.3% of direct visits, though other sources suggest figures between 30% and 40%.
To track AI traffic, analysts can create segments based on referrer domains and UTM parameters. For Yandex Metrica, this involves using regex filters for domains like chatgpt.com, perplexity.ai, and claude.ai. GA4 includes a built-in 'AI Assistant' channel for specific services, but custom channel groups are required for broader tracking.
Estimating 'hidden' AI traffic requires filtering direct visits based on behavioral patterns. Recommended criteria include excluding homepage entries and specific landing pages, while focusing on engaged sessions with at least two page views or a duration of 30 seconds or more to filter out bots and link previews.
The methodology has limitations, including the inability to distinguish AI traffic from organic search for Yandex Alice, potential data loss in mobile apps, and the risk of false positives from non-AI direct traffic. Analysts are advised to present these figures as estimates rather than absolute values.
