Analyst finds no reliable sign of AI traffic hiding in Direct
Analyst Kevin Indig tested, rather than assumed, the common claim that AI-referred visits hide inside a site’s Direct traffic channel. Across a 12-site sample tracked January to September 2026, total Direct traffic rose 140.9% but engaged Direct sessions — visits where someone actually read the page — rose just 2.3%, with one site logging 409,000 Direct sessions of which only 198 counted as engaged, per Indig’s Growth Memo. Testing weekly AI visibility against branded-search and direct traffic across four brands and 632 time-shifted comparisons, correlations as high as 0.72 collapsed once tested against randomly shuffled alignments. Indig contrasts this with Ramp’s own internal analysis, which attributed 30-40% of its direct traffic to answer engines by filtering bot and email-campaign patterns from long-tail blog URLs before comparing what remained to organic-search behavior.
Why it matters: It complicates a widely repeated claim — including a GA4 study already on this record finding 22.4% of AI Overview citation clicks misattributed to Direct — by testing it directly and finding the signal isn't reliably detectable with current analytics methods.
The record: ChatGPT
Glossary: Answer engineAI visibility
Via Kevin Indig / Growth Memo ↗
Posted to the wire October 21, 2026. Edited by Joe Balewski.