Scrunch recommends tracking two types of AI traffic—AI agent visits and AI referrals—to figure out which AI tools are sending the most traffic to a website.
Additional context: Tracking AI agent traffic reveals how often different AI platforms crawl a website for different purposes. Tracking AI referral traffic reveals how often humans click through to a website from different AI platforms.

For example, a Scrunch user would track both by:
AI agent traffic
AI referral traffic
There's no universal benchmark, but the share of AI bot traffic is steadily growing across industries while human traffic declines.
Scrunch recommends focusing more on the mix of traffic and what it signals versus aiming for a specific ratio.
A useful rule of thumb is to treat retrieval bot traffic as the closest proxy to human intent because it’s tied to live prompts. Training and indexing traffic are more related to general discoverability.
Users should concentrate on whether retrieval traffic is growing and whether it’s coming from the platforms they care about.
Scrunch recommends tracking brand presence, citations, referral traffic, AI agent traffic, and share of voice versus competitors as key performance indicators.
Scrunch recommends monitoring AI search trend data like brand mentions and citations consistently over 2-3 week periods to identify real trends versus one-off changes.
Scrunch recommends estimating how many prompts to track for AI search using the following approach: X [# of topic clusters] x Y [12-15 questions related to each topic cluster] = Z [# of AI search prompts to track]. The primary goal is to get a representative sampling of data across all customer journey stages via a mix of branded and non-branded prompts.