Agent Traffic
Agent traffic is the visitors that aren’t people, where the serve-or-block call became a revenue decision.
What agent traffic is
Agent traffic is the visitors that aren’t people: AI agents browsing, comparing, and buying on shoppers’ behalf, a growing slice of sessions that converts differently, reads differently, and forces a choice older bot thinking never had to make.
Why agent traffic matters
Bot management was built on one assumption, non-human means unwelcome, and shopping agents break it: the scraper stealing your catalog and the agent buying from it can look similar in logs, but one is a threat and the other is a customer holding someone’s budget. Blocking all automation now means turning away sales; the store’s stance on agents became a revenue decision wearing a security setting.
Managing the non-human customer
- Distinguish intent: hostile automation blocked as ever, declared shopping agents treated as buyers
- Serve the machine-readable store: schema, honest availability, text policies agents can parse
- Keep checkout passable: human-only gates at payment turn away the agents you wanted
- Measure separately: agent sessions skew every human-behavior metric they’re mixed into
Frequently asked questions
How do stores tell good agents from bad bots?
Imperfectly, for now: legitimate agents increasingly declare themselves and verification standards are forming, while hostile automation still hides. The working posture is layered, verify what declares itself, watch behavior for the rest, and reserve hard blocks for demonstrated harm rather than non-humanity.
Does agent traffic wreck analytics?
Mixed in, yes: agents don’t scroll, hesitate, or respond to urgency, so blended reports misread both audiences. Segmenting known agent sessions keeps human CRO honest and starts the more interesting series, what share of revenue arrives by proxy, and is it growing.