kya labs · Methodology
How we collect our research.
What we collect, how we test, and how to interpret the findings.
Why we study this
AI agents increasingly act as an interface to the web. We study what they do in defined situations so readers can understand their behavior without confusing observation with a prediction about every user or merchant.
How a study runs
Studies may use direct page requests, browser-driven visits, or search-sourced AI responses. We identify which method was tested; a result from one does not establish what every AI assistant can see.
What counts as evidence
Getting onto a site, reading useful information, navigating a journey and completing a task are separate observations. A readable page does not prove a completed shopping task. Each study states its endpoint and the evidence supporting it.
Dates, boundaries, and limits
We distinguish successful observations, verified barriers, failures of our collection tools and missing evidence. Results describe the dates, sites, tasks and methods tested. Changes in models or collection methods can affect comparisons over time.
Privacy and aggregation
Public research is aggregated and anonymized. Named investigations are separately scoped. We distinguish observed actions, an agent’s explanation, and analyst interpretation, and redact sensitive values where possible.
Questions about a study?
For a deeper question about scope or interpretation, start a conversation through Custom Studies.