
About
About Barkhausen AI
Visibility in AI assistants — whether an answer names you, and how often — is now consequential and, done properly, measurable.
MIT Museum · Public domain
The work
Conventions fix what a valid measurement of AI visibility must satisfy — a maturity model, metric definitions, and sampling requirements. Research reports and whitepapers examine how audiences use AI assistants, and where common measurement practice fails. Public-data studies — surveys of what is directly observable on the open web — publish their full datasets under an open license.
Principles
- Evidence. Every statistic carries its sample size and time window, names its source, and — for anything inferential — states its uncertainty. Sources are primary and checkable.
- Sources over assertion. Claims rest on public evidence, not on the standing of whoever makes them.
- Nothing is silently changed. Claims are not edited in place; corrections are dated and left visible.
- No commercial evaluation. No product is reviewed, ranked, or recommended; no vendor is endorsed.
- Open by default. Text, figures, and datasets are released under CC-BY-4.0, free to quote and reuse with attribution.
About the name
The Barkhausen effect is the observation that a magnetic material, placed under a steadily increasing field, does not magnetize smoothly but in a series of discrete jumps. Visibility in AI assistants behaves similarly: steady work produces change that appears in jumps. The metaphor gives Barkhausen AI its name and its recurring figure — the step curve.

Contact
Correspondence, comments on conventions, and corrections: contact@barkhausen.ai.