Better has no AI in your clinical work. Here's the standard that would have to change that.
We're not anti-technology — we're pro-evidence and pro-consent. So instead of a vague “no AI” promise, we're publishing the actual bar, in advance and in writing.
Every major EHR now sells an AI add-on that records or transcribes therapy sessions. Each comes with retention policies, consent complexities, and a monthly fee. We weren't comfortable putting that between clients and their privacy — so we didn't build it.
The bar
AI is off our roadmap for anything that touches protected health information — session support, note-taking, progress notes, clinical measures — until all of the following are true:
Even then, we'd also have to see it as genuinely valuable to care — a bar of its own. Therapy has taught us not to say “never,” so we offer something sturdier: published conditions, plain principles, and communication you can hold us to. We review this standard every six months and publish the result on this page, dated. We don't foresee AI in Better's clinical work — and it cannot happen quietly. Until every condition is met, Better's clinical tools stay 100% human.
Our commitments
Will Better ever add AI?
Not until it clears a bar we've published in advance: real regulatory precedent from state boards, the APA, and the FDA, plus a genuinely low-environmental-impact way to run it — and we'd have to see it as genuinely valuable to care. None of that is true today, and we don't foresee it. If it ever shipped, it would be an option, not a default: opt-in only, human-in-the-loop, consent-first, and your data wouldn't train models — ours or anyone's. Until then: 100% human, by design.