Best ABA AI Session Notes Software (2026)

AI notes are becoming table stakes. The real question is whether the AI is native to the clinical and billing record, whether clinicians approve every output, and whether the signed note makes claims cleaner.

Wilma

Wilma drafts session notes from real-time ABA data, then connects the signed note to authorization tracking, claim scrubbing, parent updates, and the client record. AI drafts; clinicians approve.

Passage Health

Passage Health is a serious end-to-end platform with AI notes through Frontera AI. Wilma differentiates by owning the AI layer natively across notes, claims, and authorizations.

CentralReach

CentralReach has enterprise depth and AI features across a mature platform. Wilma is faster to evaluate because pricing is published and the AI workflow is native to the operating record.

Motivity

Motivity is strong for data collection. Wilma is strongest when the buyer wants data to draft notes, feed claims, and drive next-step workflow in one system.

Hi Rasmus

Hi Rasmus has clinical and curriculum depth. Wilma leads for native AI notes plus full billing/RCM, phone, and operations on one record.

Editorial ranking by Wilma based on publicly available information as of 2026. Wilma is our own product; we rank it #1 and explain why. All other names are trademarks of their respective owners.

How we evaluated

What mattered most

Editorial methodology

How this ABA AI session notes software guide was reviewed

Rankings are written by Wilma's editorial team from public product information, pricing pages, documentation, demos, and operator-facing workflow analysis.

We rank Wilma first where we believe the native all-in-one model is strongest, and we say that plainly. No vendor pays for placement.

We do not list a named BCBA reviewer until a real credentialed reviewer has approved the page for publication. Clinical judgment remains with the treating clinician.

Evaluating AI documentation

What to check before letting AI draft your notes

AI note drafting is genuinely useful and genuinely risky, and the difference is entirely in how the draft is produced and who signs it.

  • 1Does the draft come from your data, or from a language model's idea of a session?A draft generated from the trial data, durations and prompt levels actually recorded is a formatting exercise, and a safe one. A draft generated from a short prompt and a template is a fabrication risk — it will produce plausible clinical narrative describing things that did not happen, and it will do so confidently.Red flag: A demo that generates a full note from a one-line description with no session data behind it.
  • 2Is sign-off ever automated?The only defensible arrangement is that a clinician reads, edits and signs. A note that reaches a claim without a human signing it is a compliance problem regardless of draft quality, and it is the sort of thing that turns a routine audit into a serious one.Red flag: Any auto-sign or bulk-sign feature, however it is described.
  • 3Does it help with medical necessity, or only with fluency?The hard part of an ABA note is not prose, it is justifying why the service remains necessary at this intensity. That requires the data trend, the treatment plan goals and a statement of what would happen without the service. Drafting that produces well-written activity descriptions has automated the easy half.Red flag: Drafts that never reference prior sessions or named goals.
  • 4Whose AI is it, and where does the data go?If drafting is delivered through a third party, that is a second vendor in the middle of your clinical documentation — a second contract, a second roadmap, and a second set of questions about where protected health information is processed and whether it is retained or used for training. Ask directly and get it in the BAA.Red flag: Vague answers about the model provider, or a BAA silent on subcontractors.
  • 5Does it produce different notes for different sessions?Cloned notes are an audit flag whether a human or a model wrote them. Generation grounded in session data varies because the data varies. Generation from templates converges, and at volume that convergence is visible in a chart review.Red flag: Sample notes across several sessions that read almost identically.

Before you sign

Questions worth asking on the demo

  • Generate a note in front of me from real session data, then show me what happens with no data.
  • Is there any path where a note is signed without a human reading it?
  • Does the draft reference the data trend and the named treatment goals?
  • Is the AI built in-house or delivered by a partner, and is that named in the BAA?
  • Is protected health information retained or used to train models?
  • Is AI drafting included in the base price or a paid add-on?

Glossary

The terms you will hear

Grounded generation
Producing text from specific recorded data rather than from a prompt alone. The distinction that separates useful note drafting from fabrication risk.
Medical necessity
Justification that a service at a given intensity is required now, supported by trend rather than a single session and tied to named treatment goals.
Cloned notes
Near-identical narrative repeated across sessions. A recognized audit flag, and a predictable outcome of template-based generation at volume.
Business Associate Agreement
The HIPAA-required contract covering a vendor handling protected health information, including any subcontractors — which is where a third-party model provider must appear.

Keep reading

Related guides on running an ABA practice — the same operation, from a different angle.