AI in ABA (2026): What's Actually Shipping

Every ABA software vendor now says “AI.” This guide maps what they actually document — compiled from a July 2026 review of current vendor materials across the platforms in our directory, with the same rules as everything here: no vendor involvement, no invented claims, and a clear line between what’s shipping and what’s marketing.

The headline finding

AI in ABA isn’t arriving as standalone tools — it’s arriving inside the platforms you already evaluate. Of the major practice management and data collection vendors we reviewed, most now document at least one AI feature, and the pattern of where they deployed it first is revealing: documentation and billing — the highest-volume, lowest-clinical-risk work — long before anything touching clinical decision-making.

The map, by function

Session notes: the beachhead. The most widespread deployment by far. CentralReach documents note drafting (NoteDraftAI) and note auditing (NoteGuardAI — reviewing for errors and compliance risk); Rethink documents Session Note AI; Artemis documents AI session notes; Raven documents one-click notes generated from collected session data; Theralytics documents generative drafts for narrative summaries; Portia documents a clinical note generator. The logic is sound: documentation load is the industry’s most-cited burnout driver, and note drafting keeps a human signature in the loop by design.

Claims and billing. CentralReach documents claim-auditing and claim-accelerating AI; Rethink documents BillAI. Auditing claims before submission is arguably the most defensible AI use case in the entire stack — it’s error-checking against rules, exactly what the denial traps demand.

Scheduling. CentralReach documents AI-powered scheduling (ScheduleAI); TheraDriver positions its entire suite around it (“build optimal schedules with just 1 click”) alongside care-operations analytics. ABA scheduling is a genuine constraint-optimization problem — credentials, authorizations, locations, cancellations — which is to say, a problem computers were always going to be good at.

Treatment plans: where caution rises. Artemis and TheraDriver document AI treatment-plan drafting. This is the frontier where “AI drafts, clinician owns” stops being a workflow nicety and becomes the entire ethical question — a drafted note describes what happened; a drafted plan proposes what should happen. Evaluate these features with your clinical leadership in the room.

The regulated tier: diagnostics. Separate from platform features, ABA-adjacent AI has an FDA-regulated lane: Cognoa’s Canvas Dx holds first-of-its-kind FDA authorization as an AI-based autism diagnosis aid, and EarliPoint has stacked successive FDA clearances for its eye-tracking evaluation. Note the difference in kind: these carry regulatory review; platform AI features carry none.

Who doesn’t lead with AI: Motivity and Hi Rasmus currently market clinical workflow strength rather than AI features, and Passage documents automation (auto-summarized session data, automated graphing) without the AI label. Worth saying plainly: not marketing AI is not a deficiency — in a hype cycle, restraint can be a signal of focus.

Before you turn any of it on

Five questions, in order. Where does clinical data go — will the vendor sign a BAA, and is your data used for model training? Is there a mandatory human-review loop — AI drafts, clinicians sign, and the system enforces that order? What does your payer say — policies on AI-generated documentation are embryonic and evolving; ask your largest payers directly rather than discovering their position in an audit. What’s the audit trail — can you show what was drafted versus what the clinician changed? What’s the honest time saving — pilot with a small team and measure documentation minutes, because a draft that needs heavy editing saves no one anything.

What nobody publishes yet

Accuracy benchmarks. Error rates. Head-to-head note-quality comparisons. Every vendor documents that they have AI; none document how good it is in measurable terms. Until that changes — and our survey will start collecting practitioner-reported experience — treat every AI feature as a pilot, not a purchase.

FAQ

Is AI note-writing compliant? It can be, with the right structure: a BAA in place, clinician review and signature enforced, and documentation that satisfies your payers’ requirements. The compliance risk isn’t the drafting — it’s unreviewed drafting.

Will AI replace RBT data collection? Nothing documented today points that way; shipping features draft, audit, schedule, and summarize around human-collected data.

What’s the best AI tool for ABA? Mostly the wrong question in 2026 — AI arrives as platform features, so the real question remains which platform fits your practice (start here), with its AI features evaluated as one criterion among many.

Compiled from vendor documentation reviewed July 2026. This page updates as the category moves — corrections: [email protected].