Five. That is how many distinct documentation types the American Occupational Therapy Association maps to a single episode of care: the occupational profile, the evaluation report, the intervention plan, contact and progress notes recorded across the course of treatment, and a discharge or discontinuation report (American Occupational Therapy Association, documentation guidance, checked August 2026). If you are writing all of that yourself between clients, with no administrative staff to hand any of it to, an AI scribe sounds like the obvious fix. The problem is that most of the AI scribes marketed to rehab therapists were not built around that five-stage structure. They were built to write one note, a session summary formatted the way a general medical SOAP note looks, because that is the shape of documentation the underlying language model saw most of during training.

This site has already covered occupational therapy and speech-language therapy together twice, once on billing automation and once on safeguarding vulnerable caseloads. Both were right to treat the two disciplines as facing the same regulatory and administrative pressure. But bundling OT and SLT into one buying decision for an AI scribe glosses over a real structural difference in what each discipline actually has to produce, and in how Medicare pays for it. Solo practitioners evaluating a tool this month deserve the more specific version.

What AOTA actually asks you to produce

AOTA's documentation guidance groups the work into four stages that generate five distinct outputs: screening, evaluation (which produces the occupational profile and the evaluation report), intervention (the intervention plan, contact notes, and progress notes across the episode of care), and outcomes (the discharge report, which summarizes the change in a client's ability to engage in occupations between the initial evaluation and discharge) (American Occupational Therapy Association, documentation guidance and occupational profile template, checked August 2026). Every evaluation is required to include the occupational profile and an analysis of occupational performance, meaning the client's own goals, roles, and environment sit inside the clinical record from day one, not just a diagnosis and an impairment score.

For a solo practitioner, that is five separate writing tasks per client relationship, each with a different structure and a different audience, spread across weeks or months rather than produced once. A tool that only speeds up the fifth of those, the routine progress note, is solving the smallest part of the actual paperwork load.

So what for you: when you evaluate an AI scribe, ask specifically whether it can produce a usable draft of an occupational profile and a discharge report, not just a session note. If a vendor cannot show you either, they have automated the easy 20% of your documentation and left the harder 80% for you to keep writing by hand.

Speech-language therapy's documentation runs on a different logic

Speech-language pathology documentation is typically organized around tracking several concurrent goals within a single session, each measured with its own cueing level, accuracy percentage, or standardized test score, alongside full evaluations that can include swallowing assessments with specific diet recommendations (industry SLP documentation and AI scribe guides, checked August 2026, vendor-adjacent sources; treat the framing as descriptive rather than as an independently verified clinical standard). That is a fundamentally different unit of measurement from OT's occupational profile, which tracks a person's function and participation in valued activities across an entire episode of care rather than a percentage-accuracy score inside a single 30-minute session.

The practical effect is that an AI scribe tuned to catch cueing levels and accuracy percentages well may still produce a poor occupational profile, because it was never trained to ask the questions an occupational profile requires: what does this person need and want to do, and what is stopping them. Conversely, a tool built around function-and-occupation narrative writing may miss the granular, per-goal metric tracking an SLT's payer expects to see in every note. Neither gap shows up in a five-minute vendor demo built around a single polished example.

So what for you: if a vendor's pitch leans on "we work great for OT, PT, and SLP," ask them to show you, side by side, an OT evaluation report and an SLT session note their tool actually produced for a real (de-identified) client. A single demo note proves the tool can write English. It does not prove it understands either discipline's structure.

Medicare treats the two disciplines differently too, and not in the pairing you'd expect

Here is the detail that surprises most solo OT and SLT clinicians: for Medicare billing purposes, speech-language pathology's threshold is shared with physical therapy, not with occupational therapy. The CY 2026 KX modifier threshold is $2,480 for occupational therapy services on their own, and a separate $2,480 for physical therapy and speech-language pathology services combined, meaning PT and SLP draw from the same annual pool while OT tracks entirely apart from both (Centers for Medicare and Medicaid Services, CY 2026 therapy threshold update, confirmed via the PA Providers Association, February 2026, re-confirmed August 2026). The targeted medical review threshold above that follows the same split: $3,000 for occupational therapy, and $3,000 for the combined physical therapy and speech-language pathology pool (American Speech-Language-Hearing Association, Medicare Reimbursement of Speech-Language Pathology Services, checked 2026).

This site's content, and most vendor marketing aimed at solo allied health practices, groups OT and SLT together constantly, and for good reason: caseloads, consent questions, and safeguarding duties genuinely overlap. But if your practice also delivers or refers physical therapy, your SLP claims are drawing down the same $2,480 pool as your PT claims, while your OT claims are not touching it at all. Tracking both disciplines in one undifferentiated spreadsheet is an easy way to miss the point where you need to start attaching the KX modifier.

So what for you: if you bill more than one of these three disciplines, keep OT threshold tracking entirely separate from PT and SLP tracking. The evidence points toward treating this as two distinct running totals, not one, from the first day of the calendar year.

What this means when you're shopping for an AI scribe

A comparison of AI scribe vendors marketing to rehab therapists found a range of monthly per-user pricing from roughly $75 to $150, with most tools built as an overlay that sits on top of whichever EHR you already use rather than a native part of it (OT Potential, "AI Scribes Compared: for OT, PT, and SLP," Sarah Lyon OTR/L, published February 2026, updated July 2026; pricing is vendor-published and subject to change, flagged accordingly). That same review raises a specific warning worth taking seriously for OT: if a scribe's underlying language model is solely a general-purpose model rather than one trained specifically on clinical documentation for your discipline, treat that as a red flag, because a general-purpose model carries the structure and bias of whatever it was trained on first, not the structure AOTA actually requires of you.

In practice, that means the demo question that matters most is not "how much time will this save me," which every vendor will answer well. It is "was this model trained on occupational therapy documentation specifically, including the occupational profile format, or was it adapted from a general medical or physical therapy model afterward." Ask for a straight answer, in writing if you can get it, before you commit a client session to any tool.

So what for you: budget your evaluation time toward the two hardest outputs, the occupational profile and the discharge report, rather than the easiest one. A tool that nails a routine progress note but cannot draft either of the other two is not actually solving your biggest time cost, even if the sales demo feels impressive.

The consent step that applies whichever tool you choose

Whatever AI scribe you land on, AOTA's Policy E.19 on the ethical use of artificial intelligence, effective April 2025, sets one requirement that does not change: when an AI tool is used during a recorded session for transcription, documentation, or data analysis, you must inform the client in advance of the recording and its intended use, the scope of data collection, and how the information will be stored or shared, and obtain informed consent (American Occupational Therapy Association, Policy E.19, effective April 2025). This is an association ethics policy rather than federal law, but it is the operative professional standard you are held to, and it applies regardless of which vendor's tool is running in the background.

So what for you: build this into the same conversation where you introduce any new AI tool to a client, document that the conversation happened in the chart, and repeat it only if you materially change which tool you use or what it does with the recording. One documented conversation at the start of a new tool's use covers you far better than an assumption that a signed general intake form already handles it.

None of this means the underlying case for AI in a solo OT practice is weak. It means the case is more specific than "get an AI scribe," and the vendors selling into rehab therapy have every incentive to blur that specificity because a single undifferentiated product is cheaper to build and market than three discipline-specific ones. Evaluate against AOTA's actual five-stage structure, track your Medicare threshold separately from PT and SLP, and ask every vendor to prove their tool was built for occupational therapy specifically rather than adapted for it after the fact. If you want an independent read on where AI genuinely pays off in your specific caseload, scored without a vendor relationship behind the recommendation, the AI Opportunity and Growth Assessment does exactly that, and the next step is a free 20-minute discovery call.

Not sure whether an AI scribe actually fits your documentation structure, or just looks good in a demo? A short call is usually enough to tell. Book a 20-minute call.

The Clinical AI Briefing

One practical AI insight for healthcare practices every week. No hype. Evidence and outcomes only.

Related: RCSLT's new AI principles name audio, video and transcript data directly · The $2,480 threshold: what AI billing tools actually fix for solo OT and SLT practices · AI HIPAA rule delayed to 2027. Safeguarding duty isn't.