Physical Therapy Prior Authorisation: Track Visits With AI

Have the treating therapist write the clinical facts, let AI turn them into a clean request in your template, and track every approval in a sheet with visits approved, visits used, an end date and an alert. A person sends the request through the payer's own channel. AI drafts and organises; the therapist signs and the payer's rules decide.

Payer pages do a good job of listing forms and rules, and the rules differ enough that no tutorial can give you a universal one. What they do not give you is a working routine: who writes the request, how it is checked, where the answer goes and who notices when visits run low. That routine is where practices lose revenue, because a patient treated beyond the approved visits is a claim at risk.

Follow me on Instagram@sagnikteaches

The four moments where authorisations go wrong

Most losses come from four moments, not from the request itself. Seeing them helps you design the tracker around what actually fails.

Connect on LinkedInSagnik Bhattacharya

The first moment connects to your front desk. If you use an AI receptionist, the tutorial on referral and insurance questions for a physical therapy receptionist covers collecting insurance details at the first call, which is when the requirement check should happen.

Subscribe on YouTube@sagnikteaches

Building the tracker first

Start with the tracker rather than the request, because it gives the whole routine a spine. A spreadsheet is enough for a small practice. Use one row per authorisation, not per patient, since one patient may have several over time.

Two formulas do most of the work. Visits left = Approved - Used, and an Alert column that reads "Request now" when visits left is 2 or fewer, or the end date is within 14 days, whichever comes first. Apply conditional formatting so those rows turn amber, and red when visits left is zero. If the spreadsheet side is new to you, the ChatGPT and Excel guide shows how to have an assistant write formulas like these, which you then test on a few rows by hand.

Twelve active authorisations on one tracker

Imagine a clinic with four therapists and 12 patients on active authorisations on one Monday. An extract of the tracker, with illustrative numbers and a review date of 10 March, looks like this.

On this Monday, three of five rows need action: A-205 has one visit left, A-209 is running out of calendar before visits, and A-214 has none left. The front desk does not book A-214's next appointment until the therapist has requested an extension or the patient has agreed in writing to another arrangement. That one rule, "no visits left means no booking without a decision", prevents most unauthorised care.

Across the clinic's 12 rows, say 5 show amber or red. That is a manageable list for a Monday huddle: each flagged row gets an owner and a date. The tracker did in ten minutes what used to rely on someone remembering.

Drafting the request: therapist writes the facts, AI shapes them

A prior authorisation request is mostly a structured clinical summary. Payers vary in what they require, but most ask for similar content: diagnosis, relevant history, functional limitations, objective findings, treatment goals, the proposed plan with frequency and duration, and evidence of progress for an extension. The therapist has all of this in their notes. The job is to get it into the payer's format clearly and without padding.

Ask the therapist to write or dictate the facts in their own words, with no identifiers. Then use a prompt like this.

You are helping a physical therapy clinic turn a therapist's notes
into a prior authorisation request. Use ONLY the facts in the notes.
Do not add diagnoses, test results, scores or goals that are not in
the notes. If a field required by the template is missing, write
"THERAPIST TO COMPLETE:" and name the field.

Output in this layout, with these headings:
Reason for request (one sentence)
Relevant history
Current functional limitations (bullet list)
Objective findings (bullet list, with measurements as given)
Goals (each with a time frame, as given)
Proposed plan: frequency, duration, total visits requested
Progress to date (extensions only)
Why the requested visits are needed (3 to 4 sentences)

Plain, factual language. No exaggeration. Under 400 words.

Therapist's notes (no names or identifiers):
[paste notes]

An illustrative piece of the output, for a request about a knee rehabilitation case:

Current functional limitations
- Unable to climb a flight of stairs without a handrail.
- Walking tolerance limited to about 10 minutes.
- THERAPIST TO COMPLETE: Reported pain level, if required by the payer.

Proposed plan
Twice weekly for four weeks, then weekly for four weeks: 12 visits.

The model flagged a missing field rather than guessing at a pain score. That is the behaviour to look for in every draft. Once the therapist has filled the gap, the clinic adds identifiers locally in the payer's form or portal, not in the chat. Placeholders and the payer form do the rest.

What the therapist must check before it is sent

Sign-off is a short, specific check, not a skim. Give the therapist a five-point list and have them tick it on the draft.

1. Every fact is mine. Nothing in the draft is a diagnosis, measurement or goal I did not supply.
2. Numbers match the record. Measurements, dates and visit counts agree with the chart.
3. Language is accurate, not inflated. No wording that overstates severity to improve the odds of approval.
4. The request matches the plan of care. Frequency, duration and goals agree with what the patient has been told.
5. Nothing is missing that the payer requires. Checked against the payer's current form.

Point three deserves emphasis. A model smooths and strengthens, and a request that overstates need is a compliance problem, not a clever tactic. Insist on factual wording. The therapist's signature, or an electronic equivalent recorded in your system, goes on the final version, and the version actually sent is stored with the tracker row so you can show what was asked for.

Payer rules differ: record them once, use them often

The same patient condition can need no authorisation with one payer, a simple online check with another and a clinical review with a third. Visit limits, validity periods, required forms and turnaround times also vary, and payers update them. Rather than relying on memory, keep a short payer sheet in the same workbook with one row per payer and a handful of fields.

The source and date column is the discipline that saves you. A rule you cannot trace is a rule you cannot trust. Review the sheet whenever a payer sends a notice and at least every quarter. A comparable approach for licences and renewals is described in a compliance calendar for licences, insurance and certificates.

Extensions and denials: two further drafts

Two more situations use the same method. In an extension request the new element is progress to date, which must come from the therapist's re-evaluation: measurements at the start, measurements now, what has improved and what has not, and why further visits are needed. Ask the model to present the before and after side by side, and to add nothing the therapist did not measure. An illustrative before and after might read: stair climbing "unable without a handrail" at the start, "four steps without a handrail" now, goal "full flight" by visit 16. The model organises; the therapist supplies the numbers.

For a denial, the clinical team decides whether to appeal. If they do, AI can help draft the letter from the denial reason and the therapist's response points, following the payer's published appeal route and deadlines. Record the deadline in the tracker the day the denial arrives. The denial grouping method in AI checks for claims and time entries also helps you see whether denials cluster on a payer or a diagnosis.

Running the whole workflow with no names in a chat

The whole workflow can run with no names in any AI chat. The tracker uses reference numbers. The therapist's notes have identifiers removed before pasting. The identifiers are added when the draft goes into the payer's form, outside the chat. If your practice software has its own AI drafting built in, ask the vendor what data it keeps and where; keeping data inside one system is better than copying it out. Where you do use a general assistant, choose a business plan whose terms say content is not used for training (ChatGPT Business and Claude Team both say so on their pricing pages, checked November 2026). Check your own data-protection duties or ask your adviser before starting. The comparison in physiotherapy clinic software with built-in AI is a good place to start looking at what you already have.

The Monday huddle: ten minutes over the tracker

A tracker only works if someone reads it at a fixed time. Put it on the Monday huddle agenda. The practice manager shares the sheet filtered to amber and red rows and goes down them one at a time, asking the same three questions: who owns the next action, what is the date and what does the patient know. Using the example above, the huddle might produce this.

Notice that none of these decisions is for an AI. The tracker gives the room the facts, and the huddle makes the calls. Ten minutes weekly across a clinic of four therapists is a small price against even one unauthorised course of care.

A quick sum: what one missed extension costs

To see whether the routine is worth the effort, do the arithmetic with your own figures. Say a visit is paid at $90 and a patient has three visits delivered after approval ran out. If the payer declines those visits, that is $270 unpaid, plus the staff time to appeal or write it off, plus a patient who may be billed unexpectedly, which is a relationship cost no figure captures. If the clinic has a missed extension once a month, that is $270 times twelve, or $3,240 a year, from one failure mode. A tracker that takes ten minutes a week (about nine hours a year) is cheap against that. These numbers are illustrative; use your own average payment per visit and your own count of missed extensions from last year's denials.

What to say when visits run out

The hardest moment is telling a patient that the next visit cannot be booked yet. How it is said matters. Here are two versions of a message the front desk might send, both drafted with AI help and checked by the manager.

The better versions are accurate, name a next step and leave the payment discussion to a person. Whether and how a patient can be asked to pay for visits beyond an approval depends on the payer's contract and the patient's agreement signed in advance, so that conversation belongs with the manager and not a template. A line on the intake form, written once and checked by your adviser, saying what happens when approved visits run out, prevents most awkward calls.

A patient who changes plan part-way through

A less obvious case: a patient's employer changes insurer in the middle of a plan of care. The old authorisation does not carry over, and the new payer may have different rules. The tracker handles it with a new row, a status of "Pending" and a requirement check on the new payer, while the old row is closed with its final count. The front desk learns of the change when eligibility is checked at the first visit of the month, which is why that monthly check is on the routine. The therapist's facts need only a refresh, not a rewrite, so the AI draft takes minutes: paste the previous request notes plus the progress to date and ask for the new payer's layout. Without that workflow, the first sign is usually a rejected claim six weeks later.

A four-week set-up for a small practice

A small practice can set this up over four weeks without disrupting patients.

1. Week 1. Build the tracker and the payer sheet. Enter every current authorisation from your payers' decision letters, and ask each therapist to confirm their own patients' rows.
2. Week 2. Add formulas and alerts. Run the Monday huddle for the first time. Expect to find at least one authorisation nobody knew had expired.
3. Week 3. Write the request template and the therapist checklist. Draft two real requests with the AI prompt, using placeholders, and compare time taken with the old way.
4. Week 4. Add the requirement check to booking and the visits-used pull from the schedule. Review what changed and what the team found hard.

After four weeks you will know whether the routine fits, and which parts could move into your practice software. Doing it by hand first means you buy software, if you do, knowing exactly what it needs to do.

Typos, stale counts and assumed approvals

• The tracker is updated monthly. Visits used goes stale and alerts fire late. Fix: pull visits used from the schedule export every Monday, or move to software that does it.
• Approved visits are copied wrongly. A typo turns 8 into 18. Fix: store the payer's decision letter beside the row and have a second person check each entry against it.
• The model fills a gap. It adds a plausible functional limitation. Fix: the "use only the facts" instruction and a therapist check that traces every bullet to the chart.
• Authorisation is assumed. A patient on a new plan is treated without a check. Fix: no first appointment is confirmed without a recorded answer in "Auth required?".

When to leave the spreadsheet

A tracker sheet suits a practice with up to a few dozen active authorisations and a front-desk person who looks at it weekly. Beyond that, or when several people edit it at once, errors creep in. Signs to move on: two people overwrote each other's edits, visits used is out of date more often than not, or you spend over an hour a week keeping it alive. At that point check whether your practice software has an authorisation module, using the guide to outgrowing spreadsheets as a checklist. Most of the routine above (requirement check, therapist-written facts, alerts with owners, sign-off) transfers directly into software.

Whether you stay on the sheet or move to software later, start with the tracker, add the alert rule, and run it for a month before you automate any drafting.

Questions about authorisation workflows

Can AI submit the request to the payer?

Not safely as a general rule. Let AI draft the wording and organise the tracker, but a person submits through the payer's portal or form and keeps the confirmation. Payers have their own required forms and rules, so the draft is a starting point for that form, not a replacement.

What goes in the tracker?

Patient reference, payer, authorisation number, approved visits, visits used, start and end dates, next review date and status. Add a column for who owns the follow-up. Use a reference number rather than the patient's name wherever the tracker is shared or copied into any AI tool.

How do I set expiry alerts?

Use spreadsheet conditional formatting or a calendar reminder: flag a record when two visits or fewer remain, or when the end date is within 14 days. Pick thresholds that give your clinicians time to request an extension before the last approved visit, and adjust them as you learn.

Does every payer need prior authorisation?

No. Requirements differ by payer, plan and sometimes by employer, and they change. Verify each patient's requirement at the first contact from the payer's own provider page or by phone, and record the answer and date in the tracker.

Who must sign off the clinical justification?

The treating therapist. AI can help structure the justification from the therapist's own notes, but the clinical reasoning, goals and expected progress must be the therapist's, checked and signed before the request goes out.

Further reads

• How Physiotherapists Use AI to Write Treatment Notes Faster — Clear treatment notes are the raw material for a strong request.
• Therapy Practice Management Software: What the AI Add-Ons Cost — What practice software AI add-ons cost.
• AI Tools for Physiotherapy Clinics: What Each One Actually Does — Where authorisation tracking sits among clinic AI tools.
• Physical Therapy Patient Retention: AI Messages Between Visits — Visit limits shape how you message patients between visits.
• How to Set Up Human Review for AI Work Without Slowing Down — Make therapist sign-off a routine rather than a rush.
• Prior Authorisation With AI: Drafting Letters and Tracking Status — The same draft-and-track idea for other clinic types.
• AI Denial Management for Medical Billing: Group, Appeal, Track — Export your remits, group denials by reason code, draft appeals with AI and track outcomes on one sheet, with a worked month in numbers.
• AI for Nurse Practitioners: Prior-Auth Letters and Practice Policies — Prior-auth letters, patient messages and policy drafts for an NP-owned practice, with a clear line on what needs a signed BAA.
• Physical Therapy Policies and Procedures: A Clinic Manual Drafted With AI — A PT clinic manual drafted with AI: intake, consent, cancellations, records and complaints written from your routines, with a clinician sign-off log.
• Using ChatGPT for Physical Therapy: Referring-Doctor Updates and Outreach — Build a library of referral update and outreach templates in ChatGPT, then fill the patient details from your own records, never the chatbot.
• AI Use in Physical Therapy: Which Uses Need Clinician Sign-Off — A clinic owner's ladder for AI in physical therapy: which uses run alone, which need a clinician to sign, and which stay human.
• AI for ABA Therapy Practices: Seven Admin Jobs and Three to Keep Human — Seven admin jobs an ABA practice can give to AI, three that stay with the BCBA, and the business associate agreement check before any client data moves.
• Will Payers Accept AI-Written Physical Therapy Documentation? — Payers judge the note, not the tool: what reviewers look for, and the review, signature and audit-trail routine that makes AI drafts safe.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT Business pricing page (openai.com) and Claude pricing page (claude.com), both checked 4 November 2026, for no training on business content by default. Payers publish their own prior authorisation forms and rules, which differ and change, so every threshold and timescale here is illustrative and must be replaced from each payer's current provider pages. Workflow design is the author's own.