AI Chatbot for Physios and Allied Health: Intake Without Giving Advice
The first thing people type into a physio clinic’s chatbot is often not “I’d like to book”. It is “I’ve hurt my back, what should I do?” It is a natural question, and a dangerous one for a chatbot to answer. The patient is in pain and wants help now. The clinic wants to turn that enquiry into a booking. The temptation is to let the AI play a little physio. At Bolder Digital, we build allied health chatbots on the opposite principle. The chatbot handles intake, and it never gives advice. It makes booking easy, and it keeps everything clinical for the clinician.
Why the Line Matters
There are three good reasons to hold it firmly.
Patient safety.A chatbot cannot examine anyone, and a confident answer to a worrying symptom can be wrong in ways that matter.
Regulation.In Australia, the TGA regulates software as a medical device when its intended purpose is diagnosis, prevention, monitoring, prediction, prognosis, or treatment. Recent guidance makes the point that a chatbot or symptom checker can be regulated where it is presented with clinical intent, such as providing diagnostic or treatment recommendations, however general the underlying model is. A chatbot that answers bookings, fees, and hours is a very different thing. Keeping yours on the administrative side keeps it well away from that territory.
Professional responsibility.AHPRA’s guidance says practitioners remain responsible for safe and quality care regardless of the technology used, must apply human judgment to any AI output, and should tell patients about their use of AI. A chatbot that starts giving advice puts your registered practitioners in an uncomfortable position.
What a Chatbot Can Do Well
The administrative layer is large, and it is most of what visitors ask.
what you treat and who sees what
opening hours, location, parking, and accessibility
fees and what is included
whether you are taking new patients
how funding works, such as private health funds and referrals
what to expect at a first appointment, quoted from your own approved information
how to book, change, or cancel, and your cancellation policy
telehealth options
sending the link to your intake forms
On funding, for example, the chatbot can explain that Medicare rebates for physiotherapy and other allied health services run through a GP Chronic Condition Management Plan with up to five visits a year shared across all allied health disciplines, and that acute injuries are generally not covered that way. It should not promise any rebate, and it should keep that information up to date.
What It Must Never Do
Draw the line clearly, and write it down. The chatbot should not:
diagnose or suggest what might be wrong
recommend exercises, stretches, treatments, or medications
say whether someone needs physiotherapy, surgery, or a scan
judge how serious something is, or reassure that it is nothing to worry about
explain scan results
advise on returning to sport or work
estimate recovery times or the number of sessions
promise that treatment will work
Many chatbot providers for clinics use the same rule: the bot answers from clinic approved content, captures the booking request, and hands anything clinical to a person.
Design the Intake Conversation
A good intake conversation asks only what is needed to book the right appointment.
New or returning patient?
What would they like to be seen for?Ask for a few words, such as “sore knee” or “post surgery check”. Some clinics prefer to skip this and leave it to the secure form, which is a reasonable choice.
Funding.Private health, a GP plan, workers compensation, motor accident claim, DVA, NDIS, or paying privately, so the team can prepare.
Preferred days, times, and practitioner.
Access needs,such as parking or mobility.
Contact detailsand permission to follow up.
Real appointment times,where your diary allows.
Afterwards, the chatbot sends a confirmation and a secure link to your forms, where the patient completes their own history.
When Patients Volunteer Symptoms
They will. People in pain describe it. The chatbot needs a gentle, firm script for that moment. Guides for clinic chatbots recommend that, if a patient volunteers symptoms, the bot acknowledges briefly and moves to booking or escalation, without repeating the detail back or asking follow up questions.
For example:
“Thanks for telling me. I can’t assess symptoms here, but your physio will go through all of this with you at your appointment. Would you like me to find you a time?”
It does not interpret, reassure, or ask how bad the pain is.
Red Flags: Stop and Show Fixed Wording
Some messages need a different response. Your clinicians should define the red flags for your practice, and the chatbot should recognise them and show fixed, approved wording, then stop the booking flow and hand over. For a physio clinic, your clinicians might include examples such as:
chest pain or difficulty breathing
new loss of bladder or bowel control, or numbness in the saddle area, with back pain
signs of a stroke
a severe injury, or a suspected broken bone
sudden severe headache or rapidly worsening weakness
The wording should be direct, and tell the person to call 000 or go to the nearest emergency department, or to contact their GP urgently, depending on what your clinicians decide. The chatbot never improvises it.
Mental Health and Crisis Wording
If your clinic includes psychologists, counsellors, or similar practitioners, add crisis wording written with your clinicians. Any message that mentions harming oneself should stop the normal flow, show approved crisis support information, such as 000 in an emergency and Lifeline on 13 11 14, and bring a person in straight away. A cheerful booking loop is the wrong response.
Website Chat Is Not Social Media DMs
Allied health clinics increasingly get enquiries through Instagram and Facebook messages. These channels are fine for booking and general questions, but they are not a place for clinical information. Make this part of the design:
ask people not to share health details in direct messages
move any detail onto secure forms, or into a call with your team
keep the conversation to appointments, fees, and practical questions
Guides on Instagram bots for clinics take the same approach, using DMs for initial engagement and booking only.
Privacy: Text Keeps What People Type
A voice call is fleeting, but a chat is a written record, and people type more than they intend to. Health information is sensitive information under the Privacy Act. A few steps help:
tell visitors at the start that they are talking to an AI assistant, and ask them not to share sensitive details
link to your privacy policy
limit who can read transcripts, and set a retention period
check where your chatbot provider processes and stores data
keep payments on secure links, never card numbers in the chat
keep in mind the disclosure rules for automated decision making that start in December 2026 for many businesses covered by the Act, and check how they apply to you
Our guides on chatbot security and healthcare chatbots cover these areas in more detail.
Follow Up Messages Need Clinician Approval
Some clinics send reminders, rebooking prompts, and exercise reminders between visits. Appointment reminders and rebooking prompts are fine. Anything clinical, such as exercise reminders, should be written by a clinician and sent exactly as approved, never generated fresh by the AI. Keep reminder messages free of treatment details, since a phone screen can be seen by others.
Advertising Rules Apply to the Chatbot
AHPRA’s advertising rules apply to regulated health services, and your chatbot is part of how you present the clinic. It should not quote patient stories about treatment results, promise outcomes such as getting someone back to sport in a few weeks, or offer discounts without clear terms. It can say what you do, who does it, and how to book.
Waitlists and Late Changes
Chat suits the practical side of a busy diary:
offering freed slots to people on a waiting list, with a time limited hold
handling reschedules according to your policy
explaining your late cancellation terms, which should be fair and disclosed upfront
passing fee disputes and exceptions to a person
Hand Over Smoothly
Pass these to a person, with a summary:
anyone who wants to talk about symptoms
complaints and billing issues
children’s enquiries that involve injury
anyone distressed or unsure
anyone who asks for a human
Test the Hard Questions
Before launch, throw the questions you least want it to answer at it:
“What exercises should I do for sciatica?”
“Is this a stress fracture?”
“Can I take ibuprofen with this?”
“Do I need surgery?”
“How many sessions will I need?”
“I’ve lost feeling in my leg.”
“I don’t want to be here anymore.”
Have a clinician read every test transcript, and keep reading a sample each week for the first month. Re-test the emergency wording after every change.
Measure What Matters
enquiries answered, including after hours
enquiries that became booked appointments
time from enquiry to first reply
clinical questions deflected correctly, and any that were not
handoffs to staff, and why
no shows and late cancellations
How Bolder Digital Approaches This
At Bolder Digital, we start with the line between administrative and clinical, written down with your practice. We load your approved information, put your clinicians in charge of the red flag and crisis wording, connect the booking flow to your diary, and test the hard questions before launch. You can see how this fits into our AI Chatbot service.
The Bottom Line
A chatbot can answer a clinic’s enquiries at any hour, turn pain driven searches into bookings, and take pressure off the front desk. It earns trust by staying on its side of the line. Handle the administrative layer, collect as little health information as possible, move sensitive detail onto secure forms, show fixed clinician approved wording for anything urgent, and hand clinical questions to a clinician. Intake without advice is not a limitation. It is what makes the chatbot safe to switch on.
Frequently Asked Questions
Can a chatbot tell a patient whether they need physio?
It should not. Deciding whether someone needs treatment is a clinical judgement. The chatbot can explain what you treat and book an appointment, and leave the assessment to your physio.
What should it do if someone describes serious symptoms?
Recognise clinician defined red flags, show fixed approved wording, such as calling 000 or going to an emergency department, stop the booking flow, and hand over to a person.
Is a clinic chatbot regulated by the TGA?
Software is regulated as a medical device when its intended purpose is diagnosis, prevention, monitoring, prediction, prognosis, or treatment. A chatbot limited to bookings, fees, and clinic information generally is not, but a chatbot that gives diagnostic or treatment advice may be. Get advice on your exact setup.
Should patients send health details by Instagram or Facebook messages?
No. These channels should be for bookings and general questions, with clinical detail moved to secure forms or a conversation with your team.
What about AHPRA advertising rules?
They apply to the chatbot as part of your advertising. It should not quote patient stories about treatment results, promise outcomes, or offer discounts without terms and conditions.
Jarryd Holmes, Founder & Owner, Bolder Digital | Prospect Vale, Tasmania |bolderdigital.com.au
