AI Adoption in Allied Health

AI has moved from novelty to baseline in allied health. More than 60% of practitioners are now using it in some form, and Zanda’s own platform data shows that use deepening rather than leveling off.

This report draws on responses from 682 allied health practitioners across Australia, the UK, the US, and beyond to document what that shift actually looks like on the ground: who is using AI, what they are using it for, what results they are seeing, and what questions the profession has not yet answered.

Five findings define the current moment:

This survey was distributed to Zanda’s practitioner user base, and completed by 682 allied health practitioners between March and May 2026. This report also draws on platform data from Zanda itself. Survey findings and platform data are presented separately throughout: survey findings reflect the 682 respondents specifically, while platform data reflects Zanda’s broader user base of 44,000 practitioners.

Given sample sizes vary by country and subgroup, country-level and age-level findings should be read as directional rather than precise, particularly for markets with smaller samples.

Definition of “allied health”

This report uses “allied health” broadly, covering the range of specialties represented in our respondent base. Alongside psychology, counseling, physiotherapy, occupational therapy, speech pathology, and massage therapy, respondents also included podiatrists, dietitians, exercise physiologists, chiropractors, and practitioners across a range of complementary and allied disciplines. In some markets, including the UK, “allied health professions” refers to a narrower, formally regulated list of professions. Readers in those markets should note the broader scope used here.

Definition of “AI usage”

For the purposes of this survey, “using AI” means using any AI-powered tool, whether built into practice management software or a standalone product like ChatGPT or Claude, for tasks such as clinical note drafting, treatment planning, written communications, marketing, or admin work. It does not include AI features practitioners may be using without realizing it, such as spell-check or basic scheduling automation.

There is a version of the AI-in-healthcare story that positions the technology as something arriving from the outside, disrupting a stable profession, and forcing practitioners to respond. The data in this report tells a different story.

AI did not arrive uninvited in allied health. Practitioners invited it in, because the alternative, spending the equivalent of a part-time job each week on documentation, was no longer workable. The technology did not create a problem. It arrived at a moment when a long-standing problem had become acute enough that practitioners were willing to try something new.

Understanding that context is the key to understanding everything that follows. The adoption rates, the usage patterns, the beliefs about outcomes, the emerging concerns about consent and safety: all of it makes more sense when you start from the premise that AI entered this sector through the door of desperation rather than the door of innovation.

This report documents where the profession stands today. It is designed to be useful for practitioners, for practice owners, and for anyone building the tools and policies that will shape how AI develops in allied health over the next decade.

To understand why AI entered allied health so quickly, you have to understand what practitioners were doing with their time before it arrived.

The cumulative picture is striking. A significant portion of the allied health workforce is spending hours each week on work that has nothing to do with the reason they entered their profession. And it has been getting worse, not better.

Each individual compliance change, each new funding body requirement, each insurer documentation standard seemed manageable in isolation. The aggregate effect is that practitioners in 2026 are carrying an administrative burden that would have been unrecognizable to the profession a decade ago.

One Australian psychologist and owner-practitioner said, “I would like a ratio of about 85% direct client contact and 15% admin, report writing, case notes, session preparation and communication. Currently it feels the opposite.”

The relationship between documentation burden and burnout is not subtle. The data shows burnout risk reaches nearly 50% among practitioners spending 21 or more hours per week on notes alone. The documentation crisis and the burnout crisis are connected at the root.

AI entered allied health through the documentation door, and that is still overwhelmingly where it lives. Zanda’s own usage data traces the same shape from the platform side. Over the past year, usage of our AI note-taking tool grew 88.2%, and usage of our AI note-editing tool grew 92.4%, both nearly doubling in twelve months.

Three quarters of AI users are using it for clinical note drafting. Nearly as many for written communications and reports. The secondary uses, particularly treatment planning and clinical decision support, represent where the technology is heading rather than where it currently sits.

At Zanda, our approach to AI has been shaped by what this data confirms: that practitioners do not need more technology for its own sake. They need fewer roadblocks between them and their clients. Every AI feature we build is tested against one question: does this give practitioners time back, or does it just add complexity?

The clinicians who most need AI are the ones too burned out to start

Not everyone who wants to use AI is using it. 12% of practitioners in this survey describe themselves as interested but not yet adopting. Their profile represents both the challenge and the opportunity for the sector.

These practitioners are convinced, just not yet converted. Getting this group from interested to active does not require more evangelism about AI’s potential. It requires accessible integration, a low barrier to entry, and an immediate practical payoff that justifies the short-term investment of learning time. The practitioners most likely to benefit from AI are precisely the ones least resourced to pursue it independently.

What is holding the remaining third back more broadly? The concerns that come up most consistently in the data are legitimate ones.

These are not reasons to avoid AI indefinitely. They are questions the profession needs better answers to, and reasons why the on-ramp to adoption matters as much as the technology itself.

The tipping point has passed. AI is no longer something allied health practitioners are considering. It is something most of them are already doing. And given how thoroughly AI has embedded itself into the tools practitioners use every day, the more surprising statistic is how many are not using it.

Nearly 40% of practitioners are not using AI. In a sector where the documentation burden is leading to greater burnout, and where the software practitioners rely on for scheduling, billing, and note-writing is embedding AI directly into core workflows, that number stands out.

More than half of non-users already feel behind on technology, and 12.2% describe themselves as interested but not yet using it, meaning the desire is there but the follow-through has not happened yet.

This tracks with what we see on our own platform. As of July 2026, 43.9% of companies using Zanda have BizzyAI features enabled, with 28.8% specifically using our AI note-taking tool. The company-level number sits below the survey’s individual-level adoption figure, which makes sense: company-wide feature rollout tends to lag individual practitioner behavior, and the survey captures use of any AI tool, not only ones built into Zanda specifically.

Allied health’s heaviest AI users are in their fifties

When practitioners and commentators imagine the early AI adopters in any profession, they typically picture the youngest cohort. The data tells us differently.

Daily AI adoption is remarkably consistent across age groups, sitting between 27% and 33% for every cohort except the 60+ group. Surprisingly, the 50-59 age group edges out under-30s as the heaviest daily AI users.

This may be because mid-career practitioners running established practices tend to carry the heaviest administrative loads. They have strong financial motivation to recover time, enough clinical experience to know exactly what they want AI to do, and enough practice infrastructure to implement it systematically.

The 60+ drop-off is the more notable shift, and it likely reflects proximity to retirement rather than any resistance to technology. Across the board, what varies more meaningfully than age is the context practitioners are working in and the pressures they are navigating.

Why AI adoption rates differ across the US, UK, and Australia

Among the three markets with sufficient sample sizes to report with confidence, there are meaningful differences in adoption that reflect structural differences in how healthcare is organized and funded.

Australia leads on daily use. On overall adoption, the US and Australia are close enough to call it even. The UK has the highest proportion of practitioners saying no outright, suggesting adoption is at an earlier stage there. What drives the differences across markets is difficult to isolate from this data alone, though the broader pattern is consistent with what we see elsewhere in the survey: where administrative pressure is highest, the pull toward AI tends to be strongest.

This may be the most important operational finding in this report. It’s worth noting this relationship likely runs in both directions: consistent use probably drives better results, but it’s also true that practitioners who see early wins are the ones who keep using AI daily, while those who don’t tend to fall back to occasional use.

Consistency and active experimentation matter more than which tools get used. But the easier the tool is to understand and the lower the barrier to entry, the more likely practitioners are to stick with it. The practitioners seeing transformational results have rebuilt their documentation process around AI rather than adding it as an occasional extra step.

AI isn’t working for everyone. 12.7% of AI users report no measurable reduction in documentation time at all. The ones seeing nothing have tried it a few times and moved on without changing their underlying workflow. This finding reframes the question practitioners should be asking. The question is not “does AI work?” The evidence suggests it does, substantially, for those who commit to it. The more useful question is “am I using it in a way that gives it a chance to work?”

The same principle holds when you look at practitioners’ general approach to technology.

Active experimenters are getting six times the documentation savings of reluctant adopters. This indicates that a posture toward adoption is the deciding variable.

Among practitioners using AI, only a third describe themselves as very or extremely confident. The largest group sits at moderately confident, and more than one in five report low or no confidence at all.

This matters because confidence shapes how practitioners engage with AI. Those who feel capable enough to rebuild their workflow around it tend to see transformational results, while those who feel uncertain often use it tentatively and see limited returns.

But confidence is not the only driver. The data also shows that urgency matters just as much. The 50-59 cohort uses AI most intensively despite being less confident than under-30s, suggesting that when the documentation burden is heavy enough, practitioners will commit to AI regardless of how fluent they feel.

Under-30s are the most confident age group, while the 60+ cohort is the least confident. These practitioners are using AI regularly but without full confidence in their ability to use it well. That gap between use and confidence is where the profession could do the most to support practitioners.

AI users are more likely to cut documentation time significantly than to see no change at all

The 20% no-reduction group in the US and UK is probably a mix of two things: practitioners using AI too occasionally to see returns, and practitioners whose documentation burden is dominated by compliance formats that AI tools don’t handle well yet.

The most plausible explanation connects to structural differences in how documentation is organized across markets, though the precise drivers are difficult to isolate from this data alone. What the finding does suggest is that AI documentation tools perform differently depending on the nature of the documentation they are being asked to handle. Where notes are primarily narrative and clinical, the tools tend to perform better. Where documentation is dominated by structured compliance requirements, insurer forms, and funding body reporting, the current generation of tools may be hitting a ceiling that better integration alone cannot solve.

The profession has arrived at a working consensus on AI as a productivity tool. It has not arrived at consensus on AI as a clinical one.

61.1% of practitioners believe AI improves their efficiency. Only 34.3% believe it improves patient outcomes. That 27-point gap is the most revealing finding in this section. Practitioners are broadly confident that AI saves them time and makes their work more manageable. They are uncertain whether that translates into better care for clients. When nearly half the profession answers “neutral” or “unsure” on a question this important, that response carries its own meaning. The jury is still out on what AI actually does inside the clinical relationship, and practitioners are saying so directly.

The picture shifts when you look at daily users specifically. Their confidence in AI’s efficiency benefits is near-universal, and their belief in client outcomes benefits is significantly higher than the overall average. Experience with AI generates both more confidence and more openness. The practitioners who have committed to integration are more optimistic about what AI can do across the board.

The youngest clinicians have already decided AI is good for patients (but the rest of the profession hasn’t)

The gap between generations on the patient outcomes question is one of the sharpest age-based divisions in the entire dataset.

Under-30s have effectively resolved this question for themselves. It’s important to note that this is among the smaller age cohorts in the survey, so read the sharpness of this split as a signal worth watching rather than a settled generational verdict.

Every other age group is still working through it. The 30-39 cohort is the most evenly split, suggesting that the practitioners currently in the thick of building their careers and their practices are uncertain about what AI means for the therapeutic relationship. What the data suggests clearly is that the profession’s future relationship with AI in the clinical context will be shaped significantly by the practitioners now entering the field. They are not waiting for the evidence to come in. They are proceeding on the basis that AI is a net positive for their clients, and the rest of the profession is watching to see if they are right.

There is also a clinical concern worth surfacing that sits underneath the numbers. The concern that appeared most frequently and most independently across the free-text responses was this: AI-generated clinical notes describe what was said in a session but not the clinical reasoning behind what the practitioner chose to do.

This distinction matters enormously in practice. Clinical notes are professional documents that record clinical judgment: the observations a practitioner made, the hypotheses they considered, the intervention they chose and why, and the outcomes they were tracking. If AI-generated notes capture the content of a session without the clinical architecture underlying it, they may create records that look thorough while missing the information that actually matters for continuity of care, supervision, and medico-legal protection.

A second concern appeared independently across multiple responses: the possibility that outsourcing note-writing to AI removes a cognitive process that matters for clinical development.

These concerns reflect the kind of careful clinical thinking the profession is known for. What makes a clinical note clinically adequate in the context of AI-generated documentation is not yet settled, and working through that question thoughtfully will matter for practitioners, supervisors, and regulators alike.

Clients are already using AI before they walk through their practitioner’s door. 70% of practitioners already have clients arriving having consulted AI first, and that number is expected to grow.

At the far end of the distribution, 6% of practitioners are already operating in an environment where AI-informed clients are the norm rather than the exception. That is where the broader profession is heading. The consultation dynamic has already changed for them. The practitioner is no longer the first source of health information their client encounters. Managing that shift, knowing when AI-sourced information is helpful context and when it is misleading noise that needs to be carefully unpacked, is becoming a core clinical skill.

The experience is not uniformly negative. Several respondents described clients arriving better prepared, with clearer language for their symptoms and more specific questions. AI can be a useful primer. The challenge is that its quality and accuracy varies significantly, and practitioners rarely have time in a consultation to assess which they are dealing with before the conversation is already underway.

More than a third of practitioners have already had a client push back on their clinical judgment, citing something an AI tool told them. There is no single right way to navigate a client who arrives having been told something different by an AI tool. Sometimes the AI was wrong and the practitioner needs to explain why. Sometimes the AI identifies something worth taking seriously. Sometimes the client’s trust in the AI is itself clinically relevant information about how they relate to authority and expertise.

The AI your client confides in at 2am has no duty of care, and allied health is sounding the alarm

The open-text responses in this survey raised three interconnected concerns about AI in the clinical relationship that go beyond the question of whether the technology is useful. They concern whether the profession has adequate frameworks to manage it safely.

Mandatory reporting

Multiple practitioners raised, independently and without prompting, a scenario that every mental health and child welfare practitioner needs to be thinking about. What happens when a client discloses suicidal ideation, a safeguarding concern, or an acute crisis to an AI chatbot they have been using as a therapy supplement between sessions?

The AI has no mandatory reporting obligations, no duty of care, and no ability to contact emergency services. It does not necessarily follow safe messaging guidelines and it may not reliably screen for risk. For clients who cannot afford frequent sessions, it may already be functioning as a primary point of contact for their mental health, available at 2am when a practitioner is not, and with none of the professional accountability that a practitioner carries.

This is not a hypothetical risk. It is a structural gap that exists right now, in every market this survey covers.

Client consent and data privacy

When a practitioner uses an AI tool to draft clinical notes from a session, is the client aware that the content of their appointment is being processed by a third-party system? Do they understand what data is being stored, where, and for how long? Does the informed consent process most practitioners currently use adequately cover AI processing of clinical information?

AI consent has become one of the more talked-about topics in allied health, but in practice it does not need to be complicated. At minimum, clients should know which AI tools are used in their care, what happens to the information those tools touch, and what their options are if they’d rather opt out.

Skill erosion

A third concern, raised independently by multiple respondents, is subtler but potentially just as significant. The cognitive work of clinical documentation, translating a session into written professional judgment, is itself a form of clinical practice. Removing that process from the workflow may have consequences for clinical development, particularly for early-career practitioners, that are not yet understood or measured.

The good news is that these conversations are already happening, everywhere and at every level. Practitioners who are engaging with AI, noticing what works, questioning what does not, and staying curious about what comes next are helping to define what responsible AI use in healthcare actually looks like.

We surveyed 682 allied health practitioners because we wanted to understand, rigorously and honestly, what AI is actually doing in their practices. The results surprised us in some places and confirmed what we have long believed in others.

The data in this report tells a story that is more optimistic than the headlines around AI in healthcare might suggest.

Allied health practitioners are using AI to take back the hours that documentation, administration, and compliance have been quietly consuming for years. The practitioners using AI most intensively embraced it because they needed to find a way to keep doing the work they love without losing themselves to the paperwork surrounding it. What the data shows, underneath the adoption rates and the usage statistics and the belief gaps, is a profession that is fundamentally resilient.

At Zanda, we believe AI has the potential to transform how allied health practices work, by removing administrative burden and giving practitioners more time and energy for the work that actually matters. That belief shapes everything about how we build.

BizzyAI, our AI suite built specifically for allied health, lives entirely within the Zanda platform. Client data never moves between systems, never passes through a generic AI tool, and is never used to train models. We hold ISO 27001 certification, the world’s leading standard for information security management, and we are independently audited for compliance with HIPAA, GDPR, and regional data protection standards across every market we serve. When a practitioner records a session and BizzyAI generates a note, that data is processed securely in real time with nothing retained afterward.

We built it this way exactly because of the concerns practitioners raise in this survey about data privacy and consent. Keeping AI functionality inside a purpose-built healthcare platform, rather than asking practitioners to copy and paste between tools, is how we address those concerns in practice.

The questions this report raises about consent frameworks, mandatory reporting gaps, and clinical documentation standards are ones we are actively thinking about too. We do not have all the answers, but we believe that practice management software has a responsibility to be part of the conversation, and to build products that make it easier for practitioners to do the right thing.

Allied health is navigating one of the most significant shifts in how care is delivered and documented that the profession has seen. The practitioners in this survey are doing that in real time, often alone and often under significant pressure. We hope this report is useful to them, and we hope the tools we build are too.

About the authors

Fiona Harrington

Head of Strategy, Marketing & Growth, Zanda

Fiona Harrington is a strategic marketing leader and entrepreneur with a passion for using technology to help small businesses run better. With an MBA from the University of Melbourne and experience at Accenture as a strategy consultant, she combines analytical thinking with a deep understanding of communication, leadership, and growth. Previously, Fiona co-founded Word of Mouth Online, Australia's leading business reviews platform, which she and her partner grew to over 400,000 reviews before its acquisition in 2015. Today, she channels her expertise into customer communications, building strong teams, and leading with purpose at Zanda.

Chloe Pierce

Brand & Content Lead, Zanda

Chloe Pierce is Brand and Content Lead at Zanda. She holds a degree in biological sciences and began her career as a research assistant at the Pew Charitable Trusts, before moving into strategic and educational content roles.