Artificial intelligence is often described as one of the most transformative technologies in modern medicine. Its strongest advocates believe AI could eventually help diagnose diseases earlier, accelerate drug discovery, and even contribute to cures for some of the world’s most serious medical conditions.
That future may indeed be possible. But the present reality is far less polished.
One area where AI is already being deployed widely is medical transcription: tools designed to listen to conversations between doctors and patients and automatically generate clinical summary notes. In theory, these systems reduce administrative burden, save physicians time, and allow healthcare professionals to focus more directly on patient care.
In practice, however, many of these tools remain unreliable.
In a recent special report, Ontario’s Auditor General’s Office provided concerning details from its investigation into AI use in the public sector, including AI tools used by medical professionals to summarize information discussed during meetings with patients. The Auditor General discovered that the province’s procurement agency had approved vendors offering AI Scribe systems that generated inaccurate medical notes which included incorrect information, AI hallucinations, and incomplete information.
Spokespersons from the Ministry of Public and Business Service Delivery and Procurement and the Ministry of Health emphasize safeguards are in place, including the requirement that doctors to manually review AI-generated notes. Officials say there have been no reports of patient harm linked to the approximately 5,000 physicians in the province currently using AI “scribe” tools.
In this blog post, I explore how AI is reshaping the way some physicians document patient interactions, examine the findings highlighted in the Auditor General’s report, and outline key questions patients may want to ask if they learn their doctor relies on AI transcription or “scribe” technology during appointments.
How Do Doctors Take Notes?
Creating written records of medical visits is an essential part of healthcare.
Physicians and other medical professionals document patient symptoms, clinical observations, test results, diagnoses, and treatment plans to ensure continuity of care and to monitor changes in a patient’s health over time. These records are often essential for future appointments, specialist referrals, prescriptions, and emergency treatment decisions.
At the same time, maintaining detailed documentation can be time-consuming. Administrative workload has become a major source of frustration for many physicians and is frequently cited as a contributor to professional burnout.
To be more efficient, some medical professionals use:
- shorthand abbreviations or symbols for common medical terms
- voice to text dictation software
- structured SOAP notes to organize information. SOAP stands for Subjective (patient information), Objective (clinical findings), Assessment (diagnosis), and Plan (prescriptions, referrals and/or follow-up)
AI-powered scribes are the most recent addition to the toolkit. These tools use AI technology to transform ambient conversations between physicians and patients into a text record, but unlike older “voice to text” programs, they can organize and summarize information the doctor can review before filing it into a patient’s records.
Every form of medical documentation has the potential to introduce errors into a patient’s record. Handwritten notes can be illegible or misread. Dictation software can misunderstand words or omit information. Physicians themselves may occasionally make incomplete or inaccurate entries.
AI systems add another layer of concern: the possibility of “hallucinations,” where software generates information that sounds plausible but is incorrect, incomplete, or entirely fabricated. In a healthcare setting, such inaccuracies can carry serious consequences, particularly if flawed records influence future diagnoses or treatment decisions.
What Did The Auditor General’s Office Report Say About AI Scribes?
Auditor General (AG) Shelley Spence broadly investigated AI use in Ontario’s public service and public sector institutions, including Supply Ontario, the province’s procurement agency. While reviewing its pre-approval process for vendors, the AG’s Office found fault with all 20 AI scribe system approved vendors. Every product demonstrated at least one inaccuracy, if not more, during the procurement testing phase, including:
- Hallucinations – Nine (9) out of 20 vendors’ systems fabricated information, including recording suggestions for treatment plans not discussed. For example, the AI scribe notes indicated suggestions made to refer patients to therapy or to order blood tests. The AG report notes: “Evaluators also noted hallucinations that could impact patients’ health. For example, the submitted notes included statements that there were ‘no masses found’ or that there was presence of anxiety in the patient, although this information was not discussed in the recordings.
- Incorrect information: 12 out of 20 vendors’ systems generated notes which captured a different drug than prescribed.
- Incomplete information: 17 out of 20 vendors’ systems missed key details about mental health conditions mentioned in one of two simulated recordings, while six (6) out of 20 vendors’ systems partially or fully missed generating notes of these key details across two tests.
Additionally, 11 of 20 approved vendors failed to submit any third-party audit reports, System and Organization Controls reports, or International Organization for Standardization (ISO) 27001 certification, and five vendors failed to submit threat risk assessments and privacy impact assessments as required by the request for bids.
AG Spence was reportedly concerned enough that she explicitly asked her own doctor to review AI scribe notes after a recent appointment. Although healthcare workers using tools such as AI Scribes must still meet their obligations under law and according to professional standards guidelines, including manual review of system generated notes, the AG report highlights that doctors are not required to attest they had verified system generated notes through a sign-off feature in the AI Scribe systems.
The report also delved into privacy concerns with these programs, noting that there was no supporting evidence to confirm patient data accessed by these systems was processed and stored in Canada.
How Have The Government And Critics Of The Report Responded?
During a press conference, Ontario’s Minister of Public and Business Service Delivery and Procurement- Stephen Crawford, told reporters that the errors noted in the AG report were discovered during the testing phase of the process and that doctors using this type of product “oversee every aspect of it.” “So, every decision that is made that comes out of any artificial intelligence anywhere is overseen by a professional.”
A spokesperson for the Ministry of Health stressed that using these products is “entirely voluntary and requires patient consent, including informing patients how AI will be used, before it is used during an appointment.”
In an article published by Canadian Healthcare Technology, CSA Group public policy fellow Will Falk, author of a recent report on AI in Canadian healthcare, stressed that the industry has known for years that 2024-era scribes hallucinated and that “every serious deployment assumes clinician review of AI generated notes.”
Falk said the government and its medical partners “should be congratulated on this program, not subject to scary headlines and attacks by ill-informed accountants.” He added that “twenty-eight (28) per cent of Canadian physicians now use AI scribes, three-quarters of them daily, and the College of Family Physicians of Canada is asking Ottawa to fund scribes for every family practice. The policy question is no longer whether ambient AI belongs in Canadian healthcare. It is whether we govern it formally or push clinicians back into unmanaged grey usage with consumer tools.”
Dr. Paul Forman, a family physician from Markham was quoted in the same article “I strongly believe these systems have tremendous value and can significantly augment physicians, improve workflow efficiency, reduce cognitive burden, and support safer, more guideline-aligned care. However, responsibility ultimately still rests with the physician. AI should be viewed as another clinical tool – whether an AI scribe or clinical decision support system – that can help make us more effective and informed, but never replace physician judgment, accountability, or careful review.”
Should Ontarians Be Concerned About These Findings?
Both humans and artificial intelligence systems are capable of making mistakes. Errors in patient records existed long before the introduction of AI scribes, and inaccuracies will continue to occur regardless of whether these technologies are used.Falk’s assertion that AI belongs in the healthcare system, and the only question now is whether governments regulate it or allow unmanaged consumer tools to proliferate, assumes that policy makers have enough reliable information to make informed decisions about how these systems should be used and governed.
Beyond the transcription errors identified in the Auditor General’s report, concerns remain about the “black box” nature of many AI systems.
This lack of transparency matters because subtle patterns in AI-generated documentation could influence patient care in unintended ways. Systems may include excessive irrelevant information while overlooking details that later prove medically significant. The risks may also fall unevenly across different populations. Patients who already face barriers in healthcare — including linguistic, cultural, or socioeconomic obstacles — could be disproportionately affected if AI systems misunderstand speech patterns, accents, or communication styles.
In fact, studies have already shown that accented speech can increase error rates in both traditional voice-to-text systems and newer AI scribe technologies.
There are also broader concerns about how increased reliance on AI may affect physicians themselves. While automation may reduce administrative burdens and help address burnout, emerging research suggests there can be trade-offs when people outsource cognitively demanding tasks to machines. Over time, excessive dependence on AI systems for summarization, analysis, or decision support may contribute to a gradual erosion of professional skills and critical engagement — a phenomenon some researchers describe as cognitive atrophy.
What Can You Do If You’re Concerned About AI Scribes?
In Ontario, patients must be informed if a medical provider wants to use an AI scribe to record and summarize their visit, and they must give their consent.
If you are concerned about the accuracy of this tool or any potential for a breach of privacy, you can ask your medical provider questions to determine whether this technology’s benefits will ultimately outweigh its potential for harm.
Some potential questions may include:
- When do you review the AI Scribe’s summary? (Immediately after meeting the patient, on breaks between patients, at the end of the workday?)
- Is the original ambient sound recording stored and accessible for the doctor’s reference?
- Do you conduct occasional spot checks to determine if your original notes of the patient encounter are comparable to what the AI scribe has produced?
- Is the data processed and stored securely in Canada?
- Is the data anonymized? Is identifying patient information kept separate from the summary note produced?
- If I do not consent to using the AI Scribe, how will you produce notes from this meeting?
- Can I review notes created by the AI scribe? What is the process for correcting any inaccuracies?
We’re Here To Help.
No healthcare system – and no healthcare professional – can avoid errors 100% of the time. However, when using a novel tool it is essential that appropriate safeguards are in place to mitigate the potential for errors.
If you or a loved one has been seriously injured due to a medical error, the team of medical malpractice lawyers at Gluckstein Lawyers can inform you about your legal rights and options. Please contact us to schedule a no cost, no obligation initial consultation and tell us what happened.