The Ontario auditor general published a report in May 2026 on the use of AI medical scribes in healthcare that uncovered widespread issues with accuracy and completeness across all 20 vendors reviewed [1].
The audit tested each AI scribe in simulated patient-doctor conversations and found that every system made errors in at least one scenario [1]. Nine of the AI scribes hallucinated patient information, which means they fabricated or inserted data that was not present [1]. Twelve vendors incorrectly transcribed patient details, such as medication names or referral notes [1].
Significantly, 17 AI scribes failed to capture critical mental health details during the simulations, missing information that could impact diagnosis and treatment [1]. The errors ranged from hallucinated referrals to incorrect medication information, raising concerns about risks to patient care [1].
The report concluded that these inaccuracies could lead to inadequate or harmful treatment plans, potentially affecting patient health outcomes [1].
An international technology news outlet also reported on the audit, citing concerns over the AI scribes’ mistakes with basic patient facts [2].
The audit reflects thorough testing of Ontario government-recommended AI vendors, highlighting the need for more rigorous validation before integrating these tools into clinical workflows [1]. No specific timeline was provided for next steps or responses from vendors.
