Study highlights potential and risks of AI dental tools

Dental Tribune International

Mon. 17 November 2025

save

LONDON, England: One of the key promises of artificial intelligence-based (AI-based) tools in dentistry is the expediting of time-consuming tasks. The authoring of electronic health records is one area where AI could enable significant time-savings; however, clinicians must be able to rely on consistent and error-free documentation. According to a team of researchers who tested the accuracy of various AI-based automatic speech recognition (ASR) systems, current systems still require clinical oversight.

Aiming to assess the lexical, transcriptional and semantic accuracy of ASR tools in clinical settings, the researchers dictated 200 orthodontic clinical records that incorporated a range of technical and dentistry-specific language and used the recordings to test ten ASR systems. Four of the systems were commercially available: Heidi, DigitalTCO, Dragon Medical One and Dragon Professional Anywhere. Another five were application programming interfaces that link different software apps: Amazon, Google, Speechmatics, Whisper and GPT-4o Transcribe. Finally, one coupled the GPT-4o Transcribe app interface with the GPT-4o model used for error correction, called GPT4oTranscribeCorrected.

The researchers found that formatting and minor grammatical errors were common in transcripts generated by all of the systems, as were Class 3 errors, meaning-altering mistakes that had the potential to negatively affect clinical care. Class 3 errors accounted for between 0.21% and 4.15% of the total errors made by the systems. GPT4oTranscribeCorrected generated the fewest Class 3 errors, 2% of its transcripts featuring at least one, and Dragon Medical One generated the most, 66% of its transcripts featuring at least one. Furthermore, the systems had difficulty with many dentistry-specific terms. For example, “Essix” was incorrectly transcribed in 97.5% of instances. Other terms such as “palatally”, “mesially” and “buccally” were misinterpreted in most cases as well. The researchers found that speaker accent had only a minor link with errors made and that background noise in the recordings led to poorer performance in all systems tested.

Discussing the results, the researchers outlined the potential of ASR systems to streamline clinical documentation, and they highlighted improvements that could be made to the ASR systems, such as the use of confidence indicators to flag potentially incorrect terms. However, they emphasised that “the most important safeguard is maintaining a ‘human-in-the-loop’ workflow to verify transcripts, as clinicians move from authors to editors of their notes”.

Lead author of the study Dr Ruairi O’Kane, a researcher at the Centre for Craniofacial and Regenerative Biology at King’s College London, said in a university press release: “AI speech tools can streamline documentation and improve efficiency, but we must remain vigilant. Even subtle transcription errors can potentially impact patient care.”

The study, titled “Transcription accuracy of automatic speech recognition for orthodontic clinical records”, was published online on 3 November 2025 in Journal of Dental Research, ahead of inclusion in an issue.

Topics:

General Dentistry

Tags:

Artificial intelligence Dental research Documentation Electronic health records

To post a reply please login or register

Future-ready dental graduates: Embracing the next wave of digital transformation in Australia

MELBOURNE, Australia: Encompassing innovations such as virtual care, remote monitoring, mobile health apps and artificial intelligence (AI), digital health ...

FDI and Henry Schein working to introduce integrated electronic health records to dentistry

GENEVA, Switzerland: Electronic health records (EHRs) have the potential to greatly improve the overall quality of dental care and research. FDI is ...

Dentsply Sirona and Pearl partner to deliver AI-assisted radiographic support

CHARLOTTE, N.C., US: Dentsply Sirona, the world’s largest diversified manufacturer of dental products and solutions, and Pearl, a leader in dental ...

AI-generated versus human-crafted smiles: Aesthetic preferences in digital smile design

MANGALORE, India: Digital smile design has begun employing artificial intelligence (AI), raising questions about the quality and acceptability of ...

The urgent ethical challenges of AI-driven caries detection

KARACHI, Pakistan: Artificial intelligence (AI) is transforming dentistry, offering powerful tools to detect caries with speed and precision once ...

Artificial intelligence in prosthodontics: From concept to clinical reality

Artificial intelligence (AI) has moved from theory into daily prosthodontic practice, reshaping diagnosis, treatment planning and design workflows. My ...

Education

Live webinar
Mon. 17 November 2025
5:00 pm UTC (London)