Health · Medical Research · 6 hrs ago
Study warns AI tools often misidentify validated blood pressure monitors
Researchers at the University of Montreal tested Google Gemini, ChatGPT, Copilot and Perplexity on 324 home blood pressure monitors.
Of those devices, 179 had not been validated for accuracy.
Three tools gave correct recommendations between 63% and 83% of the time, while Gemini was correct between 86% and 91% of the time.
Even the best-performing tool was often wrong, the researchers said.
An unvalidated monitor may give inaccurate readings, which could lead to incorrect diagnoses or treatment decisions.
The American Heart Association says people should choose a validated device listed on validatebp.org and use it correctly, with regular monitoring as part of their care.
A study found that three of four AI search tools often recommended home blood pressure monitors that had not been validated for accuracy.
Researchers tested Google Gemini, ChatGPT, Copilot and Perplexity on 324 monitors, including 179 that had not been validated.
Copilot, ChatGPT and Perplexity gave accurate recommendations 63% to 83% of the time, while Gemini was accurate 86% to 91% of the time.
Researcher Anna Soriano warned that inaccurate readings could lead to inappropriate diagnoses or treatment decisions.
The findings were presented at the American Heart Association’s annual hypertension meeting.
- Who
- Researchers, including Anna Soriano of the University of Montreal, tested the tools.
- What
- A study found that AI tools often misidentified whether blood pressure monitors had been validated for accuracy.
- When
- The findings were presented on Thursday. The article was published on October 12, 2026.
- Where
- At the American Heart Association’s annual hypertension meeting.
- Why
- Researchers said people may use AI to help choose home blood pressure monitors, and inaccurate readings could affect diagnosis or treatment.
This story does not have two clearly opposing sides.
Researchers asked four AI tools about 324 home blood pressure monitors, including 179 that had not been validated for accuracy.
The study findings were presented at the American Heart Association’s annual hypertension meeting.
- Monitors tested
- 324 home blood pressure monitors
- Not validated
- 179 monitors
- Tools tested
- Google Gemini, ChatGPT, Copilot and Perplexity
- Accuracy range
- Copilot, ChatGPT and Perplexity: 63% to 83%; Gemini: 86% to 91%
- Validation registry
- validatebp.org
What people are saying
We found that most AI tools performed only slightly better than if you had flipped a coin for each question
Using that device could result in inaccurate blood pressure readings, which could lead to inappropriate diagnosis or treatment decisions
In addition, home blood pressure devices must be both validated -- listed on the national registry validatebp.org -- and accurate
Sources
Related news
Pope Leo to have surgery to remove an early-detected lung nodule
Pope Leo XIV to have an early-stage lung nodule removed
Bolivian medical writing cast Indigenous people as both disease risks and resilient
Author describes spending hours on new NIH biographical sketch forms
Pope Leo XIV to have surgery to remove early-stage lung nodule
Pope Leo XIV to have surgery for an early-stage lung nodule
UK doctors transplant frozen testicular tissue to try to restore fertility
Early lung nodule removal can offer high chance of cure, oncologist says
Digital tools improve asthma treatment adherence in children
Early statin use linked to lower dementia risk in people with type 2 diabetes
Investigational tracer detects early cardiac amyloidosis in phase III study
