· Apr 30, 2026
Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli argue in Google DeepMind for an AI co-clinician
Markdown · JSON · Original source
Published April 30, 2026, Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli describe ’s AI co-clinician research initiative and hypothesize that “triadic care” could place alongside patients and physicians, under physicians’ clinical authority. They present the system as a way to amplify doctors’ expertise while keeping clinical judgment with physicians. The authors report zero critical errors in 97 of 98 primary care queries. But in simulated telemedical consultations, expert physicians performed better overall, particularly at identifying “red flags” and guiding critical physical examinations. The authors say these systems are best used to support practitioners, not replace clinical judgment. Their case for an AI co-clinician rests on its role as a supervised member of the care team; they describe research, not demonstrated care delivered to patients. The page does not report results from the further evaluations the authors say they are advancing.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Apr 30, 2026 (2026-04-30T12:14:15.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-07T07:28:24.010Z. First observed: 2026-09-23T06:16:54.953Z. Retrieved: 2026-10-24T06:01:37.583Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 24, 2026; originally published April 30, 2026. The telemedical evaluation described was a simulation using hypothetical encounters and patient-actors, not care delivered to real patients. The authors describe phased collaborations and future evaluation; the page does not report results from those further evaluations. The authors state that the research collaborations are not intended for diagnosis, treatment, or medical advice.
Open question
How will further evaluations across different healthcare settings assess whether the system can support care while remaining under expert supervision?
Source-stated
The authors present AI co-clinician research as an effort to explore AI support for doctors and patients.
That's why, today, we are announcing our AI co-clinician research initiative, to explore how AI could better amplify doctors’ expertise and deliver higher quality care to patients.
Source-stated
The authors hypothesize that AI agents could support patients under their physicians’ clinical authority.
We hypothesize that the next evolution of healthcare delivery will entail “triadic care” where AI agents can help patients in their care journeys under the clinical authority of their physician.
Source-stated
In objective analysis, the system recorded zero critical errors in 97 of 98 primary care queries.
In objective analysis of 98 realistic primary care queries, our system recorded zero critical errors in 97 cases, improving over two AI systems widely used by physicians.
Source-stated
In simulated telemedical consultations, expert physicians performed better overall than the AI system, particularly in identifying red flags and guiding critical physical examinations.
We assessed over 140 aspects of consultation skill and found that expert physicians performed better than the AI system overall, particularly in identifying "red flags" and guiding critical physical examinations.
Source-stated
The authors say the system is currently best used to support practitioners rather than replace clinical judgment.
This finding suggests these systems are currently best used as supportive tools for practitioners rather than replacements for clinical judgment.
Source-stated
The authors say their research collaborations are not intended for diagnosis, treatment, or medical advice.
Our research collaborations are not, at this stage, intended for use in the diagnosis, cure, mitigation, treatment, or prevention of disease, or to provide medical advice.
Source and contributor context
Google DeepMind · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
AI agent
An AI system that chooses and carries out actions toward a goal, often using software tools.
Google DeepMind
Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Editorial provenance
Version 1. Publication ID: e96c261f-8236-0296-95fe-444c9962cde6.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-07T07:23:40.436Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: ce3fc8ecde7d8d973c9c2681eeb3c244381e878e585f66b2ad11144d11fc4173. Source text hash: bb516f312587983508789a7f85ceb79497553a9f55502bf1e6044bf7fa709753.
Revision and source-check history
Version 1 · 2026-11-07T07:28:24.010Z
Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli argue in Google DeepMind for an AI co-clinician
Published April 30, 2026, Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli describe ’s AI co-clinician research initiative and hypothesize that “triadic care” could place alongside patients and physicians, under physicians’ clinical authority. They present the system as a way to amplify doctors’ expertise while keeping clinical judgment with physicians. The authors report zero critical errors in 97 of 98 primary care queries. But in simulated telemedical consultations, expert physicians performed better overall, particularly at identifying “red flags” and guiding critical physical examinations. The authors say these systems are best used to support practitioners, not replace clinical judgment. Their case for an AI co-clinician rests on its role as a supervised member of the care team; they describe research, not demonstrated care delivered to patients. The page does not report results from the further evaluations the authors say they are advancing.
Original source
· Article
Source details
Original source date: Apr 30, 2026 (2026-04-30T12:14:15.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-07T07:28:24.010Z. First observed: 2026-09-23T06:16:54.953Z. Retrieved: 2026-10-24T06:01:37.583Z.
Limits
Text as retrieved on October 24, 2026; originally published April 30, 2026. The telemedical evaluation described was a simulation using hypothetical encounters and patient-actors, not care delivered to real patients. The authors describe phased collaborations and future evaluation; the page does not report results from those further evaluations. The authors state that the research collaborations are not intended for diagnosis, treatment, or medical advice.
Open question
How will further evaluations across different healthcare settings assess whether the system can support care while remaining under expert supervision?
Source-stated
The authors present AI co-clinician research as an effort to explore AI support for doctors and patients.
That's why, today, we are announcing our AI co-clinician research initiative, to explore how AI could better amplify doctors’ expertise and deliver higher quality care to patients.
Source-stated
The authors hypothesize that AI agents could support patients under their physicians’ clinical authority.
We hypothesize that the next evolution of healthcare delivery will entail “triadic care” where AI agents can help patients in their care journeys under the clinical authority of their physician.
Source-stated
In objective analysis, the system recorded zero critical errors in 97 of 98 primary care queries.
In objective analysis of 98 realistic primary care queries, our system recorded zero critical errors in 97 cases, improving over two AI systems widely used by physicians.
Source-stated
In simulated telemedical consultations, expert physicians performed better overall than the AI system, particularly in identifying red flags and guiding critical physical examinations.
We assessed over 140 aspects of consultation skill and found that expert physicians performed better than the AI system overall, particularly in identifying "red flags" and guiding critical physical examinations.
Source-stated
The authors say the system is currently best used to support practitioners rather than replace clinical judgment.
This finding suggests these systems are currently best used as supportive tools for practitioners rather than replacements for clinical judgment.
Source-stated
The authors say their research collaborations are not intended for diagnosis, treatment, or medical advice.
Our research collaborations are not, at this stage, intended for use in the diagnosis, cure, mitigation, treatment, or prevention of disease, or to provide medical advice.
Source and contributor context
Google DeepMind · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
AI agent
An AI system that chooses and carries out actions toward a goal, often using software tools.
Google DeepMind
Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Editorial provenance
Version 1. Publication ID: e96c261f-8236-0296-95fe-444c9962cde6.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-07T07:23:40.436Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: ce3fc8ecde7d8d973c9c2681eeb3c244381e878e585f66b2ad11144d11fc4173. Source text hash: bb516f312587983508789a7f85ceb79497553a9f55502bf1e6044bf7fa709753.
passed · 2026-11-07T07:23:40.436Z · anthropic/claude-opus-5.5
Publication: e96c261f-8236-0296-95fe-444c9962cde6
attribution — pass: The reading correctly credits Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli, writing on Google DeepMind's blog. It keeps the reported results and hypotheses as the authors' statements. The page is a single article, not a listing.
Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli
summary — pass: The reading names the actor and the initiative, and explains triadic care in plain terms. It balances the 97-of-98 query result against physicians outperforming the system in the simulations. Minor omission: it leaves out the authors' point that the AI matched or exceeded primary care physicians in 68 of 140 areas, which softens the source's emphasis on progress. That does not prevent understanding. The question is a genuine open issue.
AI co-clinician performed at a level comparable to or exceeding primary care physicians (PCPs) in 68 of the 140 assessed areas.
date — pass: The evidence instant 2026-04-30T12:14Z falls on April 30, 2026 (UTC), matching the source dateline. Planned evaluations are framed as future work with no reported results, and there is no hindsight framing. The evidence is a publisher page, so no archive disclosure is needed.
April 30, 2026 Science
support — pass: Each claim matches a verbatim excerpt. The limitations about simulated encounters, patient-actors and non-clinical use are source-supported. The statement that the page reports no results from the further evaluations is accurate. No benefits are overstated.
we are currently advancing a phased approach with academic and research collaborators across globally diverse healthcare settings
media — pass: No media is selected, and no verified candidates were supplied, so nothing was invented.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMind