What exactly can ChatGPT for Clinicians do?

The new offering, which OpenAI officially announced on April 22, 2026, is not just a renamed ChatGPT Plus. It is a separate workspace optimized for healthcare tasks — from documentation through medical research to consultations directly in the moment of patient care. The system runs on the latest GPT-5.4 model, which has higher limits and more reliable outputs in healthcare settings than the standard version.

Among the key functions are:

• Trusted clinical search — responses with citations from millions of peer-reviewed medical sources, including journal names, authors, and publication dates.
• Skills for repeatable workflows — doctors can set up recurring tasks such as writing referral letters, prior authorization requests, or patient instructions.
• Deeper medical literature research — comprehensive overviews with citations within minutes, adjustable according to trusted sources.
• Automatic CME credits — verified clinical queries can automatically count toward physicians' lifelong education without the need for extra courses.

According to a 2026 survey by the American Medical Association (AMA), 72% of physicians already use artificial intelligence in clinical practice, an increase from 48% the previous year. OpenAI states that the use of ChatGPT in clinical settings has more than doubled in the past year.

Safety first: 99.6% of responses marked as safe

Healthcare is an area where tolerance for errors approaches zero. OpenAI therefore deployed a strict evaluation process: before launching the tool, medical advisors tested 6,924 conversations from common practice. The result was that 99.6% of responses were rated as safe and accurate.

In a specific test of 355 cases where three independent physicians established reference citations, ChatGPT for Clinicians cited sources more often than the human physicians themselves. Nevertheless, OpenAI emphasizes that the tool should serve as an information support, not replace professional judgment and clinicians' experience.

Regarding data protection, several key guarantees apply:

• Conversations are not used for training models.
• Multi-factor authentication is available to secure sensitive information.
• For accounts authorized to sign a Business Associate Agreement (BAA), support is available for HIPAA-compliant use when working with protected health information (PHI).

New HealthBench Professional benchmark

Along with the tool's launch, OpenAI introduced the open benchmark HealthBench Professional, which measures the performance and safety of large language models in three areas: clinical consultation, writing and documentation, medical research. Tests contain author-physician conversations, multi-stage expert assessment, and careful data filtering.

As a baseline, human physicians were asked to develop their own answers in their field with unlimited time and internet access. The results showed that GPT-5.4 in the ChatGPT for Clinicians environment surpasses both the base GPT-5.4 and all other external models as well as human physicians. OpenAI states that one-third of test examples included deliberate "red teaming" — attempts by physicians to find models' weak points.

On third-party sites such as Stanford MedHELM or MedMarks, OpenAI's models also placed in top positions in the area of real-world healthcare use.

Difference from ChatGPT for Healthcare

Readers may be confused by the similar names. While ChatGPT for Clinicians is intended for individual physicians whose hospitals or practices do not yet have centralized AI solutions, ChatGPT for Healthcare introduced in January 2026 is an enterprise product for entire healthcare systems. Among its first adopters are, for example, Boston Children's Hospital, Cedars-Sinai Medical Center, Stanford Medicine, Memorial Sloan Kettering Cancer Center, or UCSF.

Availability in the Czech Republic and Europe: Still waiting

For Czech readers, the key information is about availability. ChatGPT for Clinicians is currently freely available only to verified clinicians in the United States — physicians (MD/DO), nurse practitioners, physician assistants, psychologists, and pharmacists. Verification takes place via the national NPI identifier through a third party.

OpenAI has announced, however, that it plans expansion to other countries. In the coming months, it wants to launch a pilot program for verified clinics outside the USA in cooperation with the Better Evidence Network, of course in compliance with local regulations. For Czech and European physicians, this means that access is not ruled out, but the date is not yet known.

It is also necessary to mention the EU AI Act, which introduces strict requirements for AI systems used in healthcare — these fall into the high-risk category. Any tool like ChatGPT for Clinicians will have to meet regulatory criteria including transparency, risk oversight, and appropriate labeling to enter the European market. The Czech market additionally emphasizes GDPR and health data processing, which may further prolong the launch process.

What does this mean for Czech healthcare?

Even though Czech physicians cannot install ChatGPT for Clinicians yet, the trend is clear: administrative burden on doctors is growing and AI tools are becoming the standard. The American experience shows that when a tool is free, verified by physicians, and integrated into common practice, adoption is extremely fast. Czech hospitals and outpatient facilities should monitor this development and consider how to incorporate similar technologies into their workflows — whether through local solutions or by waiting for global players to enter our market.

OpenAI also released Health Blueprint — a document with recommendations for responsible integration of AI in US healthcare. Similar strategic approaches could inspire Czech healthcare institutions and regulators in preparing their own frameworks for using AI in patient care.