Aipedia: Doctors and medical practices

Can doctors use AI chatbots for clinical questions?

Aipedia, by the Apex broking team · Last reviewed 2026-10-27

Doctors can use general-purpose AI chatbots as a starting point, but must not enter patient-identifiable data and must check every answer against a primary source. Advice that relied on an unchecked chatbot answer is the doctor's advice in law and before the GMC.

Part of: AI in doctors and medical practices

In short

General-purpose generative AI chatbots answer drug, dose and guideline questions fluently, but they can invent references, doses and recommendations. The GMC applies Good medical practice to AI: work within your competence and keep up to date (paras 3 and 13), protect confidentiality, and use professional judgement. Defence organisation advice is not to enter any patient personal data into generative AI, and warns that removing a name may not anonymise it. Indemnity usually treats chatbot-assisted advice like any other clinical decision, so the doctor's checking is what a claim will turn on.

How doctors use general-purpose chatbots

Doctors use public generative AI chatbots to check drug interactions, summarise long guidelines, generate differential lists for unusual presentations, draft patient information and translate explanations into plain language. These tools are not medical devices and were not designed for clinical decision-making. They produce confident, well-written answers that may be out of date, based on non-UK practice, or simply invented, a problem known as hallucination.

The GMC's research with doctors using AI found that generative AI users avoided entering confidential patient data, and that over-reliance and deskilling were among the risks doctors themselves named.

The duties that apply

Competence (GMP paras 3 and 13): the GMC's AI resource says you must work within your competence and keep up to date. You need enough knowledge to spot a wrong answer.
Professional judgement: the decision remains yours, whatever tool informed it.
Confidentiality: Domain 4 requires confidentiality in "written, verbal and digital communications". The GMC's confidentiality guidance requires you to protect information against improper access or disclosure (para 119) and take reasonable steps to keep communication methods secure (para 133).
Accuracy: information you communicate must be "accurate, not false or misleading" and checked first.

Medical defence organisation guidance goes further: do not put any patient personal data into generative AI tools, because removing a name may not be enough to anonymise a case. A rare condition, age, location and dates can identify a patient.

What can go wrong

Illustrative scenario: a doctor asks a chatbot for the dose adjustment of a medicine in reduced kidney function. The answer looks authoritative and cites a guideline, but the figure is wrong and the reference does not exist. The doctor prescribes accordingly and the patient suffers toxicity. In the complaint, the doctor cannot point to any reliable source for the dose.

Other risks include differential lists that omit a serious diagnosis, guideline summaries that mix UK and overseas practice, and patient leaflets with errors. Pasting a referral letter or discharge summary into a public chatbot to "summarise" it may be a confidentiality and data protection breach, even if the clinical answer is right.

How indemnity usually responds

A clinical negligence claim is assessed on whether your care met the standard of a reasonably competent doctor, not on the tool you used. For NHS work, CNST or CNSGP usually responds; for private work, your defence organisation or medical malpractice policy, subject to its terms. One defence organisation's published position offers assistance with AI only where a human remains the decision maker and the member complies with law and national guidance, and it excludes data breaches. A breach caused by entering patient data into a public tool is also outside NHS schemes, which exclude data protection breaches. See AI hallucination liability and data breach cover within PI.

Practical controls

Never enter names, NHS numbers, dates of birth, images or identifying case details.
Use the chatbot to find where to look, then verify in the BNF, NICE or the primary source.
Check every reference exists and says what the chatbot claims.
Prefer tools your employer or practice has approved, with data protection checks.
Do not let a chatbot answer substitute for asking a senior colleague (GMP para 67).
Record the source you relied on, not the chatbot.

How Apex helps

Apex arranges indemnity for doctors and clinics, including private work outside NHS schemes. A broker reviews your cover and how your policy treats AI-assisted advice and data incidents; see does PI cover AI mistakes.

Sources

Frequently asked

Can I use a public AI chatbot for clinical questions?

Yes, for general questions, as long as you enter no patient-identifiable data and verify the answer against a reliable source before acting. The GMC expects you to work within your competence and use professional judgement, so the chatbot cannot be the basis of a decision on its own.

Is removing the patient's name enough to anonymise a query?

Often not. Defence organisation guidance warns that removing a name may not be enough, because combinations of age, condition, location and dates can identify someone. The safest course is not to enter any patient personal data into general-purpose generative AI tools.

Will my indemnity cover a mistake based on a chatbot answer?

Usually the claim is treated like any other clinical error, subject to your scheme or policy terms. One defence organisation's help for AI depends on a human remaining the decision maker and compliance with guidance. A data breach from pasting records into a chatbot may be excluded.

Check your indemnity covers how you use AI

Tell us about your practice and the work you do. A broker will review how your current cover treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start an indemnity quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.