Key points
- Clinicians raise about 0.57 questions per patient seen and pursue about half of them, in a review of 11 studies and 7,012 questions.1
- A cited answer can be checked against its source in seconds and weighed on a recognised scale such as GRADE.3
- The check takes under a minute. Confirm the source supports the claim, then confirm it was written for a patient like yours.
- wai answers clinical questions from current guidance, recent papers and peer-reviewed research, cites every answer, and the clinician makes the call.6
Summary
Ask the question during the consultation, open the cited source, and act once it supports the answer and fits your patient. The question gets answered while the patient is still in the chair.
Every clinical question answered in the room, with the source one tap away and the patient leaving with a plan. Cited evidence answers make that routine, as long as each citation is opened before it is used.
See it in your clinicThirty minutes on your own patients and doctors.
Request a demoAsk wai a question from your own clinic list and see the answer arrive with its sources, in fifteen minutes with a clinician who uses it daily.
| What the citation points to | Starting confidence | Check before acting |
|---|---|---|
| Current national guideline graded with GRADE | As graded by the guideline panel | Version date, country, and whether the recommendation is strong or weak |
| Systematic review or meta-analysis of randomised trials | High, unless downgraded | Consistency across trials and whether the trials match your patient |
| Single randomised trial | High, unless downgraded | Size, follow-up, population and whether the outcome matters to your patient |
| Cohort or case-control study | Low, unless upgraded | Confounding, and whether a large effect justifies more confidence |
| Case report or expert opinion | Very low | Use to raise a question and look for stronger evidence |
| No source, or a source that differs from the answer | None | Look the question up directly before acting |
The question in the room
A systematic review in JAMA Internal Medicine pooled 11 studies that interviewed clinicians after each visit and found 0.57 questions raised per patient seen.1 At 20 patients a day, that is roughly 11 questions by the evening.
Clinicians pursued 51% of those questions and found answers to 78% of the ones they pursued.1 The main barriers were lack of time and doubt that a useful answer existed, and 34% of questions concerned drug treatment.1 A sourced answer in seconds lets the question be pursued in the room.
Where answers now come from
In the AMA's survey of 1,692 US physicians in January and February 2026, summaries of medical research and standards of care became the most common use of AI, at 39% of the 1,342 physicians asked, up from 13% in 2024.2
What the citation is for
WHO's 2024 guidance on large multi-modal models in health records documented risks of false, inaccurate, biased or incomplete statements.4 A citation turns that risk into a check you run in seconds, the habit medicine already applies to any reference.
A usable answer names the guideline or paper, links to it and shows which passage supports each claim. The clinician who acts on it owns the decision, as set out in who signs an AI clinical note.
Checking a cited answer in under a minute
Run four checks before acting.
First, open the source and confirm the passage supports the specific claim, down to the dose or interval quoted.
Second, check the date and the country. A guideline updated last year outranks a review from a decade ago, and a recommendation written for one formulary may differ from yours.
Third, check the population. A trial in adults with normal renal function leaves open a question about an 84-year-old with an eGFR of 28, and GRADE lists indirectness as a reason to lower confidence in evidence.3
Fourth, record the source in the note, so the next clinician and any later review can see what the decision rested on.
Weighing what sits behind the citation
GRADE, adopted by guideline bodies worldwide, rates evidence as high, moderate, low or very low.3 Randomised trials start as high and can be downgraded for study limitations, inconsistency, indirectness, imprecision or reporting bias. Observational studies start as low and can be upgraded, for example when the effect is very large.3
Expert reports of clinical experience and case reports should be labelled very low quality evidence.3 An answer resting on a current GRADE guideline carries the panel's grading with it, and an answer resting on one small study carries the weight of one small study.
How wai answers a clinical question
wai answers clinical questions in seconds from current guidance, the most recent medical papers and peer-reviewed research, and cites every answer.6 Behind each answer, wai indexes more than 2,000 clinical guidelines and searches more than 40 million evidence sources.5
The answer sits in the same workspace as the note, so the source can go straight into the record you sign, and the clinician makes the decision. The same review habit applies to the note itself, covered in our guide to AI scribe accuracy.
Where to start
For one week, write down every clinical question raised in clinic and mark the ones answered before the patient left. Compare your share with the 51% pursued in the research.1
The next week, put each question to a cited evidence tool during the consultation, run the four checks and note the source in the record. Count the questions still waiting at the end of each day as the list empties.
Bring three questions from last week's clinic. We answer them live on a video call, with every source open.
Frequently asked questions
Can a clinician rely on a cited AI answer without opening the source?
Open the source every time. The citation is there so the clinician can confirm it supports the specific claim, which takes seconds and is the core of safe use.
What if the citation differs from the answer?
Look the question up directly before acting. Catching a mismatch between answer and source is the main job of the check.
How should a clinician weigh a guideline against a recent trial?
Check whether the guideline predates the trial and whether the trial population matches the patient. GRADE starts a well-run trial as high quality evidence, and a guideline panel has weighed the whole body of evidence.3
Should the evidence used be recorded in the note?
Yes, briefly. Naming the guideline or paper behind a decision makes the reasoning visible to colleagues and to any later review of the record.
Who is responsible for a decision informed by an AI answer?
The clinician who makes the decision is accountable for it, as with any reference source.
Sources
- 1JAMA Internal Medicine, Del Fiol, Workman and Gorman, 2014Clinical questions raised by clinicians at the point of care, a systematic reviewRead 14 October 2026Read source
- 2American Medical Association, 20262026 Physician Survey on Augmented IntelligenceRead 14 October 2026Read source
- 3BMJ, Guyatt and colleagues for the GRADE Working Group, 2008GRADE, an emerging consensus on rating quality of evidence and strength of recommendationsRead 14 October 2026Read source
- 4World Health Organization, 2024WHO releases AI ethics and governance guidance for large multi-modal modelsRead 14 October 2026Read source
- 5wai, 2026See wai on your own clinic's workflowsRead 14 October 2026Read source
- 6wai, 2026AI medical scribe for independent cliniciansRead 14 October 2026Read source
Every clinical question answered before the patient leaves.
Fifteen minutes on video with a clinician who uses wai daily, on questions from your own clinic. We reply within one working day.
Partner with us
See it on your own patients, then build it with us
Thirty minutes on your own patients, doctors and rules, in your market. We reply within one working day, and the clinics that partner with us early shape what we build next.
Read how specialist opinions works in a clinic