Clinicians will increasingly encounter patients who record consultations using apps like Kin Health or Aide Mirror. Some patients will mention it openly. Others will record without saying anything — which is their legal right for personal use. The recording will be processed by AI, summarised into plain English, extracted into action items, and potentially shared with family members or caregivers.
The practical question is not whether this can be stopped. It cannot — and should not be. Patients have a longstanding right to record consultations for personal use. The practical question is how to maintain safe, clear, professional communication in consultations that may be captured and summarised by AI tools the clinician does not control.
What Patients Are Trying to Solve
Patients forget roughly half of what their doctor tells them. The forgetting rate is worse for complex consultations, stressful diagnoses, consultations involving multiple topics, and patients managing several conditions simultaneously. Elderly patients, patients with cognitive impairment, anxious patients, and patients receiving bad news are particularly affected.
Recording and AI summarisation solve a genuine problem: the patient wants to remember what was said, understand what to do next, and share the information with family members who were not present. This is not adversarial. Most patients recording their consultations are trying to be better patients — more informed, more adherent, more engaged with their care.
What Clinicians Should Say at the Start
If the patient mentions recording, acknowledge it naturally:
"That's fine — recording can really help you remember what we discuss. If anything in the summary seems unclear afterwards, do check with me or with the practice, because AI tools can sometimes simplify things that are clinically important."
If the clinician wants to proactively acknowledge the possibility:
"You're welcome to take notes or record if that helps you. I'll also summarise the key plan before we finish."
Both scripts are collaborative, not defensive. They position the clinician as an ally in the patient's understanding.
How to Reduce Misinterpretation
Structure the consultation explicitly. AI summarises what it hears. A structured consultation — presenting complaint, assessment, plan, safety-netting, follow-up — produces a more coherent AI summary than a conversational, meandering exchange. The structure helps both the AI and the patient.
State the plan verbally before ending. "So to summarise: we'll do X, you should continue Y, and if Z happens, contact us urgently." This gives the AI a clean final summary to extract — and ensures the patient hears the plan explicitly, whether or not they are recording.
Name the diagnosis and the uncertainty. "I think this is most likely A, but I want to rule out B with this test" is much safer than vague language that the AI may interpret as diagnostic certainty. If the clinician is uncertain, saying so explicitly protects against AI summaries that present working diagnoses as confirmed.
Be specific with safety-netting. "Return within 48 hours if you develop neck stiffness, visual changes, or sudden severe headache" is captured accurately by AI. "Come back if it gets worse" is vague and may be summarised as generic advice rather than specific clinical guidance that could save the patient's life.
Distinguish between recommendations and considerations. "I'm prescribing X" is a recommendation. "We could consider X in the future if this doesn't improve" is a consideration. AI may blur this distinction — converting a speculative mention into a recommendation in the patient's summary. Be explicit about what you are doing versus what you are discussing.
What to Do If the AI Summary Is Wrong
Patients may return saying "the app said you recommended X" when the clinician said something different. This is a predictable failure mode of multi-stage AI summarisation — transcription errors, inference drift, over-simplification, or hallucination of recommendations.
The appropriate response: "Let me clarify what I actually recommended. The AI may have simplified or missed something. Here is what I said and what the plan is." Review the clinical record to confirm alignment. If the AI summary creates confusion about a safety-critical recommendation, document the correction explicitly in the medical record.
Suggested Wording for Key Clinical Moments
When prescribing: "I'm prescribing [drug name] at [dose] for [indication]. Take it [frequency] for [duration]. The most important thing to watch for is [key side effect or red flag]. If you experience [specific symptom], stop the medication and contact us."
When safety-netting: "I want you to come back within [timeframe] if you notice [specific symptoms]. If [emergency red flag] happens, go to A&E or call 999 immediately. Do not wait."
When uncertain: "I'm not sure of the exact diagnosis yet. We're going to investigate with [specific tests]. I'll contact you with results by [date]. In the meantime, watch for [specific red flags] and come back immediately if any of them develop."
Each is specific enough to survive AI summarisation accurately — and clear enough to serve the patient's understanding whether or not they are recording.
iatroX helps clinicians check the guideline evidence behind their recommendations — so the verbal advice captured by any recording is as accurate, evidence-based, and guideline-concordant as possible.
Use iatroX to verify clinical recommendations before the consultation →