Key takeaways
- Tell patients at the start of every AI-handled call, in plain words, and give a way to reach a person.
- Update the privacy notice: what the tool does, what data it records, who processes it, how long it is kept and whether any decisions are automated.
- Use the same short wording everywhere: greeting, website, texts and reception scripts.
- Prepare staff with answers to the questions patients actually ask: is this a person, is it recorded, who sees it, can I opt out.
- Trust in the NHS is high but not unconditional; transparency protects it.
Why disclosure matters
Patients place a great deal of trust in their practice. In the 2026 GP Patient Survey, 92.7% said they had confidence and trust in the healthcare professional they saw or spoke to[1]. The NHS is also the most trusted organisation for handling data: 85% of UK adults trust it to act in the public's best interest[2]. That trust is an asset, and it can be lost quickly if patients feel something was hidden from them.
The same government tracker shows how fragile it can be. Trust in the NHS to use data to improve services fell from 78% to 73% between two waves, and healthcare is the area people most often say needs careful management of AI, chosen by 29%[2]. A patient who discovers halfway through a call that they have been speaking to software, without being told, is unlikely to give the practice the benefit of the doubt next time.
What the law expects you to say
When you collect personal data directly from a person, UK GDPR Article 13 requires you to tell them, among other things, who you are and how to contact you, how to contact your data protection officer, the purposes and legal basis for processing, and who will receive the data. It also requires you to tell people about the existence of automated decision-making, including profiling, where it applies[3].
Information about health is special category data under Article 9[4], so the privacy notice must also explain the condition you rely on. The eighth Caldicott Principle points the same way: inform patients and service users about how their confidential information is used[5]. A phone tool that records calls or keeps transcripts is processing personal data, and the notice must cover it.
- What the AI tool does and does not do (for example, answers practice questions and routes calls, does not give medical advice).
- What it records: audio, transcript, phone number, time, the type of request.
- Who processes the data on your behalf, and where it is stored.
- How long each kind of record is kept.
- Whether any decision about the patient is made automatically, and if not, say so.
- How to speak to a person instead, and how to raise a concern.
Where to tell patients
The call itself
A short statement at the start of every call the tool answers, before any question is asked.
The phone greeting
If the tool only takes overflow or after-hours calls, say so in the main greeting so callers are not surprised when it answers.
The website
A plain paragraph on the contact page and a link to the privacy notice section.
The privacy notice
A dedicated section covering the points above, reviewed by your data protection officer.
Texts and emails
If the tool sends messages, they should say they are automated and how to reply to a person.
Reception and the waiting room
A notice in the waiting room and a short script for staff who are asked about it.
Wording you can adapt
Keep it short, factual and calm. Patients want to know three things: that this is an automated system, what it can help with and how to reach a person.
Call opening: "Hello, you're through to Riverside Surgery. I'm the practice's automated assistant. I can help with general questions and point you to the right service. If you'd like to speak to the reception team, just say so at any time."
Website paragraph: "Some of our calls are answered by an automated assistant, for example when all our lines are busy or the surgery is closed. It can answer questions about the practice and explain how to get help. It does not give medical advice or make decisions about your care. You can ask to speak to a member of staff at any time. Read how we use and protect call information in our privacy notice."
Privacy notice section heading and opening: "Automated call handling. We use an automated assistant, provided by a company that processes data on our behalf, to answer some phone calls. Calls are recorded and transcribed so that we can check the service is working and follow up your request. No decisions about your care are made automatically."
Replace the fictional practice name with your own and make sure every statement is true for your setup. If calls are not recorded, do not say they are; if the supplier stores data outside the UK, the notice must explain it.
Offering a person, and opt-outs
The most effective reassurance is a working route to a person. Saying "you can speak to the team at any time" only helps if it is true. Check that the transfer works, that it connects within a reasonable time and that it is available in the hours the tool answers. After hours, say clearly when staff are next available.
- Let callers ask for a person in any words, not just a menu number.
- Where your telephony allows, let patients who have asked not to use the automated assistant go straight to the reception queue.
- Record the request on the patient's preferences if your process supports it, without making them explain why.
- Offer alternatives that do not involve the tool: the website, the NHS App or the front desk.
Questions patients ask, and how to answer
| Patient asks | Suggested answer |
|---|---|
| Am I talking to a real person? | No, this is the practice's automated assistant. I can put you through to the team if you prefer. |
| Is this call being recorded? | Yes, calls are recorded so the practice can check the service and follow up your request. Our privacy notice explains how long we keep them. |
| Who can see what I've said? | The practice team, and the company that runs the system on our behalf, under a contract with us. |
| Does it decide whether I get an appointment? | No. It explains how to get help; the practice team and clinicians make decisions about care. |
| I don't want to use it. | That's fine. You can ask for the team whenever you call, or use the website or the front desk. |
Make sure the answers in the tool and the answers staff give are the same. Inconsistent answers do more harm to trust than any single wording choice.
Callers who need extra care
A disclosure that works for most callers may not work for everyone. Think about the patients for whom a short spoken statement at the start of a call is easy to miss or hard to understand, and plan how they will be told and how they can avoid the tool if they prefer.
- People with hearing loss: make sure the disclosure is also on the website and in written materials, and that text relay and online contact are clearly signposted.
- People with learning disabilities or dementia: an easy read version of the website paragraph, and a flag on the record so their calls can go straight to staff where possible.
- Callers whose first language is not English: keep the opening plain, avoid jargon such as 'virtual agent', and make 'person' or 'reception' work as a request at any point.
- Carers and relatives: explain that the assistant does not discuss individual patients and will pass them to the team.
- Anxious or distressed callers: the urgent-help wording comes first, before any other explanation.
Ask your patient participation group to read the drafts before they go live. Five minutes of feedback from patients is usually enough to catch wording that sounds clear to staff but confusing to everyone else.
Example: updating a practice's materials
Example (fictional): a dental practice introduces an automated assistant for lunchtime and after-hours calls.
In this example the practice manager drafts the greeting and website paragraph, the data protection officer reviews the privacy notice section, and the principal dentist checks that no wording implies clinical advice. The team rehearses the five patient questions in a huddle. In the first fortnight two patients ask not to use the assistant; reception notes the preference and the patients' calls are answered by staff. A third patient complains that the greeting was too long, and the practice shortens it to two sentences.
Keeping it accurate
Clinicians share patients' caution. Among GPs who did not use AI, 82% were concerned about patient privacy and data security[6]. Disclosure is not a one-off task; review it whenever the tool's scope, supplier or data handling changes.
- Review the privacy notice section at least yearly and after any change of supplier or scope.
- Check the call greeting every quarter against what the tool actually does.
- Log patient questions and complaints about the assistant and use them to improve the wording.
- Make sure new staff learn the patient answers in their induction.
Frequently asked questions
Do we have to tell patients a call is answered by AI?
Yes. Transparency is a core UK GDPR requirement, and patients should know at the start of the call that they are speaking to an automated system and how to reach a person.
What should a privacy notice say about AI call handling?
What the tool does, what it records, who processes the data, where it is stored, how long it is kept, whether any decisions are automated and how to speak to a person instead.
Can patients opt out of the AI assistant?
They should always be able to ask for a person. Where your telephony allows, you can also route patients who have asked not to use it straight to reception.
How long should the AI greeting be?
Two or three short sentences: who is answering, what it can help with and how to reach the team.
References
Ipsos (GP Patient Survey provider for NHS England) (2026). 2026 GP Patient Survey results released (opens in a new tab) ↑
GOV.UK (Department for Science, Innovation and Technology) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2016). UK GDPR Article 13: Information to be provided where personal data are collected from the data subject (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR Article 9: Processing of special categories of personal data (opens in a new tab) ↑
GOV.UK (National Data Guardian) (2026). The Caldicott Principles (opens in a new tab) ↑
Nuffield Trust (2025). How are GPs using AI? Insights from the front line (opens in a new tab) ↑