Before and After the Doctor's Appointment, Not Instead of It

My rule for AI and health is short. AI can get you ready for the appointment and help you read the letter afterwards, but AI doesn't get to tell you what's wrong with you.

You know how the ten minutes go. There are three weeks of waiting, then a doctor who was kind and busy, and you nodding along.

In the car park you remembered the question you'd meant to ask. What did "we'll refer you" mean, and how long would that take? Two weeks later a letter arrived in a language that was nearly English.

AI helps before the visit and after the visit, while the ten minutes in the middle stay between you and your doctor. So why is every prompt below (the message you type to AI) built to stop AI guessing what's wrong?

Because AI can't examine you and doesn't know your history, yet AI will still give you a list that sounds sure of itself. That's what AI does with any question.

Some of the list will frighten you, though most of it won't apply. And the frightening bit is what you'll carry into the room, instead of the question you came with.

Expect AI to drift towards naming an illness, because people usually want that. When you see "this could be" or "it sounds like", type this back. "I didn't ask what it is. I asked what to ask." You may need to say it more than once.

The night before

Open a temporary chat so none of this is kept, and then paste this in.

I've got a GP appointment tomorrow about [the reason, in your words]. Don't tell me what it might be. Help me get the most out of ten minutes. First, ask me a few questions so I can describe it clearly to the doctor. For example, when it started, what makes it better or worse, and what I've tried. Then write me a list of five questions worth asking the doctor. Also give me one sentence I can say if I feel rushed.

AI will ask about things you hadn't thought to notice, such as whether the problem is worse in the morning. AI may also ask whether anything changed around the time the problem started. Answer all of these questions.

Then read the five questions AI wrote. Cross out the ones that aren't yours, and add the one you were afraid to ask, because that's always the one that matters.

Write the final list on paper or in your phone notes, and hand the list to the doctor if you like. Most doctors are glad of a list. A patient who knows what they want to ask makes the appointment quicker and better for both sides of the desk.

Is the appointment for someone else? Maybe it's a parent who comes out saying "they said it was fine", or a child who can't say where it hurts.

Then add one thing, which is a short note of what you've noticed, with dates. "More tired since August, fell twice in September, stopped going to the shop."

Doctors act on patterns, and often the person in the room can't give them one. You can.

In the room

The AI stays at home.

Take your list and the sentence AI gave you for when you feel rushed. Usually that comes out as something like "Before I go, I had one more thing I wanted to ask."

Having that sentence ready is the difference between saying it in the room and remembering it in the car park.

When the letter arrives

Clinic letters, test results and discharge summaries are written by one doctor for another doctor, and you just get a copy. The hard part is usually the words.

Black out your name, NHS number and date of birth. Then take a photo of the letter or upload it in a temporary chat, and use this.

This is a letter from my doctor. Put it into plain English, line by line, so I understand what each sentence means. Don't add anything that isn't in the letter. Don't tell me what it means for me, what it could be, or what happens next. The letter says what it says. If a word is a medical term, explain the word, not my situation. Then tell me which lines are worth asking my doctor about if I'm unsure.

You get the letter back in words you'd use, along with a short list of lines that might need a follow-up call. But AI won't tell you how things will turn out, because you told AI not to, and AI would only be guessing anyway.

AI can still get the meaning wrong. Some medical terms have two meanings and AI may pick the wrong one, so the "ask your doctor about" list is there to catch that.

If anything in the plain-English version worries you, take that worry to the surgery or to 111. The chat's job ended when you understood the words.

For a parent's letter, ask them first, because the appointment and the letter belong to them, even with the name blacked out.

Read the letter with them, not for them. Going through it together over a cup of tea is a conversation, while handing them a printout feels like a verdict.

Once you've finished

AI memory and health don't mix. (Memory is when AI keeps notes about you from one chat to the next.) That's why every step above happens in a temporary chat.

If you have memory switched on anyway, check what AI has kept and delete anything medical.

Any hour in between

Don't ask AI what's wrong with you, not at 2am and not with the letter already in your hand. The list will be partly right and partly terrifying, and none of it will come from anyone qualified. The terrifying part is what you'll remember.

Never change a dose or stop a medicine because of anything AI said, and don't skip an appointment because of AI either. If AI says something that makes you want to, ring the surgery sooner.

If it's urgent, call 999 or 111.

AI can't tell the difference between a bad night and an emergency, and AI will try anyway.