Guide 4

How to write a good prompt.

6 min read

The quality of what comes back from an AI tool is almost entirely determined by what you put in. A vague question gets a generic answer. A specific question, asked with full context, gets something you can actually use. Me. handles the context. Your job is the question.

You don't have to start from scratch

The prompt library has you covered. It holds a curated set of prompts ready to use as-is or adapt to your situation, plus community-contributed prompts rated by other Me. users. After each compilation you can rate the prompt you used. The highest-rated prompts surface to the top. Over time, the library reflects what actually works, not what sounds good in theory.

Save prompts that work for you. Refine them. A prompt you have used three times and tweaked is worth more than a new one every session.

The difference a specific question makes

Here is the same question asked three ways:

Weak: "How am I doing?"

The AI has no idea what you mean. You will get something generic about sleep, stress, and recovery that could apply to anyone.

Better: "What does my sleep data show over the last 14 days?"

More specific, but still descriptive. You will get a summary, not an insight.

Strong: "My training load has increased over the last three weeks. What does my HRV trend in the 48 hours after my long runs tell me about how well I am recovering, and is there anything in my sleep data that supports or contradicts that?"

Now the AI has a specific window, a specific metric, a specific question, and a reason to look for a relationship between two data points. That is a question worth asking.

The pattern is the same whether you are preparing for a GP appointment ("I have a cardiology review next week. Summarise my resting heart rate and any notable anomalies over the last 90 days in plain language a doctor can work with.") or tracking something specific ("I started a new medication six weeks ago. Is there anything in my sleep, HRV, or energy data that changed around that time?").

Specific time window. Specific metric or metrics. Specific question. That is the structure.

Time range and data toggles

Before you compile, you choose a time range and select which data to include. These decisions shape the answer before you ask anything.

A 7-day window, the free range, is right for recent training load, a current health change, or preparing for an appointment. Longer windows are a Pro feature: 14, 28, and 90 days, and all time. 90 days is where the slower patterns become visible, how your sleep shifts across a training block, how a medication change lands over time, what your recovery actually looks like across a full cycle. More data is not always better. A focused compilation with the right toggles on will get you a sharper answer than everything at once.

Your profile does the heavy lifting

Every compilation carries your full profile with it. You are not explaining yourself from scratch every time. The context is already there. That is what closes the gap between a useful AI conversation and a generic one.

The compilation is the starting point