AI Guides › Getting Started
Setting Up Your "About Me" File Properly
By Nigel Guy · 2 min read
Given an open text box and told "tell it about yourself," most people
either freeze and write two sentences, or overcorrect into a full
autobiography — career history, hobbies, half-relevant personal details —
on the theory that more information must produce better answers. Neither
version is actually well suited to the job this file does.
The rule: an "about me" file should hold what you'd otherwise repeat
across many conversations, and nothing else — that's the only test for
whether something belongs in it.
What belongs
- Your role and general domain. Enough for it to calibrate the level
of an answer without you restating your job every time.
- Genuine standing preferences. Directness over hedging, short answers
before detail, a house style you always want followed — the kind of
thing you've caught yourself repeating in three separate chats already.
- One or two things you're consistently working on, if they're
ongoing rather than one-off — anchoring this to something real tends to
produce more useful results than a list of abstract traits.
- Anything that changes how it should treat a whole category of your
requests — a regulatory context you always work within, a language
you always want responses in, an audience you're consistently writing
for.
What doesn't belong
- One-off project details. Those belong in a project's own context,
not a global file that follows you into unrelated conversations.
- Anything sensitive you wouldn't want surfacing in an unrelated chat
— health, financial specifics, anything genuinely private. Sensitive
categories are often off by default and need deliberate opt-in for good
reason; don't route around that by pasting them into a general field.
- A full biography. Nobody needs to know your degree, your hometown,
and your hobbies for it to write a better email — pad this file and the
genuinely useful three sentences get diluted among ten irrelevant ones.
- Instructions that only apply to today's task. State those in the
conversation itself; the file is for what's durable, not what's current.
The mechanism
- Write it in one sitting, as plain sentences, not bullet fragments.
- Keep it short enough to read back in under thirty seconds.
- Revisit it monthly and delete anything that stopped being true — a
stale preference is worse than no preference, because it actively
steers answers in the wrong direction rather than just being neutral.
- Correct it the moment you notice it's making a wrong assumption, rather
than working around the wrong assumption in every chat instead.
Guardrails
- This file affects every conversation it applies to — a mistake or a
stale entry in it has a wider blast radius than a mistake in one chat, so
it's worth the occasional review.
- Treat it like a notes app a colleague could plausibly see, not a diary —
don't put anything there you'd mind being read back to you out of
context.
- Where this setting lives, what it's called, and its exact scope can
change between versions of the product — check the current settings
menu rather than assuming this guide's description matches it exactly.
All 751 AI guides · JulieMango plans from £17/mo