The problem it solves

Every new ChatGPT conversation starts from scratch. ChatGPT doesn't know your job title, your company's tone, the project you've been managing for three months, or the fact that you always want responses in plain English and under 200 words. You either explain all of that every single time, or you get generic output that doesn't match your actual needs.

The Memory feature stores specific facts about you and uses them in future conversations — automatically, without you having to paste context in. Once it's set up, every new chat already knows the basics.

Step 1: Turn memory on

Go to chatgpt.com → click your profile icon in the top-right → Settings → Personalisation → Memory. Make sure the toggle is switched on. If it's already on but blank, that just means nothing has been saved yet.

Step 2: Add your work context

The fastest way to build a useful memory bank is to type it directly into a chat. Start a new conversation and paste in something like the example below. ChatGPT will save it automatically.

Please remember the following about me for future conversations: Role: I am a [your job title] at [type of company, e.g. mid-sized logistics company]. Audience: Most of my writing is read by [e.g. senior management and external clients]. Tone: Professional but plain. No jargon, no filler phrases. Format preferences: - Short paragraphs (3-4 sentences maximum) - Bullet points for lists and action items - Keep responses under 200 words unless I ask for more Recurring tasks: I write [e.g. weekly status updates, supplier emails, and quarterly board reports]. Save all of this.

After you send this, ChatGPT will confirm what it has saved. You can then type "What do you remember about me?" at any point to see a summary of stored memories.

Step 3: Manage and update memories

To see, edit, or delete individual memories, go to Settings → Personalisation → Memory → Manage memories. You'll see each item as a separate entry. Delete anything outdated. Add new entries as your work changes.

Memory is not unlimited — it stores approximately 1,200 to 1,400 words in total. Keep each memory entry short and factual. If memory fills up, ChatGPT stops adding new entries until you remove some. Check it every month or two and clear anything no longer relevant.

What to save and what to leave out

Good things to save in memory: your job title and context, output format preferences, tone instructions, recurring task types, things you always want included or always want avoided.

Do not save confidential information — client names, financial figures, personal details, or anything your organisation's privacy policy covers. Memory is stored in your OpenAI account. Treat it the same way you'd treat a shared notebook: only write what you'd be comfortable with others eventually seeing.

💡 This week's action

Open ChatGPT right now, go to Settings and check whether Memory is switched on. Then start a new chat and paste in your work context using the template above — your role, audience, tone preference, and one recurring task. It takes five minutes. The next time you open ChatGPT, it will already know who you are.

Memory vs Custom Instructions vs Projects

ChatGPT also has Custom Instructions (under Settings → Personalisation), which let you write standing instructions for every chat. These work across all conversations but are fixed unless you manually edit them.

Memory is more flexible — it accumulates over time and can be updated by ChatGPT as you work. Use Custom Instructions for hard rules that never change, and Memory for ongoing context that evolves. For repeating specific tasks with their own files and rules, Projects (covered in a separate guide) give you the most control.