A quick note first: the way AI tools handle memory and settings changes often, sometimes weekly. Look at the published date on this article and allow for things having moved since. For training that reflects the current versions, contact the Softwired team.
The short answer
AI tools personalise answers in 2 ways. Memory is what the tool picks up and carries between chats, either facts you ask it to remember or details it draws from past conversations. Custom instructions are standing rules you write once and the tool applies every time. You can view and switch off memory in Settings and edit or clear your instructions there too. A temporary or incognito chat gives you a one-off conversation that doesn't use or add to memory.
Chats, threads and why "new chat" matters
Each chat is a separate conversation. Within a chat, the AI can see everything said so far. Between chats it only knows what memory, instructions or a project (a folder that holds related chats, files and instructions) give it. Starting a fresh chat for a new topic clears out the old material, so the AI answers from what's relevant.
Memory: what it is in each tool
This guide covers ChatGPT from OpenAI, Claude from Anthropic, Gemini from Google, Microsoft 365 Copilot from Microsoft, Grok from xAI and Meta AI from Meta. "Not confirmed" means we could not find the detail on an official page.
| Tool | What it's called | Where to manage it |
|---|---|---|
| ChatGPT | Memory, with 2 controls: Reference saved memories and Reference chat history | Settings > Personalization |
| Claude | Memory, shown as topics you can edit, with a toggle for sensitive topics | Settings > Memory |
| Gemini | Memory, part of Personal Intelligence | Settings & help > Personal Intelligence |
| Microsoft 365 Copilot | Saved memories and chat history personalisation | Copilot Chat Settings > Personalization |
| Grok | Memory across chats | Grok settings (exact path not confirmed) |
| Meta AI | Memory ("What do you remember about me?") | Ask Meta AI, or delete the message the memory came from |
Each tool handles the detail differently:
- ChatGPT keeps "saved memories" (things you've told it to remember, or details it decides are useful) separately from your chat history. Deleting a chat leaves any saved memory that came from it in place, so delete both the memory and the chat to remove something completely. If you turn off Reference chat history, ChatGPT schedules details remembered from past chats for deletion within 30 days.
- Claude organises memory into topics you can review and edit. Memory is on by default on Free, Pro and Max plans. On Team and Enterprise plans it's off by default and your organisation controls it. Memory now carries across both chat and Cowork, Claude's mode for multi-step tasks.
- Gemini can use your past chats to personalise answers through Personal Intelligence. Google notes that deleting the relevant chats is the reliable way to stop Gemini using a topic.
- Microsoft 365 Copilot lets you view, edit and delete saved memories. Turning saved memories off stops future use and keeps the existing ones, so delete those separately.
How you'd actually use it
Tell the tool, "Remember that I manage a team of 8 in a Brisbane accounting firm and our clients are mostly small businesses." Next week, when you ask for a client email template, it starts from that context without being told again. Every few months, ask "What do you remember about me?" and clean out anything stale.
Custom instructions
Custom instructions are your standing brief. You write them, and they stay the same until you edit them, whereas memory changes as the tool learns.
| Tool | What it's called |
|---|---|
| ChatGPT | Custom instructions (up to 5,000 characters on paid plans) |
| Claude | Instructions for Claude, in Settings > Your Claude |
| Gemini | Instructions for Gemini |
| Microsoft 365 Copilot | Custom instructions |
| Grok | Custom instructions |
| Meta AI | Not confirmed |
Projects in ChatGPT and Claude have their own instructions too. In ChatGPT, project instructions override your global custom instructions inside that project.
How you'd actually use it
Paste in a starter set like this one, then adjust it to your role:
- "Use Australian English and Australian date formats."
- "I work in [role] at a [type] business in [city]."
- "Start with a two-sentence summary, then the detail."
- "If you're not sure about a fact, say so rather than guessing."
Temporary and incognito chats
| Tool | What it's called | What it does |
|---|---|---|
| ChatGPT | Temporary Chat | Doesn't create or update memories. You can choose Personalized or Unpersonalized before you start. |
| Claude | Incognito chat | A private chat kept out of your normal history and memory. Claude's help page sets out how long it's kept. Where Claude's chat and Cowork have merged into one experience, incognito chats open in the older chat view, so Claude can't create files or run code in them. |
| Gemini | Temporary Chat | Skips personalisation, stays out of Gemini Apps Activity and is excluded from model training. Google keeps it for up to 72 hours. |
| Meta AI | Incognito chat | Meta says it isn't saved or stored after you exit. |
| Microsoft 365 Copilot | Not confirmed | Not confirmed |
| Grok | Not confirmed | Not confirmed |
How you'd actually use it
You want a second opinion on a sensitive staffing situation and don't want it shaping future answers. Open a temporary or incognito chat, remove names, ask, and close it.
Watch-outs
- Use memory for preferences and background. ChatGPT's own help says not to rely on memory for large blocks of verbatim text, so keep templates in a project or a skill (a saved, reusable set of instructions).
- Work and personal accounts are separate. Memory in your personal account stays there, which keeps your work data and your personal data apart.
- On business plans, your organisation may switch memory off or limit it, so ask your IT admin if you can't find the setting.
Frequently asked questions
How do I see what an AI tool remembers about me?
Ask it directly, "What do you remember about me?", then check the memory page in Settings for the full list.
Does deleting a chat delete the memory?
It depends on the tool. In ChatGPT, saved memories are stored separately from chats, so delete both. In Meta AI, deleting the message removes the memories that came from it.
Should I turn memory off?
Keep it on for everyday work, because it saves you re-explaining your role and preferences. Review it every few months and use a temporary chat for anything sensitive.
Related guides
- What is a project in ChatGPT, Claude and other AI tools?
- Custom GPTs, Gems, skills and plugins explained
- AI tool features explained: a plain-English guide for 2026
Put it to work with AI On Ramp
Memory and custom instructions pay off when everyone in the team sets them up the same way.
AI On Ramp gets your team using AI with confidence, on their own work and across teams, while you keep its cost and risk under control. We set up the tools, then our consultants sit beside your most advanced AI users and build real in-house workflows with them. Your nominated AI Lead gets the governance tools and our ongoing support to keep the business current as the tools change. We bring deep technical expertise and decades of business change management, and we work face to face from Brisbane and along Australia's east coast. See how AI On Ramp works.
Want to talk through how this applies to your team? Book a call with Softwired.
