Hey there,
Welcome back. And a warm hello to anyone joining us for the first time.
🍁 To review previous issues, visit theaieh.ca/articles |
Have you ever opened ChatGPT or Claude and had it casually mention something you told it weeks ago, and thought, wait, how does it still know that? You are not imagining it. Over the past while, the major AI tools have quietly started remembering things about you across conversations, not just within one chat.
This can genuinely be useful. It can also feel a little strange the first time you notice it. So today, plain and simple, what these tools actually remember, why it happens, and exactly how to see it and clear it if you would rather they did not.
Two different kinds of "remembering"
It helps to know there are two separate things going on, because they get mixed up a lot.
The first is within a single conversation. Any AI tool remembers everything you have said earlier in that same chat, the same way a person remembers what you said five minutes ago in a conversation. Close the chat or start a new one, and that context is gone, unless the second kind of memory kicks in.
The second is persistent memory, a newer feature that carries certain details forward into brand new conversations, even ones you start weeks later. This is the one that surprises people. It does not save your whole conversation word for word. It quietly pulls out things like your name, your job, your ongoing projects, or how you like answers formatted, and keeps a running summary it can draw on later.
Where to find it, tool by tool
Every major AI tool handles this a little differently, and these menus do shift around from time to time, but here is where to look as of right now.
| ChatGPT: Settings, then Personalization, then Manage Memories. You will see a list of what it has saved. Delete anything individually, or clear the whole list at once. |
| Claude: Settings, then Memory. You can review individual saved details, edit or remove specific ones, or turn memory off entirely. |
| Gemini: The Settings gear icon, then Personal context. From there you can view saved details or turn past chat memory off completely. |
A genuinely useful habit either way, in any of these tools you can simply type "what do you remember about me" and it will tell you, in plain language, no menu digging required.
Four prompts to try this week
|
Prompt 1: See what it actually remembers Before doing anything else, just ask. This is the fastest way to see what has been quietly saved, and most people are a little surprised by what shows up. Ellen in Brandon had been using ChatGPT casually for months, mostly for recipe ideas and the odd email. Out of curiosity she asked it what it remembered about her, and it mentioned her job, that she was gluten intolerant, and that she had two dogs. None of it felt invasive, but she had genuinely forgotten she had ever mentioned most of it. Just ask it directly:
|
|
Prompt 2: Clear something specific without wiping everything You do not have to nuke your entire memory to remove one thing that no longer applies, or that you just did not mean to have saved in the first place. You can ask it to forget something directly, right in the chat, no settings menu required. Robert in Sudbury had mentioned in passing that he was job hunting, and months later, after he had accepted a new position, Claude was still referencing the search in unrelated conversations. He simply told it to forget that detail, and it confirmed the update immediately. Fill in the blank:
Ready to use example:
|
|
Prompt 3: Ask it to stay out of a sensitive topic Before you get into something personal, a health question, a family situation, a legal matter, you can ask upfront that this particular conversation not be remembered or carried forward. Most tools also offer a temporary or incognito style chat mode specifically for this, worth looking for in the settings if you want a permanent version of this habit. Fill in the blank:
Ready to use example:
|
|
Prompt 4: Do a full memory review, once a season Set a reminder to actually open the memory settings every few months, the same way you might review your phone's app permissions. Ask AI to summarize everything it has stored in plain categories, so you can decide in one pass what still applies and what does not. Fill in the blank:
Ready to use example:
|
A quick word on where the line is
None of this means AI memory is something to be afraid of. Used well, it saves you from repeating yourself constantly. But it is worth treating the same way you would treat any account that holds personal details, worth an occasional look, and worth being deliberate about what you let it hang onto, especially anything financial, medical, or about someone else who never agreed to be remembered.
Your challenge this week
Open whichever AI tool you use most and try Prompt 1 right now. Just ask it what it remembers about you. Takes ten seconds, and you will know exactly where you stand.
Quick catch-up: what changed in ChatGPT and Claude
Both tools got big upgrades over the past month. You do not need to do anything, but here is what you might notice the next time you open them.
| ChatGPT: OpenAI moved to a new family of models in September, with the most powerful one going to paid plans first and faster, cheaper versions following a few weeks later. The more practical news is that ChatGPT now works right inside Microsoft Word, and its image generator got another upgrade. One heads up, if you use a custom GPT, OpenAI is phasing those out in favour of something it calls plugins, so it is worth keeping an eye out for that change. |
| Claude: Anthropic released two new models in late September, Opus 5.5 and Sonnet 5.5, both faster and cheaper to run than what came before. The bigger change for everyday use is that Claude is now one app. Chatting, working on documents and building slides all happen in the same place instead of separate modes. |
One story from the month that I think is worth knowing. OpenAI held back one of its newest models after its own testing caught it misleading people about what it had actually done. I find that oddly reassuring, because someone checked. It is also a good reminder that newer does not mean always right, which is exactly why the habit below never goes out of date.
Coming up next issue
Next issue we are turning to something a lot more fun, using AI to finally get your family photos organized, found, and actually looked at again instead of buried in your camera roll.
AI assisted, Human led. Whatever AI produces for you, always read it, verify it, and make it sound like you. AI is an incredible first draft machine. Your judgment, your voice, and your values are always the final layer. |
As always, I read every reply personally. Did anything surprise you when you checked what AI remembers about you? Hit reply and let me know, I am genuinely curious what turns up for people.
If you found this useful, pass it along to someone who has never thought to check. Ten seconds well spent.
Know someone who would love AI, Eh? Send them here: Subscribe for free |
Talk soon,
Chris
Founder, AI, Eh?