The context window is what an AI can see right now, inside one conversation. AI memory is a separate app feature that saves notes about you and brings them into later conversations. The window empties when the chat ends; memory stays until you pause it or delete it.
- Context window: the AI’s working space for this chat. Your messages, pasted files and its replies all sit in it, and it has a size limit.
- AI memory: a setting in chat apps such as ChatGPT and Claude that keeps selected notes about you between chats.
- Different fixes: a chat that forgets its own start is a context window problem; a new chat that knows your job is memory at work.
Think of a desk and a notebook. The context window is the desk: everything on it is in view, but it is only so big, and it is cleared when you leave. Memory is a small notebook in the drawer, where the app writes down a few things about you to put back on the desk next time.
| Question | Context window | AI memory |
|---|---|---|
| What is it? | Is: all the text the model can refer to while it writes this reply. | Is: saved notes about you, kept by the app. |
| How long does it last? | Lasts: one conversation. | Lasts: across conversations, until paused or deleted. |
| What fills it? | Holds: the hidden instructions, your messages, files and the AI’s replies. | Holds: selected details, such as preferences and projects. Not every word. |
| Who controls it? | Control: the model sets the size; you decide what goes in. | Control: a switch in the app’s settings, where you can see and delete notes. |
| What goes wrong? | Problem: it fills up, and early parts get dropped or blurred. | Problem: it keeps things about you that you may not want kept. |
What is a context window?
The plain point: it is the AI’s short-term working space, and it covers only the conversation in front of it.
Anthropic, which makes Claude, defines it as everything the model can look at while it answers, including the answer itself. It also draws the line this page is about: the window is separate from what the model learned in training, and works as its “working memory”.1
More goes into it than you might think. For developers, Anthropic lists the hidden instructions (the system prompt), every message, any images and documents, and the tool settings.1 Its size is counted in tokens, small chunks of text, and each model has its own limit. The full explanation is on context window meaning.
The model itself keeps nothing between messages. Anthropic says its API is “stateless”: the whole conversation is sent again each time.2 The chat app does that resending for you, which is why a chatbot seems to remember what you said five messages ago.
What is AI memory?
The plain point: memory is an app feature that writes down some things about you and uses them in future chats. It is not the window getting bigger.
OpenAI says that with memory on, ChatGPT “can remember relevant preferences and details from your chats and other available sources.”3 It keeps selected notes, not a recording: “Memory does not retain every detail from every conversation.”4
Anthropic’s help page says Claude “saves memory as a set of individual topics as you chat”, and that memory is on by default for its Free, Pro and Max plans.5 You can read the notes: in Claude, “Everything Claude remembers is listed under Topics” in Settings, then Memory.5 More on the settings is on AI memory meaning.
The window is about this chat. Memory is about you. A bigger window does not mean the AI remembers you, and memory does not make a long chat easier for it to follow.
How do the two work together?
The plain point: saved notes only help if the app puts them back in front of the model, and that means placing them in the window.
The model only uses what is in the window when it writes a reply. So in this site’s reading, memory works by adding a few saved notes to the start of a new chat, next to your first message.Reasoning, 19 Oct 2026 — follows from Anthropic’s definition above: the window is “all the text a language model can reference when generating a response”, so anything the model uses, saved notes included, has to be in it. Those notes take up a little of the window, like anything else.
Apps also have other ways to carry things across chats, and they are easy to mix up with memory:
- Chat search. Claude can look through your old chats when you ask, which Anthropic describes simply: “Simply ask Claude to find what you discussed before.”5
- Custom instructions. Standing preferences you write yourself. Claude’s help centre says they “apply broadly to all your conversations”.6
- Projects. A folder of files and notes for one piece of work. Anthropic: “Projects give Claude persistent context for a specific body of work”.7
Why does a long chat forget its own beginning?
The plain point: that is the context window filling up, not memory failing. Turning memory on will not fix it.
Anthropic says chat apps such as claude.ai can handle a full window on a rolling “first in, first out” basis, so the oldest text goes first.1 Even before it is full, a crowded window works less well: “As token count grows, accuracy and recall degrade, a phenomenon known as context rot.”1 See context rot meaning.
The fix is to keep the window tidy. Start a fresh chat for a new task, paste only the parts that matter, and restate key instructions if a chat runs long. Context engineering covers this in depth.
Which one is the privacy question?
The plain point: memory. The window ends with the chat; memory is stored information about you that outlasts it.
- Pause or wipe. In Claude, pausing means it “keeps its existing memory” but “won’t use memory or make new memories”; resetting “Permanently deletes all memories including project memories”.5
- Private chats skip it. OpenAI: “Temporary chats do not create or update memories, even with personalization on.”8 Claude’s incognito chats are not saved to memory either.5
- Work accounts differ. On Claude’s Team and Enterprise plans, memory is “off for each member until they turn it on”.5
Memory is also only one of several data settings. Whether chats are stored or used for training are separate switches; what happens to what you type walks through them.
Forgetful long chat: tidy the context window. Chatbot knows too much about you: check memory.
Questions people ask
Does a bigger context window mean the AI remembers me?
No. A bigger window lets one chat hold more text, but it still ends with the chat. Remembering you between chats is the memory feature, a separate setting (sources above).
If I turn off memory, does the AI forget what I said earlier in this chat?
No. The current chat still sits in the context window, so it can use what you said a few messages ago. Turning memory off stops notes being carried into future chats; in Claude, pausing keeps existing notes but stops using them (source above).
Is deleting a chat the same as deleting a memory?
Not always. OpenAI says “Deleting a chat alone does not necessarily delete a separate saved memory created from that chat.”9 Delete the saved memory too, in the memory settings.
Is the context window the same as what the AI learned?
No. Anthropic separates the window from “the large corpus of data the language model was trained on” (source above). Training is fixed before you use the model; see training data meaning and knowledge cutoff meaning.
Where to go next
- “Just the definitions.” Read context window meaning and AI memory meaning.
- “Show me what happens when I press send.” See how does ChatGPT work.
- “How do I fill the window well?” Try context engineering and the token counter.
- “What else do people mix up?” See AI terms people get wrong.
- “I want to control what it keeps.” Read what happens to what you type and staying safe with AI.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is the main difference between the context window and AI memory?
- The context window covers one chat; memory saves notes about you between chats
- They are two names for the same feature
- Memory is the size limit; the context window is a privacy setting
Show the answer
Answer: The context window covers one chat; memory saves notes about you between chats. The page says the window is what the AI can see in the current conversation, while memory is a separate app feature that carries saved notes into later chats.
2. A long chat seems to forget what you said at the start. What does the page say is happening?
- Memory has been switched off
- The context window is filling up, so early parts get dropped or blurred
- The AI is being retrained on your messages
Show the answer
Answer: The context window is filling up, so early parts get dropped or blurred. The page explains that apps can drop the oldest text first when the window is full, and that accuracy slips as it fills, which is a context window problem, not a memory one.
3. In ChatGPT, does deleting a chat always delete a memory saved from it?
- Yes, the memory always goes with the chat
- Only on work accounts
- Not necessarily, so delete the saved memory too
Show the answer
Answer: Not necessarily, so delete the saved memory too. The page quotes OpenAI: deleting a chat alone does not necessarily delete a separate saved memory created from it.
- 1Anthropic, Context windows, undated documentation, read at source 19 Oct 2026“The ‘context window’ refers to all the text a language model can reference when generating a response, including the response itself. This is different from the large corpus of data the language model was trained on, and instead represents a ‘working memory’ for the model.” ↑a“Everything in the request counts toward the context window: the system prompt, every message in
messages(including tool results, images, and documents), and your tool definitions.” ↑b“Chat interfaces such as claude.ai can also manage the context window on a rolling ‘first in, first out’ basis.” ↑c“As token count grows, accuracy and recall degrade, a phenomenon known as context rot.” ↑d - 2Anthropic, Using the Messages API, read at source 5 Oct 2026 (as cited on this site’s how does ChatGPT work page): “The Messages API is stateless, which means that you always send the full conversational history to the API.”↑
- 3OpenAI, Memory FAQ, read at source 22 Sep 2026 (as cited on this site’s AI terms people get wrong): “When Memory is enabled, ChatGPT can remember relevant preferences and details from your chats and other available sources.” The page refused an automated re-read on 19 Oct 2026.↑
- 4OpenAI, Memory FAQ, read at source 13 Oct 2026 (as cited on this site’s common AI misconceptions): “Memory does not retain every detail from every conversation.”↑
- 5Anthropic, Use Claude’s chat search and memory to build on previous context, help article, read at source 19 Oct 2026“Claude saves memory as a set of individual topics as you chat.” Memory is “on by default for Free, Pro, and Max plans”. ↑aunder Settings > Memory, “Everything Claude remembers is listed under Topics.” ↑b“Simply ask Claude to find what you discussed before.” ↑cpause, Claude “keeps its existing memory” but “won’t use memory or make new memories”; reset “Permanently deletes all memories including project memories”. ↑d“Claude won’t remember your chats, so they won’t be saved to Claude’s memory or your chat history.” ↑eon Team and Enterprise plans memory is “off for each member until they turn it on.” ↑f
- 6Claude Help Center, What are skills?, read at source 13 Oct 2026 (as cited on this site’s skills vs MCP vs plugins): custom instructions “apply broadly to all your conversations”.↑
- 7Anthropic, Skills explained, 5 Mar 2026, read at source 13 Oct 2026 (as cited on this site’s skills vs MCP vs plugins): “Projects give Claude persistent context for a specific body of work”.↑
- 8OpenAI, Memory in ChatGPT, read at source 13 Oct 2026 (as cited on this site’s pasted something into AI page): “Temporary chats do not create or update memories, even with personalization on.”↑
- 9OpenAI, Memory in ChatGPT, read at source 13 Oct 2026 (as cited on this site’s pasted something into AI page): “Deleting a chat alone does not necessarily delete a separate saved memory created from that chat.”↑
Read at source 19 Oct 2026: Anthropic, Context windows; Anthropic, Use Claude’s chat search and memory to build on previous context.
Reused from this site’s pages, with the read dates recorded there: Anthropic’s Using the Messages API (5 Oct 2026); OpenAI’s Memory FAQ (22 Sep and 13 Oct 2026; it refused an automated re-read on 19 Oct 2026); Claude Help Center’s What are skills? and Anthropic’s Skills explained (13 Oct 2026).
The desk and notebook comparison, the table wording and the advice on tidying a chat are this site’s own ways of explaining, not claims from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK