aifree.dev

Why doesn’t it remember me in a new chat?

Chatbot memory: fixed weights, a fresh context per chat, saved notes pasted into the prompt · Beginner · 5 min · context, llm · after System prompts and instruction priority

A chatbot’s model doesn’t change while you talk, and each new chat starts with no trace of earlier ones: it knows only what is sent in that chat. “Memory” features save notes about you and paste them into the prompt.

Before you start: could you explain this to a friend?

1 = not at all · 5 = easily

  1. 1 Guess first

    You told a chatbot your name last week. Today, in a brand-new chat, it greets you by name. Where did it get your name?

  2. 2 Look

    What the model is sent, chat by chat

    1 / 3

    Chat 1, yesterday · context sent: 34 tokens

    1. system prompt (app)6 tokensYou are a helpful assistant.
    2. you12 tokensHi, I'm Lan. I'm vegetarian and I'm learning Python.
    3. it16 tokensNice to meet you, Lan! Happy to help with Python or veggie recipes.

    model weights · the same file in every chat, for every user; only read, never written

    Illustrative chat; tokens measured with o200k_base (GPT-4o), tiktoken 0.14.0, 2026-10-23

    Chat 1: the app sends everything said so far. That text is the only place “Lan” exists.

    A new chat is a new context, and memory is just more text in it.

  3. Training set the model’s weights before you arrived, and chatting doesn’t change them: the same file answers everyone. Each reply is computed from one thing, the context the app sends with that request. A new chat sends a new, nearly empty context, so yesterday simply isn’t there.

  4. 4 You are the model

    You are the model. In three new chats you see only what the app sends you. Answer each one.

    new chat 1 of 3 · all you are sent:

    1. system prompt (app)6 tokensYou are a helpful assistant.
    2. you4 tokensWhat's my name?

    Illustrative chats; tokens measured with o200k_base (GPT-4o), tiktoken 0.14.0, 2026-10-23

    You are the model. You see only this chat’s context. Answer.

  5. 5 Check

    In the picture, what is different between “memory off” and “memory on”?

  6. Memory features do this note-pasting for you. ChatGPT can draw on saved memories and past chats; Claude saves topics and carries them into new chats. ChatGPT’s temporary chats and Claude’s incognito chats save no new memories. Notes take room in the context window, and a stale note is repeated confidently until you edit or delete it.

  7. 7 Say it back

  8. 8 Mission · 5 min

    In ChatGPT Free, send “Remember that I’m vegetarian.” Open a new chat and ask “Suggest a dinner for me.” Look in Settings → Personalization → Memory for the note (if Memory is on for your account). Then start a Temporary Chat, choose Unpersonalized, and ask again. Which chat knew, and where did it read it?

    Test ChatGPT Free’s memory

Three quick ones

Drawn at random from this lesson's pool; different next time.

Done?

Now: could you explain this to a friend?

1 = not at all · 5 = easily

Answer the three quick ones first.

Get an email when new lessons come out

New lessons arrive in the weekly digest, only in a week that has one. You confirm by email first; unsubscribe in one click.

Go deeper

Sources

  • OpenAI API docs, “Conversation state” (https://developers.openai.com/api/docs/guides/conversation-state, checked 2026-10-23): “each text generation request is independent and stateless”; multi-turn chats work by sending earlier messages again
  • Anthropic docs, “Using the Messages API” (https://platform.claude.com/docs/en/build-with-claude/working-with-messages, checked 2026-10-23): “The Messages API is stateless, which means that you always send the full conversational history to the API”
  • OpenAI Help Center, “Memory in ChatGPT” (https://help.openai.com/en/articles/8590148-memory-faq; read 2026-10-23 from the Wayback Machine copy of 2026-09-19, the live page refuses automated reads): memory can draw on past chats, saved memories, custom instructions, Library files and connected apps, and features vary by plan and region; Settings → Personalization → Memory; deleting a chat does not necessarily delete a saved memory made from it; temporary chats do not create or update memories, and Unpersonalized makes one use no memories; with “Improve the model for everyone” on, OpenAI may use chats and remembered information to improve models
  • BleepingComputer, “ChatGPT rolls out memory upgrade for free users” (Mayank Parmar, 2025-06-03, https://www.bleepingcomputer.com/news/artificial-intelligence/chatgpt-rolls-out-memory-upgrade-for-free-users/, checked 2026-10-23): OpenAI started rolling out memory improvements to logged-in Free users on 2025-06-03
  • Claude Help Center, memory article (https://support.claude.com/en/articles/11817273, checked 2026-10-23): memory is on by default for Free, Pro and Max; Claude saves individual topics as you chat and carries them into new conversations; searching past chats is for paid plans; incognito chats are not saved to memory or chat history
  • The chats in the picture and the you-are-the-model step (Lan, vegetarian, Python, the dinner answers) are illustrative. Their token counts are measured with OpenAI’s tiktoken 0.14.0, o200k_base (GPT-4o), on 2026-10-23: system prompt 6, chat 1 messages 12 + 16 = 28, “What should I cook tonight?” 6, saved-memories note 12
  • aifree.dev catalog, chatgpt-free (verified 2026-10-12): free text chats on chatgpt.com

Published Oct 23, 2026 · updated Oct 23, 2026

Embed this

Paste this into a blog post or course page: the lesson's main picture in an iframe, with a link back to the full lesson.

<iframe src="https://aifree.dev/embed/explain/memory" width="100%" height="620" style="border:0;max-width:720px" title="Why doesn’t it remember me in a new chat? (aifree.dev Explain)" loading="lazy"></iframe>
<p><a href="https://aifree.dev/explain/memory">Why doesn’t it remember me in a new chat? Interactive lesson on aifree.dev</a></p>

Never miss new offers and news

New tools, credits, and AI news are added regularly. Pick how you want to hear about them.

Telegram channel
Join @aifree_dev and get new free AI offers and news right in your Telegram feed.
Join the channel
Browser notifications
Turn on push notifications and get an alert on this device whenever something new is posted. No account needed.
Email newsletter
A weekly digest of new, changed and expiring free AI offers. Double opt-in, one-click unsubscribe. See the latest issue.

Unsubscribe anytime. Privacy

Frequently asked questions

What is aifree.dev?

aifree.dev helps everyone access AI through a curated directory of free AI tools, API credits, free tiers, and startup programs for coding, image, voice, video, and students - verified regularly.

How are offers presented?

Each offer includes its access details, free models when available, and expiration information when provided. Limited-time deals stay visible so they are easy to compare.

Do education and startup programs need verification?

Yes. Providers usually ask students to verify their status with a school email, student ID, or a program like the GitHub Student Pack, and ask startups to verify company details such as age, funding, or website. Check each offer’s claim page for the exact requirements.

What kinds of free AI offers are listed?

Free tiers and free trials from AI providers, free API credits, startup and student programs, and free tools for coding, image generation, voice, and video.

How can I suggest a tool or report a broken link?

Use the submit form: paste the offer page URL and a short note. A person checks every submission before anything is listed. Broken links and changed terms go through the same form.

+

Know a good one?

Tell us about an AI tool worth knowing. Found a broken link? We want to hear that too.

Submit an offer