ChatGPT profiles you. Claude forgets you on purpose. Gemini files you inside your Google Account. These are not three memory settings, they are three opposite memory architectures. ChatGPT builds a profile about you and injects it automatically. Claude starts each chat blank and reaches into your raw past conversations only when memory is on. Gemini keeps your personalization inside your Google Account with a retention slider you control. Pick the wrong mental model and you will either over-trust what an assistant knows or wonder why it forgot you.
AI assistant memory is the system that lets a chatbot recall facts and context from your past conversations. Every figure below is checked against the vendor's own help pages and reporting. Memory features move month to month, so a dated changelog sits near the end. The architecture is the part that holds.
ChatGPT remembers you whether you ask it to or not. Claude forgets you unless you ask. That single default decides almost everything else about how each one treats your data.
The short verdict
- ChatGPT, the eager profiler. It remembers the most by default. It auto-injects a compressed profile, and its Dreaming update keeps that profile current in the background with less manual curation.
- Claude, the auditable amnesiac. Every conversation begins blank. Turn memory on and Claude shows you the searches it runs across your raw chats, with citations back to the source conversation.
- Gemini, the account-native filer. It gives the most retention control. Personalization lives in your Google Account, learns from past chats, and an auto-delete window you set to 3, 18, or 36 months, or indefinite, governs it.
Three architectures, side by side
One table shows the divide. Read each column as a philosophy, not a feature list.
| Dimension | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Default state of a new chat | Personalized from the start: profile injected automatically | Blank slate: nothing recalled until you ask or memory acts | Personalized from past chats when Personal context is on |
| How it stores you | Saved memories plus a synthesized memory built across chats | 24-hour synthesis of key insights, plus on-demand search of raw chats | Personal context profile held inside your Google Account |
| Recall mechanism | Auto-inject compressed profile, optional reference to chat history | Visible tool calls that search your actual conversations, with citations | Learns from Gemini Apps Activity to tailor future replies |
| Transparency | Readable saved memories; deleting a chat does not delete memories derived from it | High: searches appear inline and link to source chats | Activity log viewable and editable in Google Account |
| Retention control | Toggle memory and chat history off; Temporary Chat for no memory | Toggle memory and search off; Temporary Chat for no memory | 18mo default; 3/36mo; indefinite; 72h for temporary chats |
| Free tier memory | Yes; Dreaming rolling out to Free users | Yes since March 2026 (chat search stays paid-only) | Yes, tied to your Google Account |
Does ChatGPT remember conversations automatically?
Yes. ChatGPT builds a profile from two layers and injects it at the start of every chat without you asking. Saved memories are explicit facts the model chose to keep, like your name or dietary preferences. They behave like custom instructions that update themselves rather than needing manual edits, per the OpenAI Memory FAQ. Reference chat history, which OpenAI launched in April 2025, lets ChatGPT draw on patterns from your earlier conversations even when nothing was formally saved.
Say you like Thai food once, and a later question about lunch may factor it in. OpenAI lets you review your saved memories to see what the system keeps about you. You can switch saved memories and chat history off independently in Settings, Personalization, Memory. Temporary Chat gives you a session that writes nothing back.
What this means for you
ChatGPT feels like it knows you fastest, because it does. The tradeoff: personalization is the default, so an injected profile can color answers in ways you did not request. For a sensitive question, use Temporary Chat rather than trusting that nothing carried over.
Does Claude remember past chats?
Only when memory is on. Claude starts every chat with no prior knowledge. Memory is an opt-in layer on top of that empty start, and enabling it switches on two things. Chat search lets you ask what you discussed before, and Claude runs a retrieval over your actual past chats, shown as tool calls right in the conversation, with citations that link back to the original thread. Separately, Claude builds an automatic synthesis of key insights across your history that updates every 24 hours and feeds context into each new standalone chat.
Claude walls off projects. Each project keeps its own separate memory space and dedicated summary, so context stays scoped rather than bleeding between unrelated work. You can edit what Claude remembers by telling it so inside a chat, and changes apply to your next conversation. The memory feature reached the free tier for all users in March 2026, though chat search remains limited to paid plans, per Anthropic's help center and reporting from Engadget.
What changed: earlier write-ups described Claude as pure raw-chat retrieval with no summaries at all. Anthropic's current help center now documents a generated memory synthesis updated every 24 hours alongside the visible search. Both were true at different dates, which is exactly why a dated changelog matters.
What this means for you
Claude is the easiest to audit. You watch it search, see which chat a claim came from, and delete that conversation if you want it gone. If you care about knowing exactly why the assistant said what it said, this design wins. The cost is friction: free users get synthesis but cannot search their own history on demand.
How long does Gemini keep my conversations?
Gemini keeps conversations 18 months by default, changeable to 3 months, 36 months, or indefinite. Personalization lives in your Google identity rather than a standalone feature. Under Personal context in settings, Gemini learns from your past chats to personalize future responses, and you can give it standing instructions that apply across every conversation, per Google's Gemini Apps Privacy Hub. Gemini stores your conversations in Gemini Apps Activity inside your Google Account, which Google may also use to improve its models.
The control that sets Gemini apart is retention. Temporary chats, and any chats you have when Keep Activity is off, last 72 hours. Google keeps that 72-hour window for safety review even in your no-history mode, so temporary does not mean instant deletion.
What this means for you
If your life already runs on Google, Gemini's memory feels native and the retention slider is genuinely useful. The flip side is concentration: your AI memory sits next to your mail, files, and search history under one account. The convenience and the exposure are one fact. Set the retention dial deliberately rather than leaving the 18-month default in place.
A dated changelog
Here is what shifted recently, with dates so you can tell whether an older guide is stale.
- March 2, 2026: Anthropic brought memory to Claude's free tier, so every user gets the synthesis (Engadget). Chat search stayed a paid feature (Anthropic Help).
- March 26, 2026: Google launched switching tools letting you import chat history and memory from ChatGPT or Claude into Gemini, rolling out globally but not yet in the EEA, Switzerland, or the UK (9to5Google).
- June 4, 2026: OpenAI began rolling out its Dreaming update for ChatGPT memory to Plus and Pro users in the US, with Free, Go, and international users following after. It synthesizes across past chats automatically and is time-aware, rewriting a stale fact like a finished trip on its own. OpenAI reports the new system hits 82.8% factual recall, 71.3% preference adherence, and 75.1% time-sensitive accuracy (OpenAI, TechTimes).
What changed: Dreaming has a cost. Deleting a conversation no longer removes the memories ChatGPT derived from it, so you have to delete both the chat and the memory entry, and logs of deleted saved memories may persist up to 30 days (TechTimes). More automatic recall, less clean a delete.
How to choose
- Want the assistant to just know you with minimal setup: ChatGPT. Its default-on profile, now self-updating via Dreaming, does the most remembering for you.
- Want to see and verify what it knows: Claude. Visible searches and source citations make recall auditable, and the blank-slate default keeps unrelated chats from leaking in.
- Want memory governed by a clear retention dial inside an ecosystem you already use: Gemini. The 3, 18, and 36-month options plus indefinite give explicit control, if you accept centralizing data in your Google Account.
Here is what most comparison guides will not tell you: none of the three remember your stuff. They remember conversations. A photo of a receipt, a PDF lease, a voice note from a meeting, a WhatsApp thread with a plumber, none of that lives in chat memory. Chat memory captures what you typed into one app. It does not hold the documents and media that make up most of what you actually need to find again.
Where a dedicated memory app fits
MemX is built for that gap. It is a consumer memory app, a personal second brain, where you capture photos, PDFs, documents, voice notes, and WhatsApp messages, then retrieve any of it by asking in plain English. MemX answers from your own saved material and cites the source memory, so recall is verifiable the way Claude's chat search is, but across your real files rather than past prompts. It reads text inside photos with OCR, so a snapshot of a receipt is searchable by its contents. Tell it a reminder naturally and it writes the event to your calendar, recurring included.
On privacy: your data is encrypted, never used to train AI, and inaccessible to MemX staff. The free tier needs no credit card, on Android, iOS via TestFlight, and WhatsApp. Use a chat assistant for thinking and drafting. Use a memory app for the documents and moments you need to find again.
Turn on Temporary Chat (ChatGPT and Claude) or Keep Activity off (Gemini) for anything sensitive, and keep the durable record of your life, the receipts, scans, and notes, in a place built to store and cite them rather than in a chat log governed by a vendor's retention policy.
Written by Aditya Kumar Jha, who works on AI memory at MemX.
01Does ChatGPT remember conversations automatically?
Yes. ChatGPT injects a profile built from saved memories and your chat history by default, per OpenAI's Memory FAQ. After the Dreaming update it keeps that profile current on its own. To stop recall, toggle memory off or use Temporary Chat.
02Does Claude remember past chats?
Only when memory is enabled. Claude starts each conversation empty, then builds a 24-hour synthesis of key insights and can search your raw past chats with citations, per Anthropic's help center. Memory reached the free tier in March 2026; chat search stays paid-only.
03How long does Gemini keep my conversations?
By default 18 months, changeable to 3 months, 36 months, or indefinite in your Google Account, per Google's Gemini Apps Privacy Hub. Temporary chats, and chats made with Keep Activity off, last 72 hours before deletion.
04Can I move my memory between ChatGPT, Claude, and Gemini?
Mostly one direction. Since March 2026, Gemini can import chat history and memory from ChatGPT and Claude via switching tools, per 9to5Google. It was not available in the EEA, Switzerland, or the UK at launch. There is no universal two-way memory standard.
05Which AI assistant has the most private memory?
Each wins on a different axis. Gemini gives the clearest retention controls but concentrates data in your Google Account. Claude is the most auditable, showing sources. For all three, a dedicated memory app with encryption and no training on your data keeps personal files separate from chat logs.
