Cross-platform pillar
If an AI conversation appears missing, first check the right account and workspace, use native search, and check archives before assuming it was deleted. Keep original transcripts for a full record; use a reviewed handover to carry selected context into another conversation when a long thread becomes difficult to continue.
Start by checking the account, sign-in method, and workspace where the conversation was created. In ChatGPT, sidebar search matches words in chat titles and messages, including archived chats. Check Settings > Data controls > Archived chats as well. In Claude, searching past chats is available on paid plans when enabled; searches within a project are limited to that project.
Closing a browser tab does not normally delete a saved ChatGPT conversation. Regular and archived chats remain in your account unless you delete them or an applicable workspace retention policy removes them. Deleted ChatGPT chats cannot be restored, including through Support; a summary cannot recover a transcript you no longer have.
An AI context window is the amount of text a model can hold in active working memory at once, measured in tokens. Everything in the current conversation, your messages, the model's replies, and any files you pasted, has to fit inside it. Understanding this explains almost every "the AI forgot what I said" problem, and points to the fix.
A token is roughly three quarters of a word. Windows vary by model and plan and change often (last verified: July 2026): ChatGPT's window depends on the current model and plan tier, Claude's capacity varies by model, plan and product surface (see Anthropic's current context guidance), Gemini goes up to 1 million, and Grok 4 supports up to 256,000 via API and about 128,000 in the app. The whole conversation counts against it, not just your latest message, which is why long sessions fill up even when each message is short.
When the window fills you get either a hard limit (the conversation cannot accept more input) or, far more commonly, silent dropping and drift, where the oldest messages get pushed out or simply receive less of the model's attention. You are not warned. Responses just start ignoring things you established early on.
None of the major AI platforms auto-summarise your conversations for you. ChatGPT has no automatic summarisation. Claude's auto-context management (on paid plans) compresses messages when approaching the limit, but for the live conversation, not for saving a summary. Gemini and Perplexity have no automatic summarisation.
The manual workaround is to ask the AI to summarise before you leave each session, but this requires you to remember every time. thredly adds an in-page Summarize button inside ChatGPT, Claude, and DeepSeek; Gemini and Perplexity use the web-app copy-and-paste workflow. Summaries are structured (goal, decisions, outputs, next steps), stored in your account, and expire according to your plan's retention period.
Use native search to find the original conversation first. ChatGPT can search the text of your saved chats, and Claude supports searching past chats on eligible plans with the relevant setting enabled. Keep the original transcript or export for a full record. A reviewed handover carries selected goals, decisions, outputs, and next steps; it does not replace the transcript. See saving and exporting AI chat history, continuing a long ChatGPT conversation, and continuing a long Claude conversation.
Grok, like every large language model, has a context window. Grok 4 supports up to 256,000 tokens per request via API; the consumer Grok app effectively works with around 128,000. Once you approach those limits, Grok quietly drops the earliest parts of the conversation, the same pattern you see in ChatGPT and Claude. The fix is the same: compress into a structured summary, paste into a fresh thread, continue from the important context. If your long chat is a roleplay, journaling, or ongoing character thread, see how to preserve companion-chat context.
Check the right account, sign-in method, and workspace first, then use native search and check archives. ChatGPT search matches chat titles and messages, including archived chats. Claude can search past chats on paid plans when enabled, with searches inside a project limited to that project.
No, closing a tab does not normally delete a saved ChatGPT chat. Regular and archived chats remain saved unless you delete them or an applicable workspace retention policy removes them. Deleted chats cannot be restored through ChatGPT or OpenAI Support.
No. Retain the original transcript or export for a full record. Review a handover before using it to carry selected goals, decisions, outputs, and next steps into another conversation. A handover cannot recover a deleted source chat.
thredly works on the web in any modern browser. An optional Chrome extension is available for desktop Chrome and Chromium browsers. Mobile AI apps have separate storage systems.
No. The context window is how much a model can hold in one active conversation. Conversation memory is about moving the important parts of that conversation into a new chat once the window fills or the thread drifts.
Yes. A good context summary is platform-agnostic. If you worked on something in ChatGPT and want to continue in Claude, a manual summary or a thredly summary works just as well.