How to do it
Pinpoint the cause with a quick test, then apply the right remedy. Don't dismantle the connection if the problem is the long chat.
- Check whether it's slow only in one chat or everywhere. Open a new conversation and ask a trivial question. If the new chat answers fast, the problem was the old conversation (too long). If it's slow there too, go to step 3.
- If the slowdown is in the long chat, start over from a leaner one. Every message of the conversation is reprocessed at every answer: a chat with fifty messages forces the AI to reread tens of thousands of words before even starting to write. Ask for a recap, copy it, open it in a new chat.
Recap this conversation into a block of context I can paste into a new chat: goal, decisions made, constraints, and current state of the work.
- If it's slow everywhere, suspect the servers. The central hours of the workday are the most congested: millions of people use the tool together and free accounts are slowed first. Check the official status page and, if you can, try again later.
- Check the connection. An unstable network, a corporate firewall, or a VPN that routes traffic badly make the AI seem stuck. Try another network (the phone's data): if from there it flies, it was the network.
- Lighten the browser. Close the useless tabs, disable the heavy extensions, do a forced refresh of the page. A chat left open for hours in a crowded tab freezes up on its own.
Feedback: after isolating the cause, the new chat or the different network give you a fast answer within a few seconds. That's the signal that you've hit the right problem.
A concrete example
Giulia writes a novel with the AI, chapter after chapter, in the same chat for weeks. At a certain point every answer takes almost a minute to start. She thinks it's the connection's fault and changes networks: no improvement. So she opens a new chat and asks the time: it answers instantly. She gets it: the novel's conversation has become enormous, and the AI has to reread it in full at every line. She asks for a recap of the plot and the characters, pastes it into a fresh chat titled "Novel - part 2," and goes back to writing at the speed of the first day.
When it does NOT work (and how to fix it)
If the answer freezes halfway
An answer that starts and then stops mid-sentence is usually a connection drop or a load spike on the server. Write "continue": in most cases the AI resumes from where it was interrupted. If it freezes again, refresh the page and relaunch the question.
If it's slow because you're using the deep-reasoning mode
Some tools have a mode that "thinks" longer before answering, designed for complex problems: it's slow by choice, not from a malfunction. If you just need a quick answer, switch to the faster, lighter mode or model the tool offers. For simple questions it's more than sufficient and answers in a few seconds.
If it doesn't answer at all and gives a limit error
On free accounts, after many requests in quick succession a limit-reached notice can appear. It's not a malfunction: it's the free-use ceiling. Wait the indicated time, reduce the frequency of the questions, or consider whether your usage justifies a paid plan.
A tip from someone who actually uses it
Don't let conversations grow endlessly. A chat is a workspace, not an archive: when a topic is closed or the conversation gets slow, save the recap and open a new chat for the next phase. Working in short, on-topic chats not only keeps them fast, it also makes the answers more precise, because the AI doesn't get lost among thousands of old messages.
Frequently asked questions
Does paying a subscription make the AI always fast?
It greatly reduces the slowdowns from overload, because paid accounts have priority on the servers during peak hours. But it doesn't solve the slowness caused by a too-long chat or a poor connection: those depend on you, not on the plan. The paid plan speeds up the server, not your bloated conversation.
Is a long chat really slower than a short one?
Yes, and it's the cause almost nobody suspects. The AI doesn't "remember" the conversation the way we do: it rereads it from the top at every question of yours. The longer the chat, the more text it has to go through before answering. It's the reason why starting over from a new chat with a recap immediately restores the initial speed.