Why Does ChatGPT Keep Freezing? Causes and 8 Fixes (2026)
ChatGPT keeps freezing because the browser tab runs out of headroom, not because the model has stopped. The usual triggers are a long conversation that the page has to render and re-render, a tab starved of memory, a script-blocking extension, or stale site data. OpenAI outages happen too, but they stop every chat at once. Each has a fix that takes under a minute, and the order below runs from the fastest to the most involved.
The pattern most people describe is the same: ChatGPT works for a while, then a specific chat starts lagging, scrolling stutters, the reply cursor stops, and eventually Chrome offers to kill the page. That is a client-side symptom. OpenAI's own troubleshooting article for slow responses points at cache, extensions, system resources and network, alongside a check of its status page (OpenAI Help Center), and a developer-forum analysis from August 2026 looked at why long threads are so expensive to render. This guide covers what is happening, how to tell your machine from an outage in 30 seconds, and the eight fixes. If ChatGPT will not load at all, use why is ChatGPT not working instead; if it loads but every reply is slow, see why is ChatGPT so slow.
Why ChatGPT Freezes: What Is Happening in the Tab
ChatGPT is a web app that renders every message in your conversation as live page elements, and long conversations make the page expensive to draw, scroll and update. A detailed reply on OpenAI's developer forum (August 26, 2026, in a thread opened on August 21) measured an assistant reply at roughly 460 DOM elements against about 30 for a user message, on a page that loads 39 stylesheets and about 2.2 MB of CSS. ChatGPT unmounts older turns as you scroll and rebuilds them when you scroll back, so a long chat keeps the browser doing heavy style and layout work as you move through it. The same report found the stored site data above about 4 MB whenever the problem was present, and clearing it restored normal behavior until a few more turns brought the slowdown back (OpenAI Developer Community).
Three other things stack on top. Extensions that block scripts or ads can interrupt the page's own scripts, which OpenAI lists as a cause of slow and stuck sessions. A machine or browser that is low on memory will freeze any heavy tab, and ChatGPT's is heavy. And outdated cached data can cause unexpected behavior, which is why OpenAI's general advice starts with clearing cache and cookies, and its steps for a frozen page start with a hard refresh and clearing site data. Genuine outages do happen, but they stop every chat at once and show on OpenAI's status page; a freeze that hits one long thread and spares the others is not an outage.
Is It You or OpenAI? A 30-Second Check
Open status.openai.com in another tab and open ChatGPT in an incognito window. If the status page is green and the incognito window works, the problem is in your normal browser profile, and fixes 1 to 7 are where to look. Incognito disables most extensions and uses fresh site data, so it isolates the two most common causes in one step. If the incognito window freezes on the same chat, the thread itself is the problem: go straight to fixes 2 and 3.
Fix 1: Hard Refresh, Then Reopen the Chat in a New Tab
Press Shift and Reload (Cmd+Shift+R on Mac, Ctrl+Shift+R on Windows) to reload ChatGPT without the cached build, then open the frozen chat in a fresh tab. OpenAI lists the hard refresh as the first thing to try when ChatGPT is stuck loading or frozen. If the page recovers and then freezes again within a few messages, the chat is too long for the tab; continue with Fix 2.
Fix 2: Start a New Chat When the Thread Is Long
A conversation that has run for hundreds of messages will keep freezing no matter how clean the browser is, because the page has to rebuild all of those turns as you scroll. Start a new chat and bring the context with you. Very long threads can also run past the model's context window, so a fresh chat with a summary helps on both fronts.
Ask ChatGPT, in the frozen thread if it will still respond, for a 300-word summary of the decisions and constraints so far. Copy it.
Open a new chat, paste the summary as your first message, and continue.
With AI Toolbox installed, skip the copying: type @@ in the new chat and pick the old conversation. Free accounts get three Context Mentions a day (one per message) and Premium is unlimited; on both plans, long threads are summarised automatically before they are added.
Keep the old thread findable by dropping it into a folder or pinning it, rather than scrolling back to it later.
Fix 3: Collapse Long Messages to Shorten the Thread
Collapsing long messages clips each one to a short preview a few lines tall, so a long thread becomes much shorter to scroll through. AI Toolbox adds a collapse button to every message, a collapse-all button, and a keyboard shortcut (Cmd+Shift+E on Mac, Ctrl+Shift+E on Windows) that collapses or expands the whole thread. It is free on every plan and works with the conversation you already have, without starting over. Collapsed messages stay on the page, so collapsing does not remove the rendering cost the forum report describes; it makes the thread easier to move through, and Fix 2 is still the real cure for a thread that keeps freezing.
Collapsed messages stay in the conversation and in the model's context; only how they are displayed changes. Expand a message when you need it and collapse it again when you are done. Full steps are in how to collapse long messages in ChatGPT.
Keeping a long thread responsive
Native ChatGPT
AI Toolbox Free
AI Toolbox Premium
Collapse long messages, collapse all
No
Yes, plus Cmd/Ctrl+Shift+E
Yes
See how full the context window is
No
Yes, live meter
Yes, live meter
Carry a thread into a fresh chat
Copy and paste
@@ Context Mentions, 3 per day
Unlimited, plus a handoff to a fresh chat from the context meter
Fix 4: Clear chatgpt.com Site Data, Not Just the Cache
Clearing the site data for chatgpt.com removes the cookies and stored data the app accumulates, and OpenAI's steps for a frozen page specifically say to clear the site data, not just the cache. In Chrome, open Settings, Privacy and security, Third-party cookies, then See all site data and permissions; search for chatgpt.com and select Delete (Google Chrome Help). You will be signed out, so sign back in afterwards. The forum report above found that the slowdown returned after a few more turns, which is why this fix pairs with Fix 2 and Fix 3 rather than replacing them.
One long ChatGPT thread freezing your whole browser? AI Toolbox (formerly ChatGPT Toolbox) adds collapse-all, a context window meter and @@ Context Mentions to ChatGPT, free, plus more folders, unlimited search results and bulk export on Premium. 40,000+ users, 4.7/5 on the Chrome Web Store. Install AI Toolbox free →
Fix 5: Disable Ad Blockers and Script Blockers on chatgpt.com
Extensions that block scripts or ads can stop ChatGPT's own scripts mid-stream, which shows up as a reply that stalls or a page that stops responding. OpenAI names this category specifically. Open chrome://extensions, switch off blockers and privacy tools one at a time, and reload ChatGPT after each. When you find the one responsible, add chatgpt.com to its allow list instead of uninstalling it. Extensions that add features on the page, including AI Toolbox, do not block ChatGPT's scripts or requests; if you want to rule everything out at once, the incognito test in the 30-second check does that.
Fix 6: Free Up Memory and Keep the ChatGPT Tab Active
A ChatGPT tab that freezes while other tabs are fine is usually the tab that grew the largest, and Chrome's Task Manager shows exactly how much memory it is holding. Open it from the More menu, More tools, Task Manager (Shift+Esc on Windows and Linux), and select the Memory column to sort by memory use (Google Chrome Help). Close the tabs and extensions above it. OpenAI's article lists closing unnecessary programs and tabs as a fix for this reason.
Chrome's Memory Saver helps in two directions. Turned on (Settings, Performance, Memory Saver), it deactivates tabs you are not using to save memory and help active tabs run smoothly. If Memory Saver keeps discarding ChatGPT itself while a long reply streams, add chatgpt.com under "Always keep these sites active" on the same screen (Google Chrome Help).
Fix 7: Update the Browser and Toggle Graphics Acceleration
An outdated browser or a graphics driver that misbehaves with hardware acceleration can freeze rendering-heavy pages like ChatGPT, and toggling acceleration is a quick test. Update Chrome from the More menu, Help, About Google Chrome, and select Relaunch if an update is waiting (Google Chrome Help). Then open Settings, System (chrome://settings/system), and switch "Use graphics acceleration when available" to the opposite of its current state, relaunch, and try the frozen chat. If neither state helps, set it back to how it was.
Fix 8: Switch Model, Device or App
If one chat freezes only while a reasoning model is thinking, try an instant model for that request; if the whole site freezes only in one browser, use the ChatGPT desktop app or a second browser until it clears. OpenAI's own diagnostics ask you to switch models and start a new chat to separate a model-specific problem from an account-wide one, and to try a different device or network. The desktop and mobile apps render the same conversations through a different client, so a thread that freezes in the browser may open fine there, which also tells you the data is intact.
Who Hits This Most
Developers feed ChatGPT long code blocks and long diffs, and each reply comes back as hundreds of styled elements; a debugging session that has run all afternoon is the classic freezing thread. Keep one chat per bug and collapse the solved ones.
Writers and editors keep a single drafting chat alive for a whole project. Those threads freeze after weeks, not hours, and lose the brief at the same time because they have also outgrown the context window. One chat per chapter in a folder is cheaper than one giant thread.
Analysts and researchers hit it on attachments. A large PDF or spreadsheet inflates the conversation immediately, and the reasoning models that handle documents best also produce the longest replies. Export the finished analysis and start clean for the next document.
Limitations and When Freezing Is Not Yours to Fix
None of the fixes above changes how ChatGPT renders a conversation; they reduce what the tab has to do or clear the state that is tripping it. Collapsing messages shortens the page but leaves every turn in it, so a very long thread will still be slow to open. Clearing site data is a reset, not a cure, and the forum report found the slowdown returned as the chat grew. If status.openai.com shows an incident, or every chat on every device freezes at the same moment, wait it out: OpenAI's rate limits and outages are covered in the ChatGPT limits guide. And if a freeze ends with a chat that will not open at all, the history recovery guide is the next stop; nothing in a frozen tab deletes your conversation.
Frequently Asked Questions
Why does ChatGPT keep freezing?
Most freezes happen in the browser tab: a long conversation that the page must keep rendering, a tab starved of memory, an ad or script blocker interfering with ChatGPT's scripts, or stale site data from an older build. OpenAI's help center points to cache, extensions, system resources and network, plus its status page. Genuine outages stop every chat at once and appear on status.openai.com.
Why does ChatGPT freeze on long conversations?
Because every message is rendered as live page elements, and a developer-forum measurement from August 2026 put an assistant reply at about 460 DOM elements against 30 for a user message. ChatGPT rebuilds older turns as you scroll, so a thread with hundreds of exchanges keeps the tab doing heavy layout work. Start a new chat and carry the context over, or collapse long messages with AI Toolbox so the thread is shorter to scroll.
Why is ChatGPT lagging after a long conversation?
Lag that builds over a session is the same rendering cost growing with the thread, and it is usually joined by the context window filling up. Ask for a summary, open a fresh chat and paste it, or use @@ Context Mentions in AI Toolbox to reference the old chat. The free context window meter shows how close the thread is to your plan's limit before it starts drifting.
Does clearing the cache fix ChatGPT freezing?
Often, but clear the site data for chatgpt.com rather than the cache alone, which is what OpenAI's steps for a frozen page specify. Site data includes the cookies and other data the site has stored in your browser. If the freeze returns as one chat grows, the thread is the cause and clearing data will only reset it temporarily.
Why does ChatGPT freeze while it is responding?
A reply that stalls mid-stream is usually a script or ad blocker interrupting the page, a tab out of memory, or a very long thread that cannot keep up with the streaming text. Test in an incognito window, which disables most extensions and uses fresh data. If it streams fine there, disable blockers one at a time; if it still stalls, the conversation is too long for the tab.
Will I lose my chat if ChatGPT freezes?
No. A frozen tab is a display problem; your chats are stored in your ChatGPT account, so the conversation opens again in a new tab, another browser or the ChatGPT app. The exception is a reply that was still streaming when the tab died, which may need to be regenerated. To keep a copy you control, export the thread with AI Toolbox (TXT on Free; Markdown, JSON and PDF on Premium).
Does ChatGPT freeze more on Chrome than on other browsers?
Not inherently; Chrome, Edge, Brave, Opera and Arc share the same engine, so a thread that freezes one usually freezes the others. What differs is the extension set and the memory available to the tab. In Chrome, Task Manager shows which tab is holding memory and Memory Saver settings let you keep chatgpt.com active.
Bottom Line
ChatGPT keeps freezing when the tab, not the model, runs out of room: long conversations are the main cause, with memory, blockers and stale site data behind them. Check status.openai.com and incognito first, hard refresh, then treat the thread: collapse its long messages, hand it off to a fresh chat, and keep it findable in a folder. AI Toolbox adds the collapse-all control, the context window meter and @@ Context Mentions for free, and Premium adds the handoff from the context meter, more folders, unlimited search results and bulk export, so a long thread is easier to retire before it freezes.
A Full Stack Developer with 7+ years of experience building AI productivity tools. Leads product development and frontend architecture for AI Toolbox, the Chrome extension suite (ChatGPT, Gemini, Grok, and Claude modules) that helps users search, organize, and export their AI conversations.
Liked this guide? Get the next one first.
Join 40,000+ readers getting practical prompts, new features, and subscriber-only deals. No spam, unsubscribe in one click.
New features and product updates before anyone else
Time-saving prompts and workflows for ChatGPT, Gemini, Claude, and Grok
Adi Leviim