Compare
Does Claude delete chats after 30 days while ChatGPT keeps them?
The short answer is no, not in the way viral posts claim. Here is how Claude and ChatGPT actually retain chat history, what the 30-day window really means, and when policy toggles are not enough.
Snapshot
Key takeaways
Neither Claude nor ChatGPT auto-wipes your visible chat history after 30 days. Saved chats stay in your account until you delete them.
On Claude consumer plans, the famous 30-day figure is mainly a backend retention default when you do not allow data use for model improvement. Opting in can extend retention up to five years for eligible chats.
On ChatGPT, Temporary Chats and deleted chats are typically purged from systems within about 30 days, but normal history remains until you remove it.
Training use and retention are related but not identical. Turning off improve-the-model settings does not mean the vendor never stores your prompts.
If the content is sensitive, policy PDFs are a weak control. Local or private AI keeps the prompt off a vendor chat log in the first place.
The short answer
People searching whether Claude deletes chats after 30 days while GPT keeps them are usually comparing consumer Claude.ai and ChatGPT privacy. The viral framing is outdated and too simple.
Both products keep conversation history in your account so you can reopen old threads. Both also describe roughly 30-day windows for some backend deletion paths, Temporary or Incognito-style chats, or post-delete cleanup. Neither product is accurately summarized as Claude auto-deletes everything after a month while ChatGPT keeps chats forever.
What actually differs is the mix of defaults, training toggles, plan tiers, Temporary Chat behavior, and how long data may be retained if you allow model improvement. Those details change over time, so treat vendor privacy centers as the source of truth and re-check after major policy updates.
What the old Claude vs ChatGPT privacy claim gets wrong
Older blog posts often said Claude deletes chats after 30 days to protect privacy, while GPT retains chat logs indefinitely for training. That story mixed three different ideas into one slogan:
- Visible chat history in your account
- Backend retention after you delete a chat
- Whether conversations can be used to improve models
Those are not the same clock. A chat can sit in your sidebar for months, still be excluded from training, and still exist on servers. Or you can delete it from the UI and still wait for a backend purge window. Or you can allow model improvement and accept a longer retention period for eligible data.
If you care about Claude vs ChatGPT privacy, ask the precise question: is this about history you can see, logs the vendor keeps, training use, legal holds, or whether the prompt ever needed to leave your device?
Important
30 days is not a magic privacy guarantee
A 30-day retention line in a privacy FAQ does not mean every prompt vanishes from the product after a month. It usually describes a backend window, a Temporary Chat mode, or what happens after you delete. Read the setting names carefully before you paste anything sensitive.
How Claude chat retention works
On Claude consumer products such as Free, Pro, and Max, your prompts and conversations generally remain available in your dashboard so you can revisit history. Anthropic's privacy materials say you can delete conversations, which removes them from your history immediately and targets backend deletion within about 30 days, subject to legal or policy exceptions.
The more important modern detail is the model-improvement choice. Anthropic has stated that if you allow your data to be used to improve Claude, retention for eligible new or resumed chats and coding sessions can extend up to five years. If you do not allow that use, the shorter retention path, commonly described as about 30 days for that backend retention context, continues to apply.
So does Claude delete chats? Yes, when you delete them, and Temporary or Incognito-style chats are designed for shorter-lived storage. No, Claude does not simply wipe every ordinary saved chat from your account after 30 days while you keep using the product.
Commercial surfaces such as Claude for Work, Education, or API plans can follow different rules. Do not assume the consumer Claude.ai story applies to a company workspace or an API integration.
How ChatGPT chat retention works
ChatGPT also keeps normal chats in your history until you delete or archive them. OpenAI's data controls let signed-in users turn off Improve the model for everyone so conversations are not used to train ChatGPT. That toggle is about training use. It does not by itself mean chats disappear from your account.
Temporary Chats are the closer match to the 30-day meme. OpenAI says Temporary Chats are deleted from systems after 30 days, are not used to train models, do not create memories, and do not stay in your history. That is a mode choice, not the default behavior of every ChatGPT thread.
When you delete a ChatGPT conversation or your account, OpenAI describes removing it from your account immediately and scheduling permanent deletion from systems within about 30 days, unless legal, security, or similar obligations require longer retention. Team, Enterprise, and Edu plans can add admin-controlled retention settings.
In other words, ChatGPT does keep chats in the everyday sense that history is a product feature. It does not mean every prompt is retained forever with no controls. ChatGPT chat retention depends on whether the chat is normal history, Temporary Chat, deleted, used for training, or held under a legal exception.
Claude vs ChatGPT retention, side by side
For consumer chat, the practical comparison looks more like this:
- Visible history: both keep saved chats until you delete them.
- Temporary or Incognito-style chats: both offer shorter-lived modes that avoid normal history and training use.
- After you delete: both commonly describe roughly 30-day backend cleanup, with exceptions.
- Training: both expose improve-the-model style controls. Claude's consumer materials also tie longer retention to allowing model improvement.
- Enterprise or API: both can differ sharply from the free consumer story.
That is why Claude data retention and ChatGPT keep chats searches often talk past each other. One person means the sidebar history. Another means training corpora. Another means whether a deleted thread is gone from backups yet.
If your real goal is AI chat history retention with minimal vendor exposure, the bigger gap is not Claude versus ChatGPT. It is cloud chat versus local inference. Policy toggles still assume the prompt reached a remote host.
What to check before you trust either product with sensitive chats
Use this as a quick audit, not as legal advice. Settings and plan names change.
Find the training or model-improvement toggle
On ChatGPT, look for Improve the model for everyone. On Claude, check data privacy or model improvement controls.
Decide whether you need Temporary or Incognito chat
Use it for one-off sensitive prompts you do not want in history, then still avoid secrets you would not email.
Delete chats you no longer need
Deletion starts the backend cleanup clock. Leaving old threads forever is a choice, not a requirement.
Confirm which plan you are on
Free, Pro, Team, Enterprise, and API surfaces can have different retention and admin controls.
Re-read the privacy center after major updates
Retention windows and training defaults have changed before. Screenshots from old blogs go stale.
Classify the content before you paste
Customer data, health details, credentials, and unpublished strategy usually do not belong in a consumer chatbot.
Tip
If the answer must not leave your machine, skip the retention debate
Claude vs ChatGPT privacy settings matter for everyday cloud use. For material you would not put in a shared drive, prefer private local AI so the prompt never becomes a vendor chat log in the first place.
When local AI is the better answer than retention settings
Retention policies are promises about what a remote operator does after your content arrives. Local AI changes the architecture. If inference runs on your device, the default question stops being how long Claude or ChatGPT keeps the log and becomes whether anything needed to leave the laptop at all.
That does not make every local setup perfect. A local app can still sync, crash-report, or call a cloud fallback. It does make the privacy model easier to reason about when the product is honest about on-device inference.
Use cloud Claude or ChatGPT when convenience matters and the content is low sensitivity. Use Temporary Chat modes and training opt-outs when you stay in the cloud but want tighter defaults. Move to private or offline AI when the content is confidential, regulated, or simply none of a vendor's business.
For the broader map of risks, see AI privacy. For the product category people adopt when they want lower exposure, see private AI.
FAQ
Related reading
Broader risks around logging, training data, and exposure.
What private AI means when architecture matters more than policy.
Why no-account wrappers are not the same as private chat.
When keeping inference offline changes the data path.
Running models on infrastructure you control.
Private local AI on your device.
Want AI chat that never needs a vendor retention policy?
Six months of Pro free.
Unltd AI is private local AI for people who would rather keep prompts on-device than debate how long Claude or ChatGPT keeps a log.