ChatGPT vs Perplexity: can you switch training off?
Both ChatGPT and Perplexity train on what you type. OpenAI gives you a switch at privacy.openai.com under "do not train on my content". Perplexity's privacy policy describes no such switch. It keeps data as long as needed, with no fixed period.
Some buttons on this site can earn us a commission. You pay the same price.
ChatGPT
AI chatbot that trains on your chats by default unless you opt out.
Perplexity
Answer engine with citations that trains on your chats by default.
From $17/mo
| Feature | ChatGPT | Perplexity |
|---|---|---|
| AI protection | ||
| Trains by default | Yes, until you switch it off | Yes, according to its privacy policy |
| Switch it off at | privacy.openai.com, "do not train on my content" | No switch described in the privacy policy |
| Temporary chat | Yes, Temporary Chat stays out of training | Not stated |
| How long chats are kept | Not stated on the training page | As long as needed, no fixed period |
| Jurisdiction and servers | Not stated | Not stated |
| Business version | ChatGPT Business does not train on your business data by default | Not covered |
| What each provider says about its own AI feature. | ||
| Our score | n/a | n/a |
| Price | n/a | From $17/mo |
| Free plan | ||
| Best for | Anyone comparing mainstream AI assistants on privacy | Anyone who wants AI answers with citations |
| Visit ChatGPT | Visit Perplexity | |
Our verdict
ChatGPT is the safer choice for your data, as long as you switch off training on day one. Both services learn from your conversations by default. Only OpenAI tells you how to stop it. Use "do not train on my content" at privacy.openai.com, or a Temporary Chat for one-off questions.
Perplexity's privacy policy says your data helps "improve or create" its AI models. It describes no setting that stops this. It keeps data "for only as long as necessary", with no fixed period. OpenAI's training page names no retention period either.
Perplexity does make a claim easier to check, because every answer links to the sources behind it. Perplexity Pro costs $17 a month on yearly billing, $204 up front.
A private alternative: Lumo from Proton
Lumo is the AI assistant from Proton and it does not train on your chats. Its no-logs policy and zero-access encrypted chat history keep your questions and answers out of reach.
“Lumo doesn’t use your chats to train the AI models. We run the models on servers we control and never send your data to any third parties, except with your express permission in giving anonymized feedback for the Apertus 1.5 model.”LumoThis button can earn us a commission. You pay the same price.
Frequently asked questions
Does Perplexity train on my searches?
Yes, by default. Its privacy policy names "improve or create services and products, including our AI models" as a use of your data. The one exception it names is its email integration. The content of those emails is not used to train or fine-tune models. The policy describes no switch to turn training off. Look through your account settings before you share anything sensitive.
How do I stop ChatGPT from using my chats for training?
Open privacy.openai.com and choose "do not train on my content". For a single sensitive question, start a Temporary Chat instead. It stays out of your chat history and out of training. Where that setting sits inside the app is covered in OpenAI's Data Controls FAQ.
How much does Perplexity Pro cost?
Perplexity Pro costs $17 a month, but that rate applies to yearly billing, so you pay $204 at once. The Perplexity pricing page we read showed only the annual rate. Perplexity and ChatGPT both have a free tier as well.
Is ChatGPT or Perplexity better for checking if a message is a scam?
Perplexity makes the checking easier, because every answer carries citations you can open yourself. ChatGPT can search the web too. It also reads a screenshot or an uploaded file. So a suspicious email can go in exactly as you got it. Remove names, account numbers and addresses before you paste anything in. Both train on chats by default.