Does Perplexity use my data to train its models?
By default, yes. For free, Pro, and Max users, Perplexity's "AI Data Retention" setting is ON, so your searches can be used to improve its models. You can opt out by turning that toggle off in Settings, though it only affects future data. Enterprise data is never used for training, and third-party model providers are barred from training on your queries.
Why — the first-principles explanation
AI models get better by learning from real usage, so your queries are valuable training fuel. That's why most consumer AI tools, Perplexity included, default to collecting and reusing your data: it improves the product and costs them nothing extra. The default is set to benefit the company, and opting out is left to you.
Perplexity exposes this as a single control called AI Data Retention in your account settings. Leave it on (the default) and your searches may feed model improvement. Turn it off and Perplexity should stop using new queries for training. The important limit: opting out generally applies only to future data, anything already absorbed into a trained model can't be neatly pulled back out.
There are two layers of protection worth knowing. First, third-party providers: when Perplexity routes your question to outside models like GPT or Claude, its contracts prohibit those companies from training on your data. Second, Enterprise: business plans are contractually excluded from training entirely, which is why privacy-sensitive organizations use them. As a caveat, an April 2026 lawsuit alleged undisclosed tracking that would contradict these policies; it's unproven and disputed, but a reminder to control your own settings and avoid entering sensitive information.
An example that makes it click
Imagine a coffee shop that records customers' orders to invent better drinks. By default, they keep a notebook of everything you order and use it to improve the menu. That's Perplexity with data retention on.
You can walk up and say, 'please stop writing down my orders', and from then on they won't. That's the opt-out toggle. But the recipes they already tweaked using your past orders? Those don't get un-baked. And if you're a corporate account with a signed contract saying 'never record us', they honor that from day one, that's the Enterprise guarantee.
How to do it
- Open Perplexity and go to Settings, then the Account or Preferences area.
- Find the 'AI Data Retention' toggle.
- Switch it OFF to stop your future searches being used for model training.
- Delete existing threads or clear your history to remove stored queries.
- For strict privacy, use an Enterprise plan and avoid entering sensitive personal data.
Key facts
- AI Data Retention is ON by default for free, Pro, and Max users.
- You can opt out by turning off the AI Data Retention toggle in account settings.
- Opting out applies to future data; data already used in training may not be removed.
- Enterprise data is never used for AI model training, per Perplexity policy.
- Third-party model providers (e.g., OpenAI, Anthropic) are contractually barred from training on Perplexity data.
An answer engine that searches the live web and cites its sources.
Affiliate link — we may earn a commission at no cost to you.▶ The 60-second explainer (script)
Does Perplexity use your data to train its models? By default, yes. For free, Pro, and Max users, there's a setting called AI Data Retention, and it's switched on out of the box. That means your searches can be used to help improve Perplexity's AI. The good news: you're in control. Go into Settings, find the AI Data Retention toggle, and switch it off. From then on, your future searches shouldn't be used for training. One catch, opting out only affects new data. Anything already baked into a model can't easily be removed, so the sooner you flip it, the better. Two extra protections worth knowing: when Perplexity sends your question to outside models like GPT or Claude, those companies are contractually banned from training on it. And if you're on an Enterprise plan, your data is never used for training, period. Bottom line: it's on by default, but one toggle turns it off, and you should never type truly sensitive info into any AI anyway.
What authoritative sources say
People also ask
How do I stop Perplexity from training on my data?
Go to Settings and turn off the AI Data Retention toggle. This stops future searches being used for training, though past data may remain.
Is Enterprise data used for training?
No. Perplexity's Enterprise plans are contractually excluded from AI model training, which is why privacy-sensitive organizations use them.
Do GPT or Claude train on my Perplexity searches?
No. Perplexity's agreements bar third-party model providers from training on data routed through Perplexity.