Does Perplexity AI train on your data?
Perplexity · AI answer engine · official site ↗
Depends on your planZDR / DPA on enterpriseOpt-out available
Does Perplexity AI train its models on your data?
It depends — Perplexity AI trains on consumer/free-tier data by default but excludes business, enterprise and API tiers.
Perplexity uses your searches to improve its AI models by default on Free, Pro and Max, but you can switch this off in settings, and Enterprise Pro plus the Sonar API are excluded from training.
Can you opt out?
While logged in, go to Account Settings → Preferences (the AI Data Retention / "AI" toggle) and turn off AI data usage. This applies going forward only and requires you to be signed in. Enterprise Pro and the Sonar API never train on your data, so there is nothing to toggle.
Zero retention / DPA
The Sonar API maintains a strict Zero Data Retention policy: nothing sent via the API is retained and customer data is never used for training. Enterprise Pro is excluded from model training with a DPA available; consumer plans retain data and offer no ZDR. DPA ↗
What the listicles get wrong
The nuance content farms miss: the opt-out only works when you are logged in and is not retroactive, and logged-out/guest searches are used for training with no toggle at all. The Sonar API ZDR is a real hard guarantee, unlike the consumer toggle which still retains data for service/security.
Verdict by plan tier
Free / Pro / Max (consumer)Trains unless you opt outSearches are used to improve AI models by default; opt out via the AI data setting while logged in. Logged-out/guest use is collected with no opt-out.
Enterprise ProNo trainingExcluded from AI training by contract; data not used to train models, with a DPA and shorter file retention (7 days).
Sonar APINo trainingZero Data Retention: no data retained beyond the immediate request and never used to train models.
Get notified when this changes
We track Perplexity AI's data-training and retention policy. Leave your email and we'll send one note if it changes.
Frequently asked questions
Does Perplexity AI train its AI models on my data?
It depends — Perplexity AI trains on consumer/free-tier data by default but excludes business, enterprise and API tiers. Perplexity uses your searches to improve its AI models by default on Free, Pro and Max, but you can switch this off in settings, and Enterprise Pro plus the Sonar API are excluded from training.
Can I opt out of Perplexity AI training on my data?
While logged in, go to Account Settings → Preferences (the AI Data Retention / "AI" toggle) and turn off AI data usage. This applies going forward only and requires you to be signed in. Enterprise Pro and the Sonar API never train on your data, so there is nothing to toggle.
Does Perplexity AI offer zero data retention (ZDR) or a DPA?
The Sonar API maintains a strict Zero Data Retention policy: nothing sent via the API is retained and customer data is never used for training. Enterprise Pro is excluded from model training with a DPA available; consumer plans retain data and offer no ZDR.
Is Perplexity AI safe to use with confidential or proprietary data?
It depends on your plan tier. Free / Pro / Max (consumer): Searches are used to improve AI models by default; opt out via the AI data setting while logged in. Logged-out/guest use is collected with no opt-out. Enterprise Pro: Excluded from AI training by contract; data not used to train models, with a DPA and shorter file retention (7 days). Sonar API: Zero Data Retention: no data retained beyond the immediate request and never used to train models. Always confirm current terms with Perplexity before sending confidential data — this is cited public information, not legal advice.
Sources
https://www.perplexity.ai/hub/legal/privacy-policy
https://docs.perplexity.ai/guides/privacy-security
This page is cited public information, not legal or compliance advice. Whether Perplexity AI trains on your data, and any zero-retention or DPA option, can depend on your specific plan, region and contract. Always confirm current terms with Perplexity before sending confidential or proprietary data.