What Does Perplexity Actually Log? Its Privacy Policy, Decoded

Bottom line first: Perplexity logs your queries. So does Google. So does ChatGPT. The difference is _why_ they log, _how long_ they keep it, and whether your data funds an ad business. Perplexity wins on two of those three fronts — but "better than Google" is not the same as "private."

Here is what each company's privacy policy actually says, with no filler.

The Three-Way Comparison

|---|---|---|---|

| Used for ad targeting | Yes | No | No |

The headline difference: Google's logging funds a $200B+ ad business. Your search for "cheapest diabetes medication" becomes a data point that follows you around the web. Perplexity's logging funds product improvement and subscription conversions — no ad auction, no behavioral profile sold to advertisers.

That's a meaningful distinction. It's not privacy. It's a less hostile surveillance posture.

What Perplexity's Privacy Policy Actually Says

Perplexity's published privacy policy (last reviewed June 2026) discloses the following:

What they collect:

The text of every query you submit
Follow-up messages and conversation threads
Account email, if you're signed in
IP address, browser fingerprint, and device type
Click behavior within answers (which sources you open, how long you dwell)
Integrations data if you connect Perplexity to Notion, Slack, or other services

What they do with it:

Improve their models and answer quality (query content is training signal)
Safety and abuse monitoring (human review possible)
Product analytics (how features are used)

What they don't do with it:

Sell to advertisers
Build cross-site behavioral profiles for third-party targeting
Share with data brokers

The retention gap: Unlike Google (18 months, documented) and OpenAI (30 days for deleted conversations), Perplexity's policy does not publish a specific retention window for query logs. They state data is deleted when you delete your account, but the duration of standard retention is not quantified in their public policy. This is a transparency gap worth noting.

What Google Actually Logs

Google Search ties your queries to your Google account by default. The data retention settings, buried under myactivity.google.com, default to 18 months — but you can extend this to indefinite.

What makes Google categorically different from Perplexity:

The surveillance graph: Google correlates your searches with your YouTube watch history, Gmail content, Maps locations, Chrome browsing, and Android app usage. That query about a new medication goes into a profile that also knows your doctor's office address (from Maps), your health insurance company (from Gmail), and your age (from account setup). Perplexity has none of those cross-product data points.

Advertising revenue model: Google's entire profit structure depends on turning your search intent into ad revenue. There is an institutional incentive to log everything, keep it forever, and use it to sell you things. Perplexity charges subscriptions. The incentive structure is fundamentally different.

Gemini integration: When you use Google Search with AI Overviews, your queries now feed both Search and Gemini training. The policy is clear: if Gemini Apps Activity is on (default), conversations are retained for up to 36 months and may be reviewed by human contractors.

What ChatGPT Actually Logs

OpenAI's logging practices are the most documented of the three:

Free tier (default): Conversations stored, used to train future models. Human reviewers may access flagged content. This is explicitly the tradeoff for free access.

ChatGPT Plus ($20/month): Same defaults as free — training use is on unless you opt out. Go to Settings > Data Controls > "Improve the model for everyone" to disable. Opting out currently bundles history deletion with training opt-out (they cannot be separated as of June 2026).

The 30-day window: Even after you delete a conversation or opt out, OpenAI retains query content for 30 days for abuse monitoring and safety. This is disclosed in their privacy policy and is a hard floor, not configurable by users.

Enterprise/API: No training use by default. Strong contractual protections. If you're a developer building on the API, your users' queries are not used to train GPT-5.

Compared to Perplexity, ChatGPT has better transparency (documented 30-day retention window vs. Perplexity's unspecified window) but worse defaults on the free tier (training opt-out is not the default).

The Fox Angle: You're Asking the Wrong Question

"Is Perplexity private?" is the wrong frame. The right question is: private from whom, for what?

Consider the actual threat models people worry about when they use AI search:

Threat model A: Ad targeting — Perplexity wins decisively. No ad business, no behavioral profile, no cross-site retargeting. If your concern is being sold to, Perplexity is a genuine improvement.

Threat model B: Government/legal access — Perplexity is a U.S.-based company, subject to subpoenas, NSLs, and FISA orders, same as Google and OpenAI. No cloud AI service protects you here. For this threat model, only a local LLM running on your hardware (Ollama + Llama 3.3) is adequate.

Threat model C: Data breach — All three services hold query logs. A breach at any of them could expose your research history. Perplexity's smaller footprint means a smaller breach surface than Google, but the risk exists. Encrypted notes storage (Proton Drive) protects your outputs; it cannot protect what you already sent to the cloud.

Threat model D: Your employer/school seeing your queries — Depends on your account. Signed in on a work Google Workspace? IT can see Gemini queries. On a personal Perplexity account on a work network? IT can see you're using Perplexity but cannot see query content (TLS-encrypted). Different risk profiles.

Affiliate Disclosure: This article may contain affiliate links. If you make a purchase through these links, we may earn a small commission at no extra cost to you. We only recommend products we genuinely believe in. This helps support our work and allows us to continue providing free content.

The Verdict

| | Winner |

|---|---|

| Least logging for ad purposes | Perplexity |

| Best transparency on retention | ChatGPT (30-day window documented) |

| Worst default settings | Google (36-month retention, training on by default) |

| Most honest about what you're giving up | ChatGPT (free = training data, documented) |

| Safest for truly sensitive research | None of the above — use a local LLM |

Perplexity earns its reputation as a more privacy-respecting option than Google for everyday research. It's not a privacy tool. The distinction matters: reaching for Perplexity instead of a local LLM for anything sensitive gives you a false sense of protection. Know what it buys you, and act accordingly.

Build a Research Stack That Matches Your Risk

If this comparison prompted a stack review, start here:

Everyday AI search: Perplexity Pro — better answers, cited sources, no ad profile
Sensitive document analysis: Local LLM via our Open WebUI guide
Research output storage: Proton Drive — zero-knowledge encryption, nothing readable on their servers
General privacy baseline: NordVPN for IP masking when using cloud AI without an account