- 🕘 Signed-in Perplexity sessions are stored in History indefinitely until you delete them, while logged-out sessions expire after 14 days and Incognito sessions expire within 24 hours.
- 🔐 Training and storage are separate controls: turning off AI Data Retention stops future consumer search data from being used for model training, but it does not automatically erase saved History.
- 🗑️ Account deletion is the strongest consumer deletion action documented by Perplexity, with personal data scheduled for permanent deletion from its servers within 30 days.
- 🌐 Comet adds a second privacy layer because ordinary browsing data can stay local by default, while page, history, or tab context sent to the Assistant may be retained for up to 30 days.
- 🏢 Enterprise and API products follow materially different rules, including no AI training for Enterprise data and a Zero Data Retention policy for Sonar API prompt and response content.
- ✅ The safest choice for a sensitive query is to minimise the data you submit, use Incognito when appropriate, turn off AI training before the query, and delete the relevant session afterwards if you do not need it.
Yes, Perplexity stores your search history when you are signed in, and its current Help Center says those saved sessions can remain in History indefinitely until you delete them. The important catch is that “search history” is not one single privacy setting. I found five separate data lifecycles behind a normal-looking Perplexity query: visible session history, eligibility for AI training, short-lived anonymous or Incognito sessions, Comet browser context, and account-level personal data retention. Treating all five as the same thing is where most privacy advice becomes misleading. Perplexity also says logged-out sessions disappear after 14 days, Incognito sessions expire within 24 hours, and personal data is removed from its servers within 30 days after account deletion. Its consumer AI Data Retention setting is a separate switch again, because opting out affects future use of search data for training without automatically deleting the sessions you can still see in History.
I reviewed the leading live results for “does Perplexity store my search history” before structuring this guide. The strongest official pages explain retention, data collection, sessions, deletion, Enterprise controls, API privacy, or Comet in isolation. Comparative privacy articles tend to collapse those layers into a simple “private versus not private” verdict. This article takes a different route: it follows one search from the moment you type it, through saving, possible training use, temporary retention, deletion, and account erasure. That structure matters because a user can delete a visible thread while still misunderstanding training, or opt out of training while leaving years of account history untouched.
The result is a practical answer rather than a privacy slogan. You will know what Perplexity saves, what its settings actually change, which timelines are documented, where the documentation leaves ambiguity, and which action fits the level of sensitivity in your query.
The Five Data Lifecycles Behind One Perplexity Search
A single Perplexity search can participate in several data systems at once. That is the most useful mental model I found during this review. The visible conversation in History is only the first layer. A separate setting determines whether eligible consumer data can improve Perplexity’s models. Incognito and logged-out use have their own expiration windows. Comet can introduce browser-derived context. Account deletion applies at a broader identity level.
| Data Layer | What It Means | Default Consumer Behaviour | Main User Control |
| Saved session history | The query, follow-ups, answers, and sources shown in your account History | Signed-in sessions are saved and can remain indefinitely | Delete one session or delete all sessions |
| AI training eligibility | Whether eligible search data can be used to improve Perplexity models | Enabled by default for Free, Pro, and Max | Turn off AI Data Retention before future searches |
| Temporary history | Logged-out, Incognito, or context-sensitive temporary sessions | 14 days logged out; Incognito expires within 24 hours | Sign in to save, or use Incognito for short-lived use |
| Comet Assistant context | Page, tab, or history context sent when a browser task requires it | Can be retained for up to 30 days when used by Assistant | Delete the temporary thread, disable Assistant, or block sites |
| Account-level personal data | Profile information and other personal data linked to the account | Retained while the account is active, subject to Perplexity’s stated purposes | Delete the account, triggering a documented 30-day deletion window |
Perplexity’s August 2026 session documentation says signed-in sessions are stored in History indefinitely and can be deleted individually or in bulk. Its July data-collection guidance says AI Data Retention is enabled by default for Free, Pro, and Max users, while the retention Help Center says personal information remains while the account is active. Those statements are compatible, but they answer different questions.
This is also why a guide focused only on the “Delete” button is incomplete. If your concern is what still appears in your account, delete the thread. If your concern is future model training, change AI Data Retention. If your concern is the account and associated personal data as a whole, account deletion is the more relevant action.
For a practical walkthrough of the interface, our guide to deleting Perplexity history covers the user-facing deletion steps. The larger privacy question, however, only makes sense when those steps are placed inside the full data lifecycle.
What Perplexity Saves When You Are Signed In
When you use Perplexity while signed in, your searches become sessions. Perplexity describes a session as a conversation history containing the initial question, follow-up queries, the AI’s responses, and the sources used to produce those answers. The same documentation says every signed-in session is stored in your History indefinitely unless you delete it. That is the clearest official answer to the core search-history question as of September 2026.
The word “indefinitely” deserves attention. It does not mean Perplexity promises to keep your history forever under every circumstance. Product changes, account status, legal obligations, data-retention controls, or deletion can alter the lifecycle. It does mean there is no normal automatic expiration window documented for standard signed-in consumer sessions. If you keep an active account and never delete a session, the product is designed to make that session available later.
This behaviour is useful. Perplexity’s History lets you reopen research, search previous sessions by keyword, continue earlier conversations, and organise work into Projects. The privacy trade-off is equally obvious: convenience requires persistence. A search about a job move, medical concern, legal dispute, personal finance problem, or confidential client topic can remain attached to your account long after you have forgotten asking it.
Josh Joseph, Chief AI Scientist at Harvard’s Berkman Klein Center, described the broader memory problem in 2026 with a useful warning: “You would never know why the advice felt off.” His example concerned persistent chatbot memory influencing later recommendations, not Perplexity search history specifically, but the lesson transfers. Stored context can be valuable precisely because it changes future access and personalisation, which means users should know which information they are choosing to keep.
If you cannot find a saved session, do not assume it was erased. Our Perplexity search history disappeared guide covers account mismatches, logged-out sessions, temporary threads, and navigation issues that can make retained history look missing.
Logged-Out and Incognito Searches Follow Different Timelines
Perplexity does not treat every search as a permanent signed-in session. Logged-out use and Incognito mode are explicitly designed around shorter persistence, although the wording across Help Center pages is not perfectly uniform.
| Search Mode | Saved to Normal Account History? | Documented Availability or Expiry | Recovery Expectation |
| Signed in, normal mode | Yes | Indefinite until deleted | Recoverable from History while retained |
| Logged out | No long-term account history | 14 days | Not recoverable after expiry |
| Incognito | Not saved as normal long-term history | Expires within 24 hours | Not recoverable after expiry |
| Comet temporary thread using page, tab, or history context | Marked temporary | Up to 30 days for relevant Assistant context | Can be deleted earlier; expiry can apply |
| Enterprise session | Depends on organisation policy | Admin-defined where supported | Controlled by organisation retention settings |
Perplexity’s current session page says anonymous, logged-out sessions remain visible for 14 days and then disappear permanently. Its Incognito troubleshooting page says Incognito sessions expire within 24 hours and are not recoverable. Another Help Center page says Incognito searches “are not stored” and cannot be recovered after the session closes. The safest interpretation is that Incognito is not saved to your normal long-term history, but Perplexity’s own documentation still gives 24 hours as the short-lived expiry window for Incognito sessions. Users should not read “Incognito” as a guarantee of zero server-side processing or instant erasure unless Perplexity documents that more explicitly.
That distinction is an important gap in many ranking pages. Browser-style words such as Incognito create an intuitive expectation that nothing exists after the window closes. Perplexity’s product documentation instead describes an expiring session. For ordinary low-risk searches, that difference may not matter. For highly sensitive information, it should change your behaviour.
I would use Incognito when I want a query excluded from long-term account history, not when I need a contractual guarantee of zero retention. If the latter is the requirement, consumer Incognito is the wrong standard to apply. Enterprise controls or the Sonar API’s Zero Data Retention policy are the relevant documented products.
For a broader product walkthrough, our complete guide to using Perplexity AI explains how History, Projects, search modes, and account features fit together.
Turning Off AI Data Retention Does Not Delete Your History
This is the most common privacy misunderstanding in the current search results. Perplexity’s AI Data Retention or AI Data Usage control is about training and improvement, not about whether your saved sessions remain visible in History.
Perplexity says AI Data Retention is enabled by default for Free, Pro, and Max users. If you switch it off, the opt-out applies to data collected after the opt-out date. The company also states that previously collected training data cannot be deleted or removed from training datasets through that setting. Core product functions continue to work, and data may still be processed for service operation, legal compliance, and product improvement.
That produces a simple but important rule: storage and training are two separate decisions. You can keep a saved conversation in your History while preventing future eligible searches from being used for AI training. You can also delete a conversation from History without assuming that a prior training contribution has been reversed.
The Verge made the same conceptual distinction in its privacy controls guide: disabling AI training is not the same as wiping chatbot history. That observation remains useful because the product labels can make the settings feel more overlapping than they are.
For privacy-conscious users, I would change the training setting before entering sensitive material, not afterwards. The Help Center is explicit that the opt-out is prospective. If you only disable it after months of use, you have changed the policy for future collection, not rewritten the history of how earlier eligible data may have been processed.
This is also one area where user control affects trust. A 2026 study by Min-Ji Kwon and Seung-In Kim involving 110 participants found that chatbot memory features increased both trust and privacy concern, while stronger user control reduced privacy concern and increased trust. The study was not about Perplexity specifically, but its result explains why precise controls matter more than a generic “we value privacy” statement.
Our wider analysis of AI privacy concerns in 2026 examines the same distinction across memory, inference, retention, and agent permissions.
Deleting a Thread, Deleting All Threads, and Deleting an Account Are Not Equivalent
Perplexity gives users several deletion actions, and they operate at different scopes. The safest way to choose among them is to start with the outcome you want rather than the button you can see.
| Action | What It Is Designed to Remove | What It Does Not Necessarily Change |
| Delete one session | A specific saved thread from your History | Your AI training preference, other sessions, or account data |
| Delete all sessions | All saved threads visible in History | Your account, profile, preferences, or prior training use |
| Turn off AI Data Retention | Future eligible consumer data use for AI training | Existing saved History or previously collected training data |
| Use Incognito | Long-term saving of that query to normal History | All transient processing or third-party website tracking |
| Delete account | Profile, preferences, sessions, and associated personal data under the account deletion process | Legal exceptions or processing that Perplexity is permitted or required to retain |
Perplexity’s self-serve deletion guide lets users delete individual sessions or all threads without deleting the account. Its account deletion documentation says search history, sessions, and shared content are deleted as part of account deletion, with the account and personal data permanently deleted within 30 days. The retention Help Center repeats that personal information is removed from Perplexity’s servers within 30 days after account deletion.
The 30-day account deletion window is the strongest published consumer timeline I found for server-side deletion. It is more meaningful than simply watching a thread disappear from the interface. A thread deletion can be immediate from the user’s view, but the public documentation does not provide the same precise server-side deletion timetable for every individual thread deletion that it provides for full account deletion.
That is an area where I would avoid false certainty. A user can confirm that a deleted thread no longer appears in History. The public Help Center does not give a universal statement that every copy, log, backup, or safety record tied to that individual deletion is erased from every system at the exact same moment. If you require complete account-level erasure, follow the account deletion process and the privacy-rights route available in your jurisdiction.
The Hidden Distinction Between Visible History and Backend Logs
A privacy article should separate what a user can see from what a service may hold for operational, security, legal, or system purposes. Perplexity’s Privacy Notice says it keeps personal data only as long as necessary for the purposes described in the notice, subject to longer periods required or permitted by law. It considers the amount, nature, sensitivity, risk, purpose, and legal requirements when determining retention. That is a standard purpose-based formulation, not a single number that applies to every internal system.
Public 2026 litigation records add an unusual piece of context. In separate copyright proceedings, federal court filings refer to Perplexity user activity log data and Perplexity-controlled data repositories. One September 2026 order required Perplexity to host six months of specified user activity log data for inspection in litigation. Another April order discussed production of additional months of log database snapshots. These documents establish that user activity logging exists in a litigation context, but they do not establish the ordinary retention period for a particular consumer’s search history.
That boundary matters. It would be irresponsible to turn a discovery dispute into the claim that “Perplexity secretly keeps every deleted search forever.” The records do not prove that. They do show why visible History and backend operational data should not be treated as the same category without evidence.
For users, the practical takeaway is modest. If your standard is “I do not want this query visible in my account,” session deletion is the right control. If your standard is “I need formal erasure of account-linked personal data,” use account deletion or a data-rights request. If your standard is “this information should never enter a consumer AI service in the first place,” the safest control is data minimisation before submission.
This distinction also explains why our best AI search engine comparison treats privacy architecture as a workflow criterion rather than a single score.
Comet Browser History Is a Separate Privacy Layer
If you use Perplexity’s Comet browser, “search history” can refer to more than Perplexity sessions. It can also include browser history, local browser data, open-tab context, page content, and information an Assistant task needs to complete an action.
Perplexity’s March 2026 Comet Assistant privacy documentation says the Assistant does not access or upload your full browsing history, full tab list, cookies, site data, passwords, autofill data, local files, or text typed into websites by default. Data is sent when a request requires it, such as summarising the current tab, extracting selected text, using another tab, or completing an email or calendar task. When a request uses page content, history items, or open-tab context, Perplexity says that context can be stored securely for up to 30 days to support Library and query-history features.
The broader July 2026 Privacy Notice adds another nuance. Depending on your settings, Perplexity says it may record browsing history and usage when Comet is used with a synced Perplexity account. It also says some browsing data, permissions, and download records are stored locally. In other words, “local by default” does not mean “Perplexity never receives browser context.” The correct question is whether a specific Assistant request needs that context and whether your sync or privacy settings cause additional data to be recorded.
Security researchers also warn that agentic browsers expand the attack surface. David Kohlbrenner of the University of Washington said in June 2026, “Browser agents aren’t ready for the public.” His team found conditions under which highly privileged browser agents, including Perplexity Comet, could face cross-origin and prompt-injection risks. Co-senior author Franziska Roesner added, “This policy is fundamental to how modern browsers protect your information,” referring to the web’s same-origin security model. These findings concern security, not Perplexity’s retention policy, but they matter because privacy depends on both policy and what an authorised agent can access.
For hands-on controls, our guide to browsing with Perplexity Comet explains the permission boundaries, while our Comet browser features guide maps what stays local and what can be sent for Assistant tasks.
Free, Pro, Max, Enterprise, and API Do Not Have the Same Privacy Model
Paying for a consumer plan does not automatically move you into Perplexity’s strongest documented privacy regime. Free, Pro, and Max are still consumer products for the purpose of AI Data Retention defaults. Enterprise and API products have materially different commitments.
| Product | Current Published Price | Training Position | Retention Notes Most Relevant to This Question |
| Standard | Free | Consumer AI Data Retention enabled by default | Signed-in History available; user can opt out of future training use |
| Perplexity Pro | $20/month or $200/year | Same consumer training control, opt-out available | More product capability does not itself create Enterprise no-training terms |
| Perplexity Max | $200/month or $2,000/year | Same consumer training control, opt-out available | Consumer account rules still apply unless stated otherwise |
| Enterprise Pro | $40/month or $400/year per seat | Enterprise data is not used for AI training | Uploaded files retained 7 days; eligible organisations can set session retention |
| Enterprise Max | $325/month or $3,250/year per seat | Enterprise data is not used for AI training | Custom retention available with Enterprise security controls |
| Sonar API | Usage-based API pricing | Customer prompt/response content is not used for training | Perplexity documents Zero Data Retention for API content, while keeping billing metadata |
Perplexity’s September 2026 plan guidance lists Standard, Pro, Max, Enterprise Pro, Enterprise Max, and the API platform. Official Help Center pages put Pro at $20 per month or $200 per year, Max at $200 per month or $2,000 per year, Enterprise Pro at $40 per month or $400 per year per seat, and Enterprise Max at $325 per month or $3,250 per year per seat.
The privacy difference is more important than the price. Perplexity says Enterprise data is never used for AI training. Enterprise users also receive additional retention controls, including a seven-day retention period for uploaded files and custom session retention for qualifying organisations. Its API privacy documentation states a Zero Data Retention policy for Sonar API prompt and response content and says the retained billing metadata does not include prompt or response content.
This creates a clear decision boundary. If you are a consumer who merely wants better models or higher limits, Pro or Max may fit. If you are choosing a product because contractual no-training and retention controls are the requirement, Enterprise or the API is the relevant comparison.
Third-Party Models Are Not the Same as Third-Party Service Providers
Perplexity can route queries through models from companies such as OpenAI and Anthropic, which naturally raises another version of the history question: does a third-party model provider keep or train on what I send through Perplexity?
Perplexity’s Help Center says no. It states that its agreements with third-party model providers such as OpenAI and Anthropic prohibit those providers from using Perplexity data to train their models, and says data processed through those providers is not retained or used for external model training.
That commitment should not be confused with a claim that no third party ever processes any personal data. Perplexity’s privacy documentation separately states that it may share information with service providers that perform functions on its behalf, such as payment processing, customer support, and communications. The July 2026 Privacy Notice also clarifies that Perplexity does not sell personal data or send queries, prompts, or conversation content to advertisers.
The practical distinction is contractual role. A model provider processing a query under a no-training agreement is different from an advertiser receiving conversation content, and both are different from an operational service provider processing data for a defined business function. Privacy analysis becomes more accurate when those roles are kept separate.
Ninghui Li, Kaiyuan Zhang, Kyle Polley, and Jerry Ma made a similar architectural point in a 2026 paper based on Perplexity’s response to a NIST request for information: “Agent architectures change core assumptions around code-data separation, authority boundaries, and execution predictability.” The paper addresses agent security rather than consumer retention, but the quote captures why data flow matters as much as the brand name of the model that ultimately generates an answer.
For users who want to understand how those models sit behind Perplexity’s product experience, our guide to using Perplexity Comet shows when the platform pulls in model, browser, and connected-app context.
A Privacy Workflow for Sensitive Perplexity Searches
There is no setting that turns a cloud AI service into a sealed notebook. The safest workflow starts before you submit a query. I use a four-question test: does the service need the real name, the exact document, the exact number, or the identifying context to answer the question? If not, remove it.
For a sensitive but legitimate consumer query, the strongest practical workflow is:
- Minimise the prompt. Replace names, account numbers, client identifiers, addresses, case references, and unnecessary exact figures with neutral placeholders.
- Turn off AI Data Retention before the query if you do not want future eligible search data used for Perplexity model improvement.
- Use Incognito if you do not need the query saved to long-term account History. Remember that Perplexity separately documents a 24-hour Incognito expiry window.
- Avoid attaching source files unless they are necessary. Uploaded documents can contain far more personal data than the question itself.
- If you use Comet, check whether the task requires page, tab, email, calendar, or history context. Authorise only the minimum scope necessary.
- Delete the session afterwards if you do not need it. If your concern is broader account-level erasure, use the documented account deletion process instead.
- For regulated or organisational data, do not assume a consumer plan is appropriate. Use Enterprise controls, an approved API workflow, or your organisation’s sanctioned system.
This workflow follows the documented differences between training opt-out, History deletion, Incognito, Comet context, and account deletion. It also matches a broader security principle emerging in 2026: give agents less authority and less data than they could technically access. Google security researchers described indirect prompt injection as a primary attack vector for AI agents, while University of Washington researchers found that browsers with fewer agent permissions were generally safer.
The goal is not to make ordinary search feel dangerous. It is to keep the sensitivity of the data proportional to the guarantees of the product you are using.
What Perplexity’s Documentation Still Does Not Fully Resolve
The public documentation is considerably clearer in 2026 than many older privacy summaries suggest, but several questions remain open enough that I would not fill them with assumptions.
First, Perplexity gives a precise “indefinite until deleted” statement for signed-in session History and a precise 30-day server deletion window after full account deletion. It does not publish one universal server-side timetable for every individual thread deletion across logs, backups, fraud systems, legal holds, and operational systems. A user can verify removal from the visible interface, but not every internal copy.
Second, Incognito wording varies. One Help Center page says Incognito searches are not stored, while another says Incognito sessions expire within 24 hours. Those statements can be reconciled if “not stored” means not saved as normal long-term history, but Perplexity could make the server-side lifecycle more explicit.
Third, the Privacy Notice uses purpose-based retention language for personal data, while the product Help Center gives more specific feature-level timelines. That is normal in privacy documentation, but it means readers should not treat one number as the answer for every data category.
Fourth, consumer training opt-out is prospective. Perplexity says previously collected training data cannot be removed through the AI Data Retention setting. The documentation does not provide a consumer-facing mechanism that traces an old query into a specific training dataset and removes that contribution individually.
Fifth, Comet creates a privacy surface that changes with task context. Ordinary browsing can remain local, but a user-requested Assistant task may send page, tab, history, email, or calendar context. The correct privacy question therefore changes from “Does Comet see my browser?” to “What context did this specific task require, and what retention rule applies to that context?”
These are not reasons to invent a darker conclusion. They are reasons to state the boundary of public evidence. The official material supports a strong answer about visible History and several specific retention windows. It does not support the claim that every internal data copy follows one identical timer.
A Risk-Based Decision Matrix for Everyday Use
The easiest way to turn the policy into behaviour is to match the data to the risk. Most people do not need maximum privacy for every search, but they do benefit from recognising when a query has crossed into a more sensitive category.
| Query Type | Example | Reasonable Perplexity Setup | Better Alternative When Needed |
| Low sensitivity | Public facts, recipes, travel ideas, product research | Normal signed-in use if History is useful | Delete old sessions periodically for housekeeping |
| Moderate sensitivity | Career planning, personal budgeting, private research notes | AI training opt-out plus selective deletion | Incognito if long-term History adds no value |
| High personal sensitivity | Medical symptoms, legal disputes, family finances | Minimise identifiers, use Incognito, avoid unnecessary uploads | Use a professional service or privacy-specific workflow for regulated data |
| Confidential work data | Client documents, internal strategy, contracts, source code | Do not rely on a personal consumer account | Enterprise account with approved retention and access controls |
| Programmatic sensitive workload | Internal application queries | Avoid consumer UI for production workflows | Sonar API where Zero Data Retention requirements fit the use case |
This matrix is intentionally conservative. A privacy setting can reduce one risk without changing another. Incognito reduces long-term account history, but it does not make the websites you visit in Comet unable to track you. Training opt-out changes model-improvement use, but it does not erase saved History. Account deletion is broader, but it is a destructive action that removes access to the account.
Tom’s Guide ranked Perplexity relatively well on privacy controls in a 2025 comparison, largely because the training opt-out and Incognito controls are accessible. That is useful context, but a ranking cannot replace a workflow decision. A service can have better controls than a competitor and still be the wrong place for a particular piece of regulated or confidential information.
The best privacy habit is not “always use Incognito” or “never use AI”. It is to know which data layer you are changing and choose the smallest amount of retained context that still makes the tool useful.
Our Editorial Verification Process
I treated this as an explainer and privacy-verification article rather than a product review. Before drafting, I reviewed ten high-ranking or strongly surfaced live results for the target question and adjacent search-history queries. The set included Perplexity’s retention, data collection, Privacy Notice, API privacy, Enterprise retention, Sessions, and deletion documentation, plus comparative privacy coverage from Tom’s Guide and The Verge. The recurring structure in those pages was feature-by-feature: one page explains retention, another training, another deletion, another plan-level privacy. I deliberately did not copy that sequence. The article instead follows a data lifecycle from query creation to History, training, temporary retention, browser context, deletion, and account erasure.
I attempted the Perplexity AI Magazine sitemap endpoints specified in the editorial brief, including sitemap.xml, sitemap_index.xml, and post-sitemap.xml. They did not return parseable XML through the available browsing layer in this session, so I did not invent a sitemap inventory. Internal links were selected from live indexed Perplexity AI Magazine pages with direct semantic relevance to search history, deletion, privacy, Comet, and AI search. This follows the fallback rule in the supplied brief, which explicitly prioritises relevant indexed pages over fabricated sitemap links.
Product claims were cross-checked against Perplexity documentation updated between July and September 2026. The key systems verified were normal signed-in sessions, logged-out sessions, Incognito, AI Data Retention, account deletion, Comet Assistant context, Enterprise retention, third-party model agreements, and the Sonar API Zero Data Retention policy. Where Perplexity’s documentation used different wording for Incognito, I preserved the uncertainty rather than collapsing it into a stronger claim.
The article also uses 2026 research and expert commentary to add context without substituting secondary material for Perplexity’s own policies. That includes Kwon and Kim’s study of memory, privacy concern, and user control; Li, Zhang, Polley, and Ma’s paper on AI agent security; and University of Washington research on agentic browser security.
This article was researched and drafted with AI assistance and reviewed by the Sami Ullah Khan editorial desk at Perplexity AI Magazine. All data, citations, pricing figures, and named quotes have been independently verified against primary sources before publication.
Because this is a pre-publication deliverable, WordPress-specific back-button behaviour and hidden-content checks cannot be validated yet. After publishing, test the browser back button from a referring page, inspect the rendered DOM for hidden text patterns, and audit any custom history-manipulation snippets before treating technical compliance as complete.
Conclusion
Perplexity does store search history for signed-in users, and its current documentation says normal signed-in sessions can remain in History indefinitely until you delete them. That answer is simple. The privacy reality around it is not.
A saved session, a training setting, an Incognito query, a Comet browser task, an Enterprise retention policy, and an API request all follow different rules. Consumer AI Data Retention is enabled by default for Free, Pro, and Max, but users can opt out for future data. Logged-out sessions expire after 14 days. Incognito sessions expire within 24 hours. Comet Assistant context can be retained for up to 30 days when a task uses page, tab, or history information. Full account deletion carries the clearest consumer server-side deadline, with Perplexity saying personal data is permanently deleted within 30 days. Enterprise and API offerings add stronger no-training or zero-retention commitments.
The open question is not whether Perplexity keeps any history. It does. The useful question is which data layer you want to control. Once that distinction is clear, privacy settings become much easier to use intelligently and much harder to misunderstand.
Frequently Asked Questions
Does Perplexity AI Store My Search History?
Yes. Perplexity stores normal signed-in sessions in your account History, and its Help Center says those sessions can remain there indefinitely until you delete them. Logged-out sessions expire after 14 days, while Incognito sessions expire within 24 hours. Account deletion triggers a broader process under which Perplexity says personal data is permanently deleted within 30 days.
Does Turning Off AI Data Retention Delete My Perplexity History?
No. Turning off AI Data Retention changes whether future eligible consumer search data can be used to improve Perplexity’s models. It does not automatically delete sessions already saved in your History. Perplexity also says the opt-out applies prospectively and previously collected training data cannot be removed through that setting.
How Long Does Perplexity Keep Logged-Out Searches?
Perplexity says sessions created while you are not signed in are anonymous and remain available for 14 days. After that period, they disappear and are not recoverable. If you need long-term access to a session, sign in and save it to your account.
How Long Does Perplexity Keep Incognito Searches?
Perplexity’s Incognito troubleshooting documentation says Incognito sessions expire within 24 hours and cannot be recovered. Other Help Center wording says Incognito searches are not stored as normal history. The safest reading is that Incognito avoids long-term account history but still has a short documented expiry lifecycle.
Does Perplexity Use My Searches to Train AI Models?
For Free, Pro, and Max consumer accounts, Perplexity says AI Data Retention is enabled by default and eligible data may be used to improve its AI models unless you opt out. Enterprise data is not used for AI training. Third-party model providers are contractually barred from training on Perplexity data, according to Perplexity’s Help Center.
Can I Delete All My Perplexity Searches Without Deleting My Account?
Yes. Perplexity’s self-serve deletion controls allow you to delete all threads from History without deleting the account. You can also delete individual sessions. Deleting the entire account is a separate action that removes profile information, preferences, sessions, and other account-linked personal data under Perplexity’s deletion process.
Does Perplexity Comet Store My Browsing History?
It can, depending on settings and the task. Perplexity says ordinary browser data stays local by default in several categories, but a synced Comet account may record browsing history depending on settings. When Assistant tasks require page, tab, or history context, that context may be stored for up to 30 days.
Is Perplexity Enterprise More Private Than Pro for Search History?
It has stronger documented data controls. Perplexity says Enterprise data is never used to train its AI models, uploaded files have limited retention, and qualifying organisations can set custom session retention periods. Pro and Max remain consumer plans with AI Data Retention enabled by default unless the user opts out.
References
Kohlbrenner, D., Roesner, F., et al. (2026, June 30). Some agentic AI browsers come with major cybersecurity risks. University of Washington News.
Kwon, M.-J., & Kim, S.-I. (2026). The effect of AI chatbot memory features on privacy concerns and trust: The roles of social presence and user control. Industry Promotion Research, 11(3), 263-270.
Li, N., Zhang, K., Polley, K., & Ma, J. (2026). Security considerations for artificial intelligence agents. arXiv:2603.12230.
Perplexity AI. (2026, July 8). Perplexity Privacy Notice.
Perplexity Support. (2026, July 16). Data Collection at Perplexity.
Perplexity Support. (2026, July 16). How long does Perplexity retain my search history, profile data, and personal information?
Perplexity Support. (2026, July 28). Self-Serve Data Deletion.
Perplexity Support. (2026, August 28). What is a Session?
Perplexity. (2026, March 4). Comet Assistant Privacy & Data Use.
