Perplexity: our review
Perplexity is a search engine that writes a direct answer to a question and pairs it with the pages that fed it. It suits anyone who wants to save the time of opening ten tabs, provided they keep an eye on what the cited sources actually say.
What we take away from it
What counts in its favor
- Every answer comes with clickable sources: you can see where the information comes from, unlike an assistant that answers with no reference.
- A free version already usable with no account for a one-off search, with no announced limit on basic searches.
- An Enterprise account contractually guarantees that no data is used to train a model, with zero-retention agreements at third-party model providers.
What counts against it
- On consumer accounts (Free, Pro, Max), collection for model training is enabled by default: you have to turn it off yourself in the settings, and this choice does not apply to requests already sent.
- No pricing page could be reviewed during the check (a systematic 403 across the whole domain): it's impossible here to confirm the exact price of the Pro or Max plan, only some secondary tiers (Education at 10 $/month, Enterprise Pro from 40 $/month per seat) are documented in the help center.
- Data passes through servers located in the United States or outside the European Economic Area, with no announced European hosting option for consumer accounts.
Useful for a first pass on a search and coming away with sourced leads, provided you reopen the cited pages yourself before quoting the result to a client. To rule out if the requirement is strictly European hosting or a firm price confirmed in advance.
The facts
- Data hosting
- Data outside the EU · source
- Categories
- Research and monitoring
- Data verified on
- 30 July 2026
What happens to your data
- Is your data used to train their models?
- Depends on the plan, or an option you can turn off
- Subprocessors
- Requests that call an advanced model go through third-party providers (OpenAI, Anthropic, and others cited in the help center). On Enterprise accounts, zero-retention and zero-training agreements govern these sub-processors; on consumer accounts, no such agreement is publicly documented.
- Retention
- The "AI Data Retention" setting is enabled by default on Free, Pro and Max accounts. Turning it off stops future collection, but what has already been sent is not deleted retroactively. On Enterprise, uploaded files are kept for 7 days and requests never feed training.
Perplexity in detail
Perplexity doesn't present itself as a conversational assistant but as a search engine: you ask a question, it reads a batch of web pages and writes a constructed answer, with every claim linked to the page that produced it. It's the clickable citation feature that sets the tool apart from the rest of the chatbots.
What it actually does
Asking a question triggers a real web search, not an answer drawn solely from a model's memory. Perplexity picks a batch of pages, extracts what answers the question, and writes a text that links back to each of them with a small number. The most common use among a consultant or an executive is quick monitoring: a regulatory point, a competitor's news, the state of a technology. Deep Research mode goes further by chaining several queries before producing a longer report.
Why it interests us, with a reservation
Perplexity's commercial pitch, systematic citation, is real and useful: you can trace back to the source in one click, something no purely conversational assistant offers by default. But a sourced answer isn't a verified answer. The citation only guarantees that the sentence comes from somewhere, not that the page actually says what the summary attributes to it, nor that it's authoritative on the subject. On a question of French law or precise tax rules, where few online pages are genuinely reliable, the answer is only as good as the handful of sources found that day. It's a nuance the tool's marketing presentation, centred on source transparency, doesn't encourage keeping in mind. We address it in our audit and consulting engagements: citing a source doesn't excuse anyone from reading it.
Where it falls short
Two further reservations, beyond the question of the sources themselves.
The first concerns personal data: collection for model training is switched on by default on Free, Pro and Max accounts. You have to turn it off yourself in the account settings, and that decision only applies to future queries — what's already been sent stays with Perplexity. Only the Enterprise tier excludes this collection by contract.
The second concerns price: the official pricing page refused all automated access during this check, across the entire domain. We were therefore unable to confirm the exact rate of the consumer Pro and Max offerings — only the secondary tiers (Education, Enterprise) are documented in the help centre. Check the price shown before committing.
Our recommendation
Perplexity earns its place in daily monitoring, provided every answer is treated as a starting point and not a conclusion: reopen the cited sources before passing them on to a client. For an organization that requires strictly European hosting of data, or that refuses any default collection, this isn't the right choice without going through the Enterprise offering.
Who it's for, and who it isn't
Suited for
- Consultant or executive who does regular monitoring and wants to save the time of cross-checking several pages by hand
- Team that accepts checking the cited sources itself before passing them on in a client deliverable
- Organization that has the means for the Enterprise tier and wants the contractual guarantee that its requests feed no model
Not a fit
- Firm bound by a strictly European data-hosting clause: Perplexity does not offer it on its consumer plans
- Anyone who takes a sourced answer for a verified answer: a cited source can be misread, poorly summarized, or simply wrong
- Organization that wants a firm price without first checking the help center: the public pricing page could not be verified as of today
Frequently asked questions
- Is a Perplexity answer with sources reliable as is?
- No, not automatically. Perplexity shows the pages that fed its answer, which allows checking — but a cited source can be poorly summarized, outdated, or unreliable. The citation proves where the information comes from, not that it is accurate: open the source before passing it on in a professional document.
- Do my searches train Perplexity's models?
- By default, yes, on Free, Pro and Max accounts: the "AI Data Retention" setting is enabled when the account is created. It can be turned off in the account preferences, but only for future requests. On an Enterprise account, this collection does not happen, contractually.
- How much does Perplexity cost?
- The vendor's pricing page could not be reviewed during this check (it refuses all automated access). The help center does document the Education plan at 10 $/month and the Enterprise Pro plan from 40 $/month per seat (30 $ for education and non-profit institutions). Check the price of the Pro and Max plans directly on the site before committing.
- Does Perplexity cover the French-speaking web well?
- The interface exists in French and the tool indexes French-language pages like the rest of the web. But the quality of an answer depends on what actually exists online on a given topic: on a precise question of French law or taxation, where few pages carry authority, the answer is only as good as the few sources found.
- What's the difference with ChatGPT or Claude for a search?
- Perplexity builds its answer around citation: web search and clickable sources are at the center of the product. ChatGPT and Claude offer web search as an option, but their primary use remains conversation and writing. For monitoring that must rely on identifiable pages, Perplexity starts with a structural advantage.
Alternatives to Perplexity
AI assistants · Writing
ChatGPT
OpenAI's assistant — the one your teams have often installed before telling you about it.
AI assistants · Writing
Claude
Anthropic's assistant, at ease with long documents and structured writing.
Research and monitoring
Kagi
Paid search engine with no ads or tracking, with a built-in AI assistant to summarize results.
Research and monitoring
Exa
Semantic search engine designed to be plugged into AI agents rather than used directly by a human.
Our sources
- Perplexity Privacy Notice, read via web archiveaccessed on 30 July 2026
- Data Collection at Perplexity (help center), read via web archiveaccessed on 30 July 2026
- Which Perplexity Subscription Plan is right for you? (help center), read via web archiveaccessed on 30 July 2026
- What is Perplexity Pro? (help center), read via web archiveaccessed on 30 July 2026
Prices and hosting terms change often. We recheck them periodically, but the publisher's page remains the only up-to-date reference.