Perplexity

Editorial assessment

The Free plan’s limits decide most choices. The pricing page only describes them as limited usage, but the comparison table in the Help Center plan guide sets 3 Pro Searches a day and 1 Research query a month, so anyone who needs a full research report more than once a month, or several deep searches a day, will run into the Free limits quickly.

• FreeFree

  occasional look-ups where basic search with citations is enough.

• ProUSD 17 per month

  weekly or daily research, model choice, image generation and scheduled Tasks.

• Education ProUSD 10 per month

  the same Pro features at half the Pro price for verified students and faculty.

• MaxUSD 167 per month

  heavy daily use of Research and Computer, where Pro’s weekly limits are the bottleneck.

• Enterprise ProUSD 34 per seat per month

  teams that need seat management, admin-controlled billing and data that is never used for training; the current Help Center comparison lists 400 Pro Searches a week and 50 Research queries a month, while an older copy of the same article still shows 80.

• Enterprise MaxUSD 271 per seat per month

  organizations that need 4,000 Pro Searches a week and 500 Research queries a month (800 in the older copy).

• Enterprise custom pricing

  organizations that need role-based access control, their own enterprise domain or volume discounts.

Pro and Max limits are described as weekly and monthly allowances for average or advanced use rather than fixed numbers, so a heavy user should watch the usage shown in the account before upgrading.

You ask a question in plain language and Perplexity assembles an answer from web sources, with links to the pages it drew on. Pro Search runs a deeper, multi-step search for the same purpose. The links make each sentence traceable to a page you can open, which is the product’s main difference from a chat assistant used without search.

A quick way to use them: pick the claims your decision rests on, open the linked page for each, find the passage and check the date on it. Where the answer is more confident than its source, correct it in your notes. Perplexity works best as a map of where to read next, and the research category lists tools built for papers and notebooks instead of the open web.

If the work is producing drafts rather than finding sources, a writing-first assistant such as Claude may fit better; Claude vs Perplexity and the comparison table set them side by side. Our editorial policy explains how profiles in the directory are checked and corrected.

• Research mode runs dozens of searches, reads hundreds of sources and writes a report you can export to PDF or a document. Free users get limited access, and you cannot pick the model it uses.
• Spaces group threads and files under custom AI instructions and let you invite viewers or contributors. Pro lists up to 50 file uploads per Space; syncing files from connected apps is described for Enterprise users.
• Tasks run a saved question once, daily, weekly, every weekday, monthly or yearly, and deliver the result by email or push notification. Pro and Max subscribers can keep up to ten.

You can attach text, code, PDF, image, audio and video files, and speech in audio or video is transcribed so it can be searched. Each file is limited to 50 MB, long files are cut down to their most relevant parts, and visual scenes in video are not indexed. For a long report, ask about specific sections rather than expecting the whole document to be read.

No. The developer API is a separate product billed per token for building software, and its rates tell you nothing about what a Pro, Max or Enterprise seat costs. If you only need answers in the browser or app, the subscription plans above are the relevant prices.

Input: Ask: Which file formats does the Library of Congress recommend for long-term still-image preservation? Limit the answer to the current Recommended Formats Statement and request three linked claims.

1. Open each cited Library of Congress page in a separate tab.
2. Record the exact supporting passage, section title and date for three claims.
3. Mark whether each statement is a recommendation, an accepted format or your own inference.

Accept: Each selected claim has an accessible official passage with the same scope. Stop or revise: Treat a missing page, obsolete edition or stronger wording than the source as a failed claim; narrow the question and check again.