What is Perplexity good for?

Perplexity is one of the AI tools taught in our class, where we use it with Claude to build a research brief on a provided scenario and check its sources. This page explains the kinds of work it suits. Plan prices and feature lists change often, so check Perplexity’s site for current details.

What kind of tool it is

Perplexity is an AI search tool: you ask a question in ordinary language, and it searches the web and writes an answer rather than returning a list of links. The value for work is speed to a first overview. The risk is the same as with any AI writing: a fluent summary can still be wrong.

What Perplexity’s documentation and independent tests say

• An arXiv audit of generative search engines including perplexity.ai found that on average only 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence
• The Columbia Journalism Review Tow Center test of AI search tools found that Perplexity answered 37 percent of the queries incorrectly. The Columbia Journalism Review Tow Center test found that the chatbots collectively provided incorrect answers to more than 60 percent of queries. That is why the verification routine below matters.

Work tasks that suit it

Competitor and market scans. “What products does this company offer, who are its main customers, and what has it announced this year?” gives you a starting brief in minutes.

Vendor and tool comparisons. Ask for a comparison table of features and pricing from public pages, then check each price on the vendor’s own site.

Getting up to speed. Ask for an explanation of an unfamiliar regulation, technology or industry term, with sources, before a meeting.

Recurring research. In our class, attendees practice building a brief and then checking every claim and citation.

Tasks where it is weaker

• Anything about your private documents or internal data, which it cannot see unless you upload it.
• Niche facts with few good sources, where it may over-generalize.
• Questions where a single authoritative source exists (a statute, a filing, a price list): go to that source directly.

A verification routine

1. Ask for sources with each claim.
2. Open at least the sources behind the numbers, dates and names.
3. Check whether the source actually says what the summary claims.
4. Note the date of the source; old pages can dominate results.

For more on why answers go wrong, see how to stop AI hallucinations.

Perplexity, Google or a chat assistant?

• Google: best for locating a specific page, official site or local result.
• Perplexity: best for a sourced synthesis of what is online.
• Claude or ChatGPT: best for analyzing documents you provide and for writing.

Many people use two of these together: Perplexity to gather material, then a chat assistant to analyze and write. Our Perplexity vs Google comparison goes deeper.

Data caution

Do not paste confidential information into any search tool without checking your company’s rules. See whether it is safe to put company data in ChatGPT for the same questions applied to chat tools.

Sources

• arXiv, Evaluating Verifiability in Generative Search Engines (Liu et al.): “on average, a mere 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence”
• Columbia Journalism Review, Tow Center, AI search engines are bad at citing news: “with Perplexity answering 37 percent of the queries incorrectly”
• Columbia Journalism Review, Tow Center, AI search engines are bad at citing news: “Collectively, they provided incorrect answers to more than 60 percent of queries”