What is Perplexity and what is it good for at work?
Perplexity is an answer engine: you ask a question in everyday language, it searches articles, websites and journals, and it writes a single answer with numbered citations linked to the original sources. Its help centre describes Basic search for quick facts, Pro Search for deeper multi-source answers with a choice of AI models, and Research mode, which runs many searches, reads many sources and produces a comprehensive report, typically in a few minutes.
At work it fits the early stage of research: getting oriented on an unfamiliar market, finding what has been published on a regulation, collecting sources for a briefing or checking what a competitor says about itself. Its value is the citation trail. Its risk is treating a confident summary as the finding rather than as a list of pages to read.
How do I get started with Perplexity?
Go to perplexity.ai and type a question. Perplexity's getting-started article says each question starts a session, and follow-up questions stay in that session so you can refine without starting over. You can upload a file to a session and ask about it alongside web results, and adjust language, output format and tone in your profile.
If your company uses Perplexity Enterprise, sign in with that account, because Perplexity describes different data handling for Enterprise.
Your first question should be specific enough to have checkable answers: 'What are the main UK requirements for food allergen labelling on prepacked products, with official sources?' Then open the first three citations and see whether they say what the answer claims.
What is a good 10-minute Perplexity starter exercise?
Use a fictional eight-person consultancy, Northgate Advisory, preparing for a pitch to a regional garden centre chain that wants to start selling online.
Minute 1 to 3: ask a scoped question in Pro Search: what do established UK garden centre chains publicly say about their online ordering and delivery options, citing their own websites. Minute 4 to 5: follow up in the same session to turn the answer into a table with company, what they offer and the citation number. Minute 6 to 8: open every citation in the table. Mark each row as confirmed, partly supported or not supported by the page. Minute 9 to 10: ask Perplexity what kinds of sources it did not use, such as trade bodies or official statistics, and note where you would look next.
The deliverable is not the table Perplexity wrote; it is the version with your confirmation column, which is the only one fit for a client pitch.
How do I get better results from Perplexity?
Perplexity is a search tool first, so better results come from better questions and stricter checking. Perplexity's own tips are to start with a clear goal, use straightforward language, avoid asking too much at once and include important context.
Name the source type you want
Ask for official, company or academic sources explicitly, then check the trust labels and domains in the citations.
Choose the mode by depth
Use Basic for a fact, Pro Search for a multi-source answer and Research when you need a report and can wait a few minutes.
Refine within the session
Follow-ups keep context, so narrow the question step by step instead of starting new searches.
Group work in a Project
Perplexity's Projects organise sessions and files by topic, with instructions applied to every query in the Project, and they are private by default.
Verify by highlighting
Perplexity suggests highlighting answer text to check the source behind it; do this for every claim you will repeat.
What should I not paste into Perplexity, and what does Perplexity say about data?
Search questions often leak more than people expect: a query naming a client, a deal or an unannounced product can reveal confidential plans. Follow your employer's AI policy, keep client names out of personal accounts and avoid uploading confidential files.
Perplexity's help centre says that for Free, Pro and Max users, AI Data Retention is on by default, which means some data may be used to train AI models and improve search quality. You can switch it off under Account settings, Preferences, with the AI data retention toggle. Perplexity says opting out applies only to data collected after you opt out, and previously collected training data cannot be deleted. It states that Enterprise data is never used for AI training. Turn the setting off before your first work search if your policy allows a personal account at all.
Where does Perplexity fall short?
An answer with citations is not the same as a correct answer. Summaries can blend sources, overstate what a page says or lean on whatever ranks well rather than what is authoritative. That is why the starter exercise ends with a confirmation column.
Perplexity is also less suited to work that starts from your own material rather than the web, such as drafting a long document or analysing an internal spreadsheet. And because the training opt-out is not retroactive, getting the setting wrong on day one cannot be fully undone.
How does Tosheo.ai teach Perplexity inside a role path?
Tosheo.ai includes Perplexity tasks where research is part of the role: a sales representative preparing account research, a marketing manager scanning competitors' public positioning, or a real estate agent checking local planning information, all with fictional briefs.
Each lesson takes 10 to 15 minutes and ends with a research task marked against a visible rubric for question scope, source quality and verification, with specific feedback you can flag for review. The 7-day free trial needs no credit card.
Sources and scope
Perplexity Help Center, How does Perplexity work?, checked 19 October 2026 ↗
Perplexity Help Center, Getting started with Perplexity, checked 19 October 2026 ↗
Perplexity Help Center, Practical tips for using Perplexity, checked 19 October 2026 ↗
Perplexity Help Center, What are Projects?, checked 19 October 2026 ↗
Perplexity Help Center, Data collection at Perplexity, checked 19 October 2026 ↗