Master 95% of Perplexity In 16 Minutes (Full Guide)

0:00I've moved my research completely from Google to Perplexity, and it's been one of the best decisions I've made because behind the search box, it'll run the whole research job for you. The reading, the comparing, even the report at the end. So, in the next 16 minutes, I'll show you how to master 95% of this tool starting from scratch, and we'll work our way up from the easy wins to the heavy stuff, one level at a time. And by the end, you'll know which mode does which job, what's worth paying for, and

0:24the 5% where Google still wins. Perplexity is an AI search engine. And where Google leaves you a list of links to read yourself, this one reads them for you and gives you the answer straight out. So, it does both jobs at once, the searching and the reading, and it attaches the references to every claim, which is why people call it an answer engine. Take a question like, "When's the best time to see the Northern Lights in Iceland?" Google gives you 10 links, three of them tour companies with every reason to say always. Perplexity reads those same pages and gives you the months laid out, September to March with the best window

0:55in deep winter, and a little number at the end of every line. Those numbers are the whole point of the tool. You can hover over any of those numbers to see the source behind it, and if you click one, it takes you straight to the page it came from, so you can read it yourself. So, when an answer matters, you can check it in one click instead of just trusting it. And if a source looks

1:13weak, you'll notice it straight away. So, I'll follow up with, "And what about early February?" And it answers inside the same context instead of starting from zero. Under every answer, there's a row of suggested follow-ups, too, and they're decent, so the follow-up questions stay in one thread instead of five tabs. And a lot of the time, the sources themselves are trying to sell you something. So, what you can write is a prompt like, "Give me this again from independent sources only," and it rebuilds the answer without the tour companies in it. It's the same with news. If you ask about a story that broke overnight, the answer comes from

1:44pages that went up a few hours ago. One habit that pays off is glancing at what it read before you trust it because eight solid sources agreeing means the answer holds, and two random blogs means you ask it to look harder. And by the way, every prompt and template I use in this video is packed into one free doc. Grab it from the link below and follow along. Everything else is in the sidebar. New is where every search

2:06starts. Computer runs tasks for you. Spaces keeps related work together. Artifacts stores the things you've created. Customize lets you tailor your workspace. And history holds every thread you've ever run. One thing to keep straight from the start is that Google keeps maps, directions, shopping, and finding a plumber near you. Perplexity is for the questions where you need an answer you can back up. If you just need a website's address or the nearest cash machine, Google's still quicker. This is for the questions that matter. That's the core loop. You ask, you read the sources, you follow up, and everything else in this tool is a bigger version of that loop. So far though,

2:39it's answered me the way it would answer anyone, and 2 minutes in the settings fixes that permanently. So before anything else, I go into the settings under my account icon and open personalization. The box in there asks you to introduce yourself, and what you write works as standing instructions for every answer it ever gives you. Mine tells it to lead with the answer before the reasoning and keep things tight, to cite as it goes and use US context and metric. And if my question's vague, to ask me one clarifying question instead of guessing. My exact block is in the free doc below. Paste it in once and it

3:09pays you back on every search after. And there's a response length option in the same settings. So everyday questions come back short while the big ones still get the full answer. Perplexity has also been building out memory. So it carries what it learns about you between chats. That's the part that compounds. Mention once what laptop you're on or what course you're doing, and months later the answers still account for it without you repeating yourself. While you're in settings, find the AI data retention option because if you rather your questions stay out of model training, that's the switch. And turn the email notifications off or every long research

3:41run pings your inbox for the rest of time. Now the same northern lights question gets a different answer. The dates arrive first in tight bullets with US half terms factored in instead of three paragraphs of waffle before the point. It also stops guessing at what I meant. A vague question now gets one clarifying question back, which saves the answer being confidently wrong. It takes 2 minutes, you do it once, and every search you run for here gets better. Personalization changes the tone, but the little row of controls under the search box decides what the answer is even built from. The mode switch is the big one. Search is the

4:13default fast answers for everyday questions, and it's what I leave on eight times out of 10. And free gives you three of those a day, so spend them on questions that deserve them. I'll spend one of those on the in-between questions, something like which US bank account suits a student with a part-time income, where the answer needs a dozen pages compared properly, but doesn't need a whole report. Deep research is the thorough one. It runs for minutes at a time and builds you a full report. And the plus menu now focuses on adding things into Perplexity. You can upload files or images, hook up external

4:42connectors, or drop work into spaces. Say I'm buying running shoes and I want the runner consensus instead of the review site take. So, I name where I want it to look and ask what long-distance runners on Reddit say about carbon-plated shoes for beginners, worth it or overkill. That one instruction sends it into the actual forum threads, runners arguing in the replies with links straight to the posts. For a buying decision, that's the most useful data on the internet. Switch the sources over to academic and it

5:08works just as well on serious material. I'll ask what published studies say caffeine does to deep sleep, and every line of the answer footnotes to a paper instead of a blog. It's the same for numbers, ask how two chip companies' margins compare this year and tell it to work from the filings. And that's exactly where the answer comes from, and you can run both angles on one question, ask what the studies say and what people actually report using it, and the research and the real-world experience come back side by side. One more trick that works in any mode is ending the question with answer as a table, and a

5:38comparison arrives as a table instead of paragraphs, ready to drop into your notes. The model picker sits at the end of the row, and on free you don't get one. Perplexity picks the model for you. On Pro, you choose between its own sonar models and the latest from OpenAI, Anthropic, and Google. One thing worth knowing before you pick a favorite, though, is that the same model behaves differently inside Perplexity than it does in its own app because everything

6:01here is tuned for search and citations. So, I leave it on best and let it root. Pro adds image and video generation on top as well, though it's a bonus here rather than the reason you'd pay. You can attach files to a search, too, PDFs and images with a lower cap on free. Attach the module handbook or a contract and ask against it, and it reads the file and the web together. So, what does this clause mean gets answered from the document and the rules around it. Pro can also connect your own accounts as sources, Gmail and Drive, so a search pulls from your documents next to the

6:31web. So, the reading list a lecturer emailed in week one or the notes doc buried three folders deep in Drive turns up inside the answer without you going hunting for it. And the mic next to the box does dictation plus a full voice mode for when your hands are busy. I keep the voice side for walking questions, the ones you'd otherwise forget before you're back at a desk. The search box hasn't changed, but now you're choosing what it reads, and the answer quality follows. A single search can only do so much, and some questions need a written plan and a lot of reading before the answer means anything. For

7:02exactly those questions, I flip the mode switch over to deep research. It's the mode for comparing two options side by side or checking a market before you put money into it. So, the split between the two is this. Search answers in seconds. Deep research is for anything you'd otherwise open 20 tabs over. The skill with this mode is the brief because

7:20vague briefs get vague reports back. Mine follows the same shape every time. What I'm trying to decide, what the report has to include, which sources to trust, and the form I want the answer in. That template's in the free doc below, too. And the sources line is the one that changes the result most. Tell it to prefer owner surveys and official filings over reviews, and the whole report ends up built on numbers instead of opinions. So, I'll run one, electric versus hybrid for a family car in the US. Total 5-year cost including depreciation, charging against fuel, insurance, and servicing with the strongest counter-argument for each side. It lays out its plan first, then

7:57it goes off and reads, and you can watch the source count climb. There's a notify option, too, so you close the tab, get on with your day, and it gives you a nudge when the report's ready. What you get is a structured report with sections and a cost table, numbered citations on every claim. It leads with the takeaways, then the detail underneath, so you get the gist before you read the whole thing. And it stays a thread, so I'll push it further, go deeper on depreciation alone, and it extends the research it already did. It does the same job on essay topics, ask what the research says about remote work and

8:28productivity, and it groups the studies by which side they land on, which is an essay plan by itself. On free, you get one research run a month, so save it for something that matters. Pro moves it to monthly limits you'd struggle to hit with normal use. So, don't waste a run on anything the normal search could answer. This mode is for questions with several parts to them. And when the report's done, open the sources tab and spot-check the two claims your decision leans on hardest. It takes a minute, and it's what the citations are for. One brief and 16 pages of reading come back as one cited report. And a report like

8:58that shouldn't get lost in your library between a shoe search and a Wi-Fi question. So, there's a place built for it one click away. Perplexity calls that place spaces. The problem they solve is repeating yourself because every chat You re-explain the situation, the constraints, and what you've already ruled out, and a few days into a car hunt or a dissertation, you're pasting the same paragraph in yet again. Space hold all of that once. So, I'll start a new space and call it car decision. It asks for instructions straight away, and what you write in that box shapes every thread you'll run inside. Mine say,

9:29"We're choosing a family car for the US. Budget is 25 grand. Two child seats have to fit, and always compare any new suggestion against the short list in the files. Then I attach the files, which means my short list spreadsheet, the research report we just ran, and now this space knows more about the decision than any single chat could. Files are for the facts, the data and documents it should work from, and the instructions are for behavior, how it should treat what's in them. Pro takes the cap to 50 files per space, so the two setups don't blur. The settings from earlier tune how it answers you everywhere. A space tunes

10:01what it knows about one piece of work. The rule I use is that if I'd end up typing it in every thread, it belongs in the instructions instead. Every thread I start inside stays grouped there. So now I ask, which of the three short-listed cars holds its value best? And it answers from my own spreadsheet plus fresh sources, filtered through my instructions, without me re-explaining a thing. And the instructions say, "Cite from my files first and flag anything older than the module allows." New threads start with the context already

10:28in them because the space holds it. Everything the project holds is still a chat, though, and you can't exactly send a chat to a client or a teacher. There's a fix for that in the plus menu from earlier, so I'll click upload files and images. [music] It's the part I'd show someone who thinks Perplexity is just search. You give it a brief, it goes away for 10 minutes or more, and it comes back having built a thing, whether that's a formatted report, a working spreadsheet, or a dashboard you can click around. So I'll feed it the car research, "Build me a one-page decision summary in plain

10:56English plus a cost dashboard comparing the three cars over 5 years." Then you watch it work, searching, then writing code, then laying out the pages, and it's the same engine from earlier, just building instead of answering. The dashboard it gives you back comes out really, really clean, and it cites the cost sources underneath. For one prompt,

11:14that's a ridiculous amount to get back. Then I'll push it one step further, turn the same research into a five-slide summary for a presentation, and it drafts the deck with the numbers already charted. And apps means what it says, describe a small tool, like a revision planner or a cost calculator, and it builds a working page you can use in the browser. Before you send anything out, it's the same habit as the rest of the tool. You check the numbers against the sources it lists because it's your report now, and the mistakes would be

11:39yours, too. Any thread converts as well. There's a convert to page option that turns a chat into a shareable page with the citations kept. So, the research you ran can go to your team or the family group chat as a link that reads like a document instead of a screenshot of a chat. You can set scheduled tasks up on the paid side, too. Mine checks EV grant news every morning and emails me anything that changed set once, runs itself. That's the pattern for anything you're waiting on, whether it's a grant decision, a rule change, or a price drop. So, you set the task once and it

12:08does the checking instead of you. All of this is the pro side. Free doesn't get the builds, and honestly, this is the feature that justifies the 20. Research goes in, something finished comes out. Each mode has proven itself on its own, and one live deadline run through all of them is a different thing entirely. Take one assignment from nothing, a term paper on whether minimum alcohol pricing

12:28worked. It starts as a normal search. What is minimum unit pricing, and where has it been tried? In 30 seconds later, I know the basics, and that Scotland is the test case. So, Scotland becomes the angle. Then comes the first question with a catch in it. Did drinking drop because of the policy or because of pandemic timing? And that's a deep research. It weighs the arguments on both sides instead of taking the first answer it finds. Now, the topic needs a space, so I make one. Put academic sources first, and flag anything older than 2020 with the module reading list going in as files. With the space

13:00holding the context, I send off the research run, a full report on the findings for and against, with effects by income group and the strongest criticisms of the Scottish data, and it joins the project files when it's done. Back at the desk, the takeaways sit up top, 40 sources underneath, and the two claims my argument leans on get spot-checked through their citations before anything gets written. Then it turns the report into a two-page summary for revision and a five-slide version for the seminar, charts included. And because new studies on this keep landing, a task runs every Monday morning and keeps me posted so the paper stays current while I write it. One

13:33topic went from a blank page to a cited deck without leaving the tool. Run that chain once on your own deadline and it becomes the default way you handle anything with one. The one annoying step left in all of this is opening a tab and Perplexity's answer is to replace the browser itself. That replacement is called Comment. The browser feature is free on Windows, Mac, and mobile. It works like Chrome, imports your bookmarks in a minute, and the address bar runs Perplexity. So every question gets a cited answer without opening a tab first. Everything from this video works inside it, too, because it's the same account just inside a browser. The

14:06agentic side, the assistant that acts on pages for you, filling things in and working across tabs, that's the paid part. The free plan gets you the browser and the answer engine everywhere you already spend your day, which for most of this workflow is the whole job. And the mobile app covers the rest. Same account, same project, so the question you ask at the bus stop is waiting in your library when you're back at a laptop, which just leaves the money

14:29question of what each plan gets you. Free covers the core loop, which means unlimited everyday searches, three pro searches a day, and one research run a month. Pro is $20 a month and it opens the model picker along with image and video generation. It also lifts the research and create limits, raises the file caps, and adds the scheduled tasks. And if you're a student, education pro is the one to know about. 10 a month with student verification and the pro searches go unlimited. And the way to decide is to stay free until you hit the research cap twice in a month because that's the sign you use it enough for

15:02pro to pay for itself. And with a student email, there's no decision to make at all. And there's one level left, the hardest one. Everything so far happened inside Perplexity's own apps. And the last level is using it inside something you build yourself. That's done through the API, and don't worry, you won't be writing any code as I'll walk you through it. The API is just the same sonar engine that's been answering us this whole video, except now you pay for exactly what you use, and a few dollars of credit covers a lot of questions. First, I grab an API key from Perplexity's API portal and top up a

15:33couple of dollars of credit. Then I open Claude, I'm using Fable 5 for this, and I paste in the build prompt. The prompt asks for one web page with a question box that sends whatever I type to Perplexity's sonar model and shows the answer with its references as links. Fable 5 writes the whole thing as a single file. I save it, open it in the browser, drop my key in, and I've got a working answer box. I'll ask it the northern lights question from the start of this video, and the same cited answer comes back, except now it's coming through a page I built. A couple of runs

16:01like that cost pennies because the sonar model is priced at about a dollar per million tokens, and a normal question only uses a few hundred. And the instructions and files habit is the one to keep because 5 minutes of it makes any AI tool work your way. That's the whole point of my school community AI fluency, getting that fluent with every tool worth using. So, join through the first link in the description and bring your research with you. Thank you for

16:23watching. I'll see you in the next one.