Course: Perplexity Tutorial for Beginners
This Perplexity tutorial for beginners teaches you to turn a question into a research brief you can defend. You will narrow the question, collect sources, check claims and write a short report that separates evidence from interpretation. The worked project compares ways a small team can organise product information. You can practise with ordinary search and a notes document; paid research modes, browser agents and file uploads are optional. The key skill is reviewing the evidence behind the answer, rather than collecting a large number of links.
Independent course by ThinkQuarry. Level: Beginner.
Course level: Beginner. Format: Seven practical lessons with prompts, exercises and a final project. Cost: This ThinkQuarry course is free to read. Optional product features may have separate availability or limits.
What you will create: A focused research question, a checked source ledger, a comparison table and a concise report with limitations.
This course belongs to our AI productivity tools and workflows learning hub. Use the hub to choose your next course.
Before you begin
Keep a notes document for your input, prompt, reviewed answer and corrections. Use public or fictional examples while practising. Check the official product interface for available features; do not purchase a plan to complete these exercises. When a feature is unavailable, use pasted text and a manual checklist. These are independent ThinkQuarry lessons, not an official provider certification.
Lesson 1: Turn a broad topic into an answerable question
Start with a decision rather than a topic label. “Research productivity” is too broad; “Which product-information workflow fits a three-person team with no budget for software?” gives the research a purpose. State the audience, location where relevant, time period and constraints. List what would change your decision. A focused question prevents a long answer full of interesting but irrelevant details. Use fictional or public examples while learning. If the question concerns sensitive business facts, remove unnecessary identifiers before submitting it.
Try this prompt
Help me narrow this research question: a fictional three-person shop needs to organise product facts using tools it already has. Budget for new software is zero. Propose three focused questions and explain which evidence would answer each. Do not recommend a product yet.Exercise and success check
Choose one question that leads to a concrete decision. Write the decision criteria before collecting sources.
Lesson 2: Ask for source types that match the question
Different claims need different evidence. For a software feature, the provider’s official documentation is a better starting point than an outdated comparison blog. For a scientific result, look for the original research paper. Ask Perplexity to identify source dates and the part that supports the claim. Do not request an arbitrary large source count; three relevant primary sources can be more useful than twenty summaries of the same announcement. When search results are thin, state the gap and change the question rather than asking the assistant to fill it with certainty.
Try this prompt
Find primary documentation relevant to our question. For each source, provide its URL, publication or update date if visible, and the feature or limitation it supports. Distinguish official documentation from commentary. Mark missing dates as unknown.Exercise and success check
Open each proposed source. Keep it only if it addresses the actual question and you can identify the supporting passage.
Lesson 3: Build a claim ledger and inspect citations
Perplexity answers include citations, but the existence of a citation does not prove that every nearby statement is correct. Open the source and check the wording, date and context. In your notes, create columns for claim, source, supporting passage, date and confidence. Use short paraphrases rather than copying large sections. A citation may support a product’s existence without supporting a conclusion about its suitability for your team. Separate those two claims. If a link is unavailable, do not count it as verified evidence.
Try this prompt
Turn this answer into a claim ledger. Split sentences that contain more than one claim. For each claim, identify the cited source and what I must verify on the original page. Leave confidence as unchecked until I review it. [Paste answer.]Exercise and success check
Check three claims manually. Mark unsupported or only partly supported claims and revise the report accordingly.
Lesson 4: Use follow-up questions to expose uncertainty
Ask what could make the answer wrong, which facts may be outdated and where sources disagree. A good follow-up narrows a specific gap rather than requesting “more detail” about everything. If two sources give different limits or interface instructions, compare their dates and intended users. Do not blend incompatible versions into one answer. Sometimes the correct conclusion is that the available evidence does not settle the question. Record the uncertainty and identify the next useful source or practical test.
Try this prompt
These two sources disagree about a feature: [paste short excerpts and dates]. Explain the disagreement without choosing a winner prematurely. Identify whether the difference could be date, plan, region or platform. List what would resolve it.Exercise and success check
The response should distinguish a documented fact from a possible explanation. Do not publish a hypothesis as the final feature description.
Lesson 5: Compare alternatives against explicit criteria
Create your comparison criteria before asking for a winner. For our fictional shop, useful criteria are cost using existing tools, ease of updating facts, access control and version history. Ask for a table with a source for each externally verifiable claim. Keep your own preferences separate from product capabilities. A recommendation can be reasonable even when it is not universal; explain which assumptions make it suitable. Where exact features vary by account, test the team’s real setup instead of relying on a generic comparison.
Try this prompt
Compare the researched options using only our checked evidence. Criteria: no new software cost, clear ownership of facts, version control and simple handover. For each cell, cite the ledger entry or mark unknown. Explain trade-offs rather than declaring a universal best option.Exercise and success check
Every feature claim should connect to a ledger entry. If one option has more unknowns, show them rather than filling the cells with guesses.
Lesson 6: Write a report that preserves evidence and limits
Draft a short report only after you have checked the important claims. Use a structure such as decision, evidence, trade-offs and next test. Clearly label interpretation and avoid implying that a source endorses your recommendation. A report should be useful to someone who has not seen the conversation. Include enough context to understand the constraints, but remove research that does not affect the decision. If a feature is unavailable, your fallback can be a simpler process using tools the team already has. That is a practical decision, not a research failure.
Try this prompt
Write a 400-word brief from this verified ledger. Sections: decision for this team, supporting evidence, trade-offs, unanswered questions and next test. Do not introduce new facts. Use ordinary source links beside factual claims. Ledger: [paste].Exercise and success check
Compare the report with the ledger. Remove any claim introduced during writing that was not checked during research.
Lesson 7: Finish with a reproducible research project
Package the question, criteria, source ledger, comparison and final brief together. Add the date you checked the sources, because product documentation can change. Ask another person to inspect one important claim and follow your reasoning. If they cannot reproduce the conclusion, improve the evidence trail. Keep a list of what should be rechecked later, such as limits, availability or prices. Research is complete when the decision has sufficient support for its purpose, not when the assistant produces a confident ending.
Try this prompt
Audit this research pack for reproducibility. Can a reader identify the question, assumptions, original sources, unsupported claims and next action? List specific gaps without adding new research results. [Paste pack.]Exercise and success check
Your final deliverable should contain at least three manually checked claims, an explicit limitation and one next action tied to an unresolved question.
Final project checklist
- Your output addresses the original task and intended audience.
- Supplied facts are preserved and unsupported claims are removed.
- Important external claims have original sources you opened and checked.
- Calculations, examples or interactive behaviour have been inspected where relevant.
- Unknown information is clearly labelled and the next human decision is stated.
Save your first and final versions. Compare the corrections you made: they reveal which part of your brief or review process needs improvement. Completion means producing a useful, checked deliverable; it does not imply an official qualification or guaranteed result.
Frequently asked questions
Are all cited answers accurate?
No. Open the original sources and check whether they support the exact claim, including its date and context.
Do I need Pro Search or an agent to finish the course?
No. The exercises use source-based research and a notes document. Advanced modes can be optional aids when available.
Can I copy a complete source article into my report?
Use your own explanation, short necessary excerpts and attribution. Respect copyright and do not reproduce a source wholesale.
Continue your learning
- AI research workflow for blog posts
- Perplexity Comet research workflow
- Grok business planning course
- Claude document analysis course
Official references and editorial note
Product guidance checked on 7 October 2026. Features and interfaces can change. The lesson prompts and practice projects are original ThinkQuarry examples; check current official guidance when a feature differs from your account.
