digest
Short Summary
Main recommendation: Use Perplexity as an AI search engine for research rather than a conversational chatbot, configuring its settings first and choosing the search sources and model that fit your query. Perplexity returns fast, citation-backed answers drawn from multiple sources, and the presenter uses it as his everyday Google replacement and as the starting point for content research. He does not recommend it for writing, which he prefers to do in Gemini or Notebook LM.
Key Points
- Perplexity is built as an AI search engine, not a standard chatbot: answers are fast and every response includes citations you can hover over, open, and verify.
- The presenter finds its answers more accurate than a traditional chatbot because they synthesize multiple sources.
- The Search mode is used about 80% of the time for everyday questions, with Research (deep research) and Labs (documents, reports, small apps) for heavier work.
- The Sources toggle lets you add Social (Reddit, Twitter) for real-time public opinion, Finance (SEC filings), or connected Gmail and Google Drive.
- For product research, Perplexity presents answers in titled, practical sections, while ChatGPT with the same model gives a more conversational text block.
Details Worth Noting
- Concrete instructions:
- Create an account at perplexity.ai using just an email address.
- In settings (click the down icon at bottom left, then Preferences), choose an image generation model; the presenter picked Seedream 4.5 over Nano Banana or GPT Image 1 because he already had those elsewhere.
- Toggle off AI data retention if you do not want your inputs used to train their models.
- Keep only the Spaces option on in the left sidebar; the presenter finds Discover and other options cluttering.
- Under Personalization, leave reference search library and reference saved memories on so Perplexity learns from past chats; you can skip manual profile details, and optionally turn off location sharing.
- Under Notifications, toggle off email alerts, which otherwise announce every completed deep research task.
- Leave the model selector on "Best," which lets Perplexity pick a model per question; the presenter found model choice made little difference for his research.
- For a daily routine, leave Thinking on, enable Social, and ask something like "check Twitter for important updates in the past 24 hours related to [topics]," then expand the sources dropdown to see which sites were checked.
- Use Tasks in settings to schedule recurring queries, such as a daily morning check of specific accounts.
- Create a Space for an ongoing project, add custom instructions (e.g., "always prioritize posts from X"), and upload PDFs, Google Drive files, or publicly accessible links so answers draw from them.
- The Sources toggle matters most for opinion research: with Social on, the presenter checked public reaction to a new iPhone-style product and got a consensus drawn from real-time posts.
- For product-purchase research, the presenter compared a massaging heated eye mask query in Perplexity and ChatGPT using the same model; Perplexity's structured sections suited his need for quick, practical answers.
- The presenter's workflow: research in Perplexity, then move the findings into Notebook LM or Gemini for scripts and writing.
- Perplexity's voice input and voice mode let you dictate or hold a spoken back-and-forth conversation, and attachments can be added through the attach button.
Bottom Line
Perplexity works best as a fast, citation-first research engine: configure preferences, choose the right sources, and use Spaces and Tasks to avoid repeating work, but hand off writing to a different tool.
