What do you like best about Perplexity?
Most of what I do in a week is research that has to hold up to scrutiny: market sizing, competitor moves, vendor shortlists, claims I am about to repeat in something I publish. Perplexity earned its spot because it answers those questions in one pass and shows me where every line came from.
The inline citations are the feature that changed my habit. Every claim in an answer carries a numbered source I can click, so I am not trusting a confident paragraph on faith the way I had to with the first wave of chat tools. When I am checking whether a competitor actually shipped a feature or just announced it, I read the answer, then jump straight to the source that backs the specific sentence I care about. Trust but verify stops being a separate step and becomes part of reading the answer.
The real-time search is the other half of why it works for this. It pulls current information rather than answering from a frozen training cutoff, so a question about a pricing change or a funding round from last week comes back with this week's facts. I have caught more than one tool giving me stale answers on things that had quietly moved, and that does not happen here, because it actually goes and looks before it replies.
Model selection on the paid plan is something I lean on more than I expected. I can point a query at GPT, Claude, or Gemini depending on the task, so I use one model for tight factual lookups and switch to another when I want a more careful read of a long document. Having that choice inside one interface means I am not paying for and juggling three separate subscriptions to get the right model for the job.
Document analysis has quietly become one of my heaviest uses. I drop in a PDF, a long report, a pitch deck, sometimes a messy screenshot of a chart, and ask specific questions against it instead of reading the whole thing front to back. For vendor research this is the fastest part of my week, because I can pull the relevant numbers out of an 80-page document in the time it used to take just to find the right section.
The follow-up flow is what makes it feel like research rather than a single lookup. An answer comes back with suggested next questions, and more useful than that, I can keep narrowing in plain language against what it already gave me, so a broad opening question turns into a tight one over three or four steps without my having to restate the context each time. When I am scoping a market I barely know, that progression from a wide question to the specific thing I actually needed is most of the value, because I rarely know the precise question to ask until the first answer shows me what is there. It rewards working a topic in layers, which is how I think through unfamiliar ground anyway.
Spaces is where the ongoing work lives.
Instead of a flat list of one-off questions, I keep a Space per topic with its own uploaded files and context, so a running competitor analysis or a market I am tracking holds together across days rather than scattering into threads I cannot find later. When I come back to a topic a week on, the context is already sitting there waiting.
A few smaller things I use often enough to call out:
- Focus modes that narrow a search to a domain, like an academic mode for published work or a finance view for market data, which cuts a lot of the SEO filler out of the results.
- Deep Research for the times I want a longer, structured report on a topic rather than a quick answer, when I am willing to wait a minute for something more thorough.
- Pages, which turns a piece of research into a clean shareable write-up so I am not rebuilding findings from scratch when I hand them off.
The Comet browser is a recent addition I did not expect to keep using and now do. It folds the same answer-and-cite behavior into actual browsing, so cross-referencing a handful of tabs or summarizing a page I am already on happens in place rather than copying text back and forth. For research that is browsing-heavy it removes a lot of the shuffling between a search box and the pages it points me to.
The interface stays out of the way, which for a tool I open this many times a day matters more than any single feature. It is quick, it does not bury the answer under chrome, and the trip from a question to a usable answer is a short one. That economy is a large part of why reaching for it has become automatic rather than a decision.
Review collected by and hosted on G2.com.
What do you dislike about Perplexity?
None of this is free of friction, and a couple of the rough spots are worth being straight about before anyone builds a workflow around it.
Accuracy on deep or niche topics is the one I watch most. On well-trodden questions it is reliable, but push into a narrow technical corner or an obscure subject and it can oversimplify, flatten a nuance, or occasionally assert something the cited source does not quite support. The citations are the saving grace, because the fix is to click through and read the source myself, but it means I do not take a surprising answer at face value on anything that matters. The habit I have settled on is to treat it as a fast first pass, not a final authority, on anything I am going to act on.
It is also not the tool I reach for when I need to actually write something long. For drafting marketing copy, a piece of content, or anything that needs a sustained voice, the output is competent but flat compared to what I get elsewhere, and I have stopped trying to force it into that role. I use it to gather and verify the material, then draft in a different tool. Knowing where its strength ends has made it more useful to me, not less.
The free tier has tightened to the point where it is really a sampler now. The daily allowance of the better search and research queries is low enough that anyone using it seriously will hit the wall fast and get pushed toward the paid plan. I happen to think the paid plan earns its cost for how much I use it, but I would rather the free experience left more room to evaluate the thing properly before paying.
The Comet browser comes with a tradeoff I would not gloss over. An AI browser that can act on pages is genuinely useful for research, but the same capability carries well-documented security questions around prompt injection, and I would not run anything sensitive like banking or confidential account work through an agent browser at this stage. For the research and summarization I use it for, the tradeoff sits fine with me. For anything where a hijacked instruction could do real damage, I stay in a normal browser. It is a real consideration to go in aware of, not a reason to avoid the tool.
A few smaller things that come up:
- Citation quality varies. Most sources are solid, but now and then an answer leans on a thin or SEO-heavy page, so a glance at what it is citing is worth the second it takes.
- On the free experience there is little visibility into which underlying model served a given answer, which matters when you are trying to work out why two similar questions came back at different depths.
- Very long documents can get summarized a little too aggressively, so for a dense report I sometimes have to ask follow-ups to pull back the detail it compressed out.
None of these have moved me off it. They are the shape of working with a tool that tries to do a lot, and most of them come down to using it for what it is good at and verifying the rest, which is a fair price for what it does well. Review collected by and hosted on G2.com.
