Hey! Ankur here, and this is the 23rd edition of Lazy AI — 5 mins of reading to help you stay ahead of the AI curve.
Today, I want to talk about how I improve accuracy with my research, and why everyone MUST use Perplexity.
Your AI “remembers” more than it “thinks”
How many times have you used ChatGPT to research something — got a confident, well-structured answer — and then found out the statistic it cited doesn’t exist?
Happens to everyone.
But here’s the thing: it’s not a bug. It’s a feature of how most AI tools are built. And once you understand it, you’ll immediately know which tool to reach for — and when.
So What’s Actually Going On?
Most AI tools — ChatGPT, Claude, Gemini — are built on something called a Large Language Models (LLM).
An LLM is an AI that has been trained on an enormous amount of text from the internet — articles, books, forums, everything. It learns patterns from all of that, and when you ask it something, it generates a response based on those patterns. It’s completing — very, very sophisticatedly.
Think of it like a very well-read friend who read everything on the internet up to a certain date, and then went completely off-grid. Ask them something recent? They’ll answer confidently — based on what they last read. Whether that information is still accurate? That’s a different question.
And when they’re asked something they don’t know, they’ll just make something up.
That’s Claude, ChatGPT and Gemini for you
But Perplexity is built differently
Perplexity isn’t really a chatbot. It’s closer to a search engine with a brain.
Every time you ask it something, it goes to the web first — before generating a single word of response. It pulls relevant pages, reads them, and then builds its answer from what it just found. The language model’s job is just to translate those search results into something readable.
The internal design principle Perplexity built this around: “You are not supposed to say anything that you didn’t retrieve.”
That one sentence explains why using Perplexity feels different. It can’t invent a source. It can only cite what it actually found. And it shows you those sources, inline, next to every claim.
You don’t need to be an engineer to talk AI. Subscribe to this newsletter — once a week, you’ll understand AI well enough to be the smartest person in the room at work.
Each issue is just 5 minutes — less than the time you spend doomscrolling before bed. Except, this actually moves your career forward. Join 8,000+ subscribers now.
But it’s not magic
I’ll be honest, because I don’t want to hype things just for engagement.
Perplexity still hallucinates. In one independent test, models were asked about a research paper that didn’t exist. ChatGPT, Gemini, and Claude all said they couldn’t find it. Perplexity fabricated an answer — and cited sources that didn’t support the claim.
How? When a search returns nothing useful, the model still tries to fill the gap. And when it does that, you’re back to the same hallucination risk as any other LLM.
That risk is lower with Perplexity than with other AI chatbots, but it’s still there.
Here’s the practical way to know: If a paragraph in Perplexity’s answer has no citation number next to it, the AI is speaking from training, not retrieval. That’s your cue to verify.
So what does this mean for you?
Here’s the thing: Perplexity reads the web for you. ChatGPT and Claude think for you. Both are useful, and neither replaces the other.
Use Perplexity when you need facts you can trace back to a real source — research, current events, checking if something is still true. Reach for ChatGPT or Claude when you need the AI to write, reason, or create something from scratch.
Here’s how I use it
I’ve started using Perplexity as the final step before I publish anything — articles, newsletters, social posts. I paste in whatever I’ve written and ask it to verify every claim I’ve made.
In all the time I’ve been doing this, I’d say more than 90% of the times, the sources are correct. With ChatGPT and Claude, that’s lower.
No, you will not find 100% accuracy anywhere, and manual checks will still be needed. But with Perplexity, the errors are lower, which just saves your time
Try it before your next piece of writing goes out. It takes 2 minutes, and it might be exactly what you need.
You don’t need to be an engineer to talk AI. Subscribe to this newsletter — every other day, you’ll understand AI well enough to be the smartest person in the room at work.
Each issue is just 5 minutes — less than the time you spend doomscrolling before bed. Except, this actually moves your career forward. Join 8,000+ subscribers now.
Before you go, I have one request
I’ve been writing for over 3 months now, and I would love to know if I’m on the right track. So I need you to help me with two things:
Tell me if the weekly articles are adding value
Tell me what you would like to read on, so that it’s just more relevant for you
It’ll take a minute to reply, but will improve the quality of my work by 10X, so if you think my work has added ANY value to your AI learning journey, I’d love to see a reply to this email with your thoughts 😊
I’ll see you next week!
Cheers,
Ankur