I Added an AI-Powered Tool to Find and Explore Related Research Papers — Here's What I Learned From Researchers

Long before I launched the discuria website, the idea in my mind was well-formed: a platform where researchers could discuss academic papers with people around the world. Now that I actually launched the website and started talking to users, I realized that while discussion is the end goal when many users use the website, at this early stage it's not going to get someone to open a new tab.

What researchers actually told me

The feedback I heard wasn't "I wish I had somewhere to discuss papers." It was something more fundamental: finding the right papers in the first place is genuinely hard.

Google Scholar returns broad results that may be years old, and journal databases are overwhelming without good filtering. And once you find one relevant paper, tracking down the web of related work around it is tedious and time-consuming.

The second thing I kept hearing: AI has already changed how researchers read papers. Most researchers now use AI tools to summarize papers and ask questions as they read. Not as a replacement for deep reading, but as a first pass to orient yourself before diving in. And I couldn't ignore this reality.

What I built in response

Better discovery: sorting, ranking, and related papers

The browse page now ranks results three ways — by relevance, recency, and "hottest" (a combination of citation count and recency). This means users can find the most cited paper on a topic, or surface what's being published right now, depending on what they need.

But the feature I'm most excited about is the interactive node graph. Search for a paper, and you'll see a visual map of related papers connected to it. Click any node to expand it, revealing another layer of connected work. It's designed for the moment when users find one good paper and want to understand the entire landscape around it, following ideas outward rather than running repeated searches.

Additionally below each paper in the reader, there's also a "You May Like This" section that surfaces related papers automatically, so discovery continues naturally as you read.

AI paper assistant, powered by Claude

There's now an AI assistant built directly into the paper reader, sitting right next to the PDF as you read. Ask it to summarize a section, explain a method, or clarify a result. It has context of the paper you're reading and responds in plain language.

What I think makes this genuinely useful, rather than just another AI wrapper, is what sits on the other side of the PDF: real human annotations and comments from other researchers. You have AI assistance on one side and human insight on the other, while the paper is in the middle. Both are available without switching tabs or copy-pasting text somewhere else.

The honest reason for these changes

I realized that building a discussion platform with no users is a chicken-and-egg problem. People come to discuss, find nobody there, and leave. The discussion will never start.

Instead of waiting for that to fix itself, I focused on building things that are useful to a single researcher, alone, on their first visit, regardless of whether anyone else is there. Good search, good discovery, good AI assistance. These features have standalone value. They don't require a community to exist first.

The discussion layer is still there, and is still the long-term vision. But now there's a reason to show up before the community exists.

Both features — the discovery tools and the AI assistant — are completely free to use. The AI assistant requires a free account. Everything else works without signing up.

If you're a researcher, a student, or just someone who reads academic papers, I hope you can try it at discuria.org. And if something doesn't work the way you'd expect, or you have a feature you wish existed, give me feedback because I’m genuinely listening.