How it works

How OpenNeedle answers your question.

No black box. Here is exactly what happens when you ask.

The process

You ask. Type your question in plain English. No account, no signup: just a text box and a button.
We search. Your question becomes a few search terms, and those hit BGPT, a scientific evidence engine that pulls real papers, claims, and their limitations from the literature. You see the exact terms we searched.
We weigh. We read the evidence and write the answer in plain English: what the science says, what it doesn't, and how sure we are. Every claim carries a number that links back to its source.
You check. Every source is listed under the answer with its title, journal, and a link to the paper. Click any number, open the study, read it yourself. You don't have to take our word for anything.

Up to 48 sources, not 3 abstracts

Most AI answers (Google's and ChatGPT's included) lean on a handful of abstracts or snippets. OpenNeedle reads up to 48 evidence records per question, each one a structured extract of the actual paper: the claims, the data behind them, who was studied, how the study ran, and what the authors admitted it couldn't show. That's where the catch hides, and abstracts leave it out.

You can check every claim

We'd rather you check us than trust us. Every claim links to the actual study, so if we got something wrong, you'll find it. The reasoning behind this method, with the numbers, is on Why we built it.

Where the data comes from

Our evidence layer is BGPT, which turns full-text scientific papers into structured records: each claim tied to the experiment behind it, the exact results, who was studied, how, and where it came from. You ask, we search BGPT for the best evidence, and the answer gets built from what comes back.