Paper Pilot

The search box covered roughly 220 million papers and returned linked references and citation details next to each result. You could read an abstract and its reference list in the same panel instead of hopping to another database. For a topic you barely know, that layout made it easier to spot the three or four papers everyone cites. Results pinned straight to a board, so search and reading list stayed in one place.

Every paper got a short recap that fit in a paragraph, plus a longer breakdown that ran step by step. You chose the depth, from a simple overview to a detailed walkthrough of method and results. The two speeds mattered for screening: recap for a fast yes or no, Full Explanation when a paper was clearly relevant but hard to parse. That split is still the most useful idea Paper Pilot shipped, and most academic research AI tools now copy some version of it.

A board was a shared page holding papers, notes and citations together. On paid plans you could chat with PDFs across the whole board, so a follow-up question reached all 50 papers at once rather than one file. Boards also supported sharing, which made them handy for a two-person project. Paper Pilot's free tier limited boards to five papers and a single board, so the comparison feature was mostly a paid thing.

Pilot AI answered questions grounded in the papers you had collected. Ask it what two studies disagree about and it pulled passages from both. Ask it for a gap in the literature and it pointed at sections that mentioned future work. Because the answers stayed inside your board, it behaved more like a research assistant than a general chatbot. The trade-off was scope: nothing outside the pinned set came into the answer.

Citations were generated as you wrote, in the format you set, and references could be inserted without retyping author names and years. Paper Pilot also kept a list of references per board, which helped when a draft needed a quick bibliography. The feature worked best inside the board's own editor. Exporting to a reference manager was still a manual step.

The writing assistant drafted sentences and paragraphs on top of your collected sources and suggested citations as you typed. On the Pro plan it was unrestricted, and it indexed both online and uploaded papers. Think of it as a literature review tool with a text editor attached: useful for a related-work paragraph, less so for original argument. Drafts still needed a human to check that each citation said what the sentence claimed.