The AI That Will Save You WEEKS of Research Time! Unriddle AI

Unriddle is positioned as a research-focused AI workspace that turns imported papers and other materials into searchable, question-answerable knowledge—complete with citations that link back to the exact parts of source documents. The core payoff is speed: instead of rereading stacks of articles to find what matters, researchers can ask targeted questions, get answers, and then click references to verify claims directly in the original text, reducing the risk of unsupported “hallucinated” details.

The workflow starts with importing. Unriddle accepts documents plus other formats—plain text, videos, audio, websites, images, and recordings—then builds a library where each item can be summarized and queried. A key feature is suggested questions that help users who don’t know where to begin. For deeper exploration, users can ask the system about specific concepts (the transcript uses an example about how the complexity of active materials affects outcomes), and every response includes clickable references that jump to the relevant passage in the paper.

Beyond single-article Q&A, Unriddle supports grouping papers into collections so users can ask questions across a curated set rather than their entire library. The transcript describes a manual grouping process (select papers, then group them under a named idea) and then querying that group—for instance, asking which devices are “best” across papers about OPV devices. The answers are paired with links back to where the information was drawn in each source, including navigation to the cited sections.

Users can also query their entire reference library at once, such as asking for the best indoor organic photovoltaic devices. Results come back quickly and include references that can be opened to check the underlying sources. One limitation noted is that chat history doesn’t appear to be saved or is at least not obvious, which can be annoying when revisiting earlier questions.

Unriddle also includes an academic writing area. It offers an AI writing assistant embedded alongside a writing canvas, with an “AI partner” that can suggest structures and help draft sections of a literature review. The writing interface is described as powerful but somewhat clunky—formatting and highlighting controls are not as intuitive as in competing tools. Still, the system provides practical actions like rewrite, paraphrase, shorten/expand, and “find citations,” with citation insertion and format settings.

A newer “record” feature lets users dictate ideas by voice and then convert the transcript into polished text suitable for peer-reviewed writing. Early use is promising for brainstorming, but the transcript highlights a frustrating gap: prompts that reference a figure described verbally may fail because the system can’t reliably connect the recording content to the requested figure context. Finally, a graph/visual element appears to be more confusing than helpful at the moment, with unclear purpose and limited discoverability.

Overall, Unriddle’s strongest value is extracting and verifying information from imported research. Its writing and recording features look like meaningful additions, but they’re still maturing compared with the smoother, more reliable paper-unriddling experience.