One Prompt = Full Literature Review (Why Every PhD Needs This Tool)

Thesis AI positions a single prompt as a shortcut to a full, fully referenced literature review—turning what typically takes days or weeks of manual reading and synthesis into a document that can be generated in minutes. The pitch is straightforward: gather a focused set of references for a new research niche, import them into Thesis AI, and then ask for a literature review that produces a long, structured draft (the example output lands at 41 pages) with citations tied back to the sources.

The workflow starts with setting up the document style inside Thesis AI. Users can choose citation format (the example uses IEEE), specify citation granularity (whether to reference papers and/or page-level details), select document language, and adjust a “temperature” control that shifts writing from strictly academic (temperature set to 0) toward more expressive phrasing (higher values). Once the chat interface is ready, the key step is feeding Thesis AI the right bibliography.

Instead of manually hunting for papers, the process emphasizes importing references from Zotero. The transcript also describes a pre-collection phase using discovery tools like Consensus and illicit (and optionally Research Rabbit) to generate reference lists for a chosen topic—here, “preparation, characterization, and testing of light harvesting additives for organic solar cells.” Results can be exported (including RIS files) and then imported into Zotero via Zotero’s File → Import flow. The next hurdle—getting PDFs attached—is handled through Zotero’s “find full text” function, which searches for accessible versions of the papers.

A notable detail is a “sneaky trick” mentioned for configuring Zotero’s PDF resolver behavior via the Zotero config editor, including an extension related to PDF resolvers and a setting that can retrieve PDFs from a source referred to as “SIHub.” The transcript explicitly frames this as potentially “naughty” and advises against it, but it underscores that the quality of the final Thesis AI output depends on having full-text PDFs available.

With PDFs attached, Thesis AI’s “import collection” feature pulls in the entire Zotero collection (the example imports 18 files). After that, the user provides a short topic description and runs the generation. The system reports writing progress and can take up to 30 minutes depending on the number of files (the transcript mentions up to 100). At the end, the output can be exported to Overleaf or continued via a new chat.

The generated literature review is presented as a structured draft with sections like an introduction and an abstract, plus clickable references that let the reader trace claims back to specific papers. The transcript stresses that this is not a replacement for critical reading; it’s a starting point for understanding mechanisms, context, and how studies relate—then verifying details in the original literature. The practical takeaway is that a researcher can go from “zero references” to a thesis-length introduction draft quickly, then use the citations to guide deeper study and eventually shape the work into a thesis, peer-reviewed paper, or review article.