After the reading comes the slow part: sorting what you noted, and writing up what each paper does. Rectoly hands both to an AI that runs on the Android tablet in your hands. It never writes a word of your notes, every sentence it writes about a paper points to the page it came from, and nothing leaves the device — unless you decide to bring your own API key.
Open Across papers, choose a collection or your whole library, and tap Organize with AI. The notes you wrote — comments on highlights, sticky notes, text boxes — are grouped under headings by topic. The AI writes only the headings; every note stays exactly as you wrote it, with its paper and page, and a tap opens the paper at that place.
Tap Save as a research theme and the result is an ordinary theme: rename the headings, move a note the AI misplaced, add sentences of your own, and export it to Markdown or Word with a reference list — or, in the Zotero edition, send it to Zotero as a note. A note that fits no heading is not dropped; it waits under a heading of its own.
Choose Paper cards… in the library’s menu and pick the papers. For each one, the AI writes out the research question, method, data, findings and the limitations the authors state — from the paper’s own text, in the paper’s own language. Each sentence ends with its page; tap it and the paper opens on the paragraph it cites.
Open the table and the cards line up: a row per paper, five columns, a page on every sentence. Share the folder and you have the table as Markdown and CSV for your spreadsheet. In the Zotero edition, every card can be saved to its paper as a Zotero note, where the plugins and tools that read Zotero can find it.
The AI writes from the paper’s paragraphs; Rectoly then checks every sentence against them. A sentence with a claim or a number that is not in the paragraphs it cites is dropped, and the card says how many were. Stated limitations holds only what the authors themselves wrote. A field the paper does not state stays empty rather than guessed.
Everything runs on this device with the same 2.5 GB model translation uses, downloaded once. Your notes and your papers are not sent anywhere. On a device that cannot run the model, or if you want a stronger one, choose Your own API key in Settings › AI — OpenAI, Claude, Gemini, DeepSeek, OpenRouter, an OpenAI-compatible server, or Ollama or LM Studio on your own network. Rectoly says what will be sent and asks first.
Card the papers in the collection and the table is the first draft of your literature review matrix: what each paper asked, how, on what data, what it found and what it admits. Every cell leads back to a page.
The notes you left while reading are grouped by topic and saved as a theme, so the shape of your argument is there before you start writing — in your own words, not the AI’s.
The model runs on the device. A manuscript you are reviewing, or a draft nobody has seen, is organized and carded without going to any AI service — unless you choose your own key, and then you are told exactly what goes.
Read with the same library and the same annotations. The app downloads from this site, and one purchase covers it together with iPad, Windows, Mac and Linux.