Write articles by chatting with your notes (using AI)

AI-assisted note research can turn a pile of sources into a usable article draft in under an hour—by combining three steps: generating an outline, pulling reputable references, and then letting an AI write from highlighted notes.

The workflow starts with choosing a topic and prompting an AI chat to produce a structured outline tailored to the writer’s needs. For insomnia, the outline is designed to include an introduction that frames the condition, a section on the problems insomnia can cause, and a deeper breakdown of underlying causes and solutions. The requested solutions are organized into practical categories—physical exercises, mental exercises, and things to avoid—ending with a structured sleep plan as a conclusion.

Next comes source selection. Because the Reflect AI chat used in the workflow can’t directly browse the open web like ChatGPT can, the process splits: Reflect generates the outline and then requests reputable sources, while ChatGPT (browser version) is used to perform the online research and return links. The user copies the outline and the source list into a new note inside Reflect, then instructs ChatGPT to research each outline section using only the provided reputable sources. The output is a set of targeted research results—organized around insomnia symptoms, causes, and treatments—meant to guide what to read and save.

The “human” part is then executed inside Reflect using a Chrome extension for capture and organization. Each linked article (for example, from Sleep Foundation and Mayo Clinic) is opened, key passages are highlighted—especially definitions, symptoms, causes, and lifestyle or treatment guidance—and those highlights are automatically stored under the note’s links. The goal isn’t to archive entire pages; it’s to extract the key information the eventual draft will need. After reviewing multiple sources (about 10 minutes in this run), the notes are tagged and filtered using Reflect’s advanced search so that the AI can answer only from the curated set of highlights.

With the research locked in, the article draft is produced by pasting the saved outline into Reflect’s “chat with my results” mode. The first draft often needs editing: the writing style may be uneven, and some note content can be missing. The user improves the draft through iterative prompts—asking for more information, adding sources, and requesting rewrites—until the output better matches the captured highlights. The result is a short, readable first draft that includes the desired sections and avoids irrelevant material the user didn’t highlight.

Finally, the draft can be saved as a new note (tagged as an article draft) and refined further inside Reflect. Even with multiple iterations, the end-to-end process—from outline to research to draft—takes roughly 30 minutes in this demonstration. The takeaway is less about publishing a finished piece immediately and more about using AI chat as a fast research-to-draft engine that turns curated notes into a coherent article skeleton ready for human polishing.