Automatically add backlinks to your notes (using AI)

Automatically turning spoken or typed notes into a web of backlinks is the core payoff: the workflow takes raw note text, identifies proper nouns (people, places, companies, projects, and similar entities), and wraps them in double square brackets so the notes system converts them into consistent, clickable backlinks. The result is that entities like “Mark Smith,” “Blackstone Coffee,” “baby bird films,” and “Winter Park, Colorado” become linked automatically—without manually hunting through the text after the fact.

The method hinges on a custom AI prompt configured inside the Reflect Notes AI editor. The prompt keeps the original text unchanged but instructs the model to surround each proper noun—covering person names, workplace names, project names, and company names—with double square brackets. It also adds formatting guidance so the editor can interpret the bracketed entities as backlink targets. Additional rules tighten accuracy: it encourages backlinks when the entity starts with a capital letter, and it explicitly tells the model not to include actions or verbs. That last constraint exists because earlier runs sometimes backlink words like “biking,” which the creator didn’t want treated as an entity.

In practice, the workflow works for both new and existing links. When the notes already contain a backlink for an entity, the system reuses it so the formatting stays consistent. When an entity doesn’t yet have a backlink, the system creates one. The transcript demonstrates this with an audio memo: after running the AI step, the text remains the same, but the entities gain “beautiful purple backlinks” once the bracketed terms are converted. The creator notes that the AI palette editor may initially display the double brackets in a “weird” formatting state, but the final replacement produces proper backlink styling.

A key advantage is timing. Backlinking is hard to do while recording audio because there’s no practical way to insert structured links mid-sentence. Manual backlinking after transcription is also tedious, and writing notes forces constant context-switching—deciding what to link while trying to capture ideas. Automating the entity detection removes that friction, making it easier to build a “second brain” where notes connect into a visual map of ideas.

The transcript also points to customization: users can clone an existing Reflect Notes prompt, change its text and name, and save it. They can further enrich created backlinks by adding context later—such as company type, logos, domains, or other metadata—while still letting the AI handle the initial linking. Over time, the accumulating backlinks generate a network view (“demo brain” in the example) that ties hundreds of notes together into a navigable structure, turning note-taking into an ongoing, low-effort habit of linking related concepts.