How to Humanize AI-Generated Content With Mem Smart Write and Edit

AI-generated drafts often come out sounding generic—like they were produced by a system rather than by a real person. The fix presented here is to “train” Mem’s Smart Write and Edit by feeding it a large library of writing that already reflects the creator’s natural voice, then using Mem Chat to generate a detailed writing-voice profile that Smart Write can apply to new drafts.

The process starts with importing existing work. The creator recommends pulling blog posts into Mem as markdown files—using Chrome extensions that download posts in markdown, or via an automation setup (Zap) that imports new files from a folder. A practical snag appears during import: post titles may not come through automatically, requiring manual title entry. The key principle is volume and authenticity: the more writing that has been produced in the creator’s own voice, the better the system can mimic patterns in vocabulary, sentence structure, and style.

After import, organization matters less than quantity, but collections make the workflow easier—especially when using Mem Chat. The creator uses two collections: one for imported blog posts and another for manuscripts from traditionally published books (including the full manuscript for “Audience of One”). The underlying rationale is straightforward: writing tools learn from the data and human input they’re given, and for a content creator, that data is the creator’s own text.

Next comes the voice profiling step. In Mem Chat, a prompt is used to ask for a description of the writing voice based on the content inside those collections. The output becomes a “writing voice” note—expanded into a more detailed profile that includes preferred vocabulary, sentence patterns, stylistic tendencies, and even references to specific high-performing posts. That voice note is then used as an instruction for Smart Write and Edit.

The tutorial demonstrates the difference between standard drafting and drafting with the writing voice applied. Smart Write first produces a baseline outline for a blog post (benefits of daily reflection), then the same task is re-run with the writing voice instruction, resulting in noticeable changes in tone and phrasing. Similar before-and-after comparisons are shown for a newsletter introduction tied to a podcast interview and for a social media post promoting a recent interview. In each case, the “apply writing voice” version sounds more like the creator—though not perfectly—because it reflects the vocabulary and stylistic habits captured from prior writing.

A final example applies the approach to opening lines for a video script on workflow optimization, where the voice-based rewrite also tends to adjust length and cadence. The recap boils the method down to three steps: import as much writing as possible, organize it into collections, then generate a detailed writing-voice description in Mem Chat and apply it through Smart Write and Edit. The creator emphasizes that this works best with a substantial content library; without enough source material, the voice-matching effect may be weaker.