ReddGrow DocsRefining Your Advocate
Teach the AI your personal writing style by rewriting sample drafts — the advocate learns your voice and applies it to every future comment.
When you first create an advocate, the AI generates comments based on the persona settings you chose: tone, role, creativity, and knowledge base. Refine takes this further — you teach the AI your actual voice by rewriting sample drafts in your own words.
Why Refinement Matters
Generic AI comments often sound like AI comments. The Refine wizard closes the gap between "sounds plausible" and "sounds like me." Users who run two or three Refine sessions typically see:
- Comments that match their natural vocabulary and sentence rhythm
- Appropriate hedging or directness depending on their style
- Correct handling of product disclosure (when to mention the product, when to hold back)
- Brand-appropriate analogies and framing
How to Access Refine
Navigate to Advocates → [your advocate] → Refine (the /advocates/[id]/refine route).
The Wizard
The wizard presents 5 sample drafts the advocate recently generated — real comment drafts from your active campaigns. For each one:
- Read the draft and the Reddit post it was written for
- Rewrite the comment in your own words — write it how you would actually say it
- If a draft is fine as-is, click Looks good to accept it without changes
- Submit when all 5 are done
There is no minimum number of rewrites required — even rewriting 2–3 out of 5 gives the system useful signal.
What the AI Extracts
After you submit, ReddGrow analyzes the difference between the original drafts and your rewrites:
| What it learns | Example |
|---|---|
| Vocabulary patterns | "leverages" vs "uses"; "significant" vs "big" |
| Sentence length | Short punchy sentences vs longer explanatory ones |
| Hedging style | "In my experience..." vs "Definitely do X" |
| Disclosure approach | How often you mention the product and how directly |
| Structural patterns | Whether you lead with empathy, a question, or a statement |
These patterns become voice guidelines that are injected into every future draft for this advocate.
Running Multiple Refine Sessions
Voice guidelines accumulate — each session adds to the model's understanding of your style rather than replacing previous guidance. You can run Refine:
- After you see a batch of drafts that feel off-brand
- After major shifts in your product positioning
- When you add a new advocate for a different persona
There is no limit on sessions.
Checking the Extracted Guidelines
After a Refine session, you can view the extracted voice guidelines on the advocate's profile page under Voice. These are plain-English descriptions like:
"Use short, direct sentences. Avoid corporate buzzwords. Always disclose affiliation when mentioning the product. Prefer humor over formality."
You can edit these manually if the extraction missed something or over-fitted to one draft.
Tips for Effective Rewrites
- Write fast — first-instinct rewrites are more authentic than carefully edited ones
- Don't translate, rewrite — start from the Reddit post context, not from the AI's draft
- Include your actual disclosure style — if you'd say "disclaimer: I work on this tool", write that
- Vary the scenarios — if all 5 sampled drafts are from the same subreddit, the extracted voice will be narrow; more diverse samples give better generalizable guidelines