building a reddit writing copilot for non-native english speakers (with a weird 10-step humanization framework)
been researching reddit ai tools for the last few weeks and honestly almost all of them fail for the same reason:
they still sound like ai 😭
reddit users detect it instantly
too polished
too structured
too “helpful”
too corporate sounding
and then boom:
downvotes or bans
english isnt my first language so ive personally struggled with this for months
every time i wanted to post/comment on reddit i had to:
copy -> open chatgpt -> rewrite -> paste back into reddit
got tired of it and started building a chrome extension that works directly inside reddit itself
NOT trying to make:
- auto spam bots
- mass posting automation
- engagement farming garbage
more like:
a writing copilot for people who know what they wanna say but cant always phrase it naturally in “reddit english”
the interesting part:
instead of just prompting chatgpt with:
“write casually”
im experimenting with a weird 10-step humanization framework i havent really seen other tools use
not gonna fully reveal it rn because thats basically the core idea 😅
but it focuses on things like:
- imperfect pacing
- disagreement patterns
- subreddit rhythm
- emotional variance
- shortening logic
- reaction-based writing
- intentional randomness
- avoiding balanced essay structures
- selective lowercase / punctuation behavior
- conversational friction
basically trying to simulate how actual redditors type instead of how ai thinks humans type
current pricing idea:
free:
20 comments/day
pro:
$5/mo
BYOK on all plans:
- openai
- claude
- gemini
- groq
- openrouter
- maybe local/opensource models later
main reason for BYOK:
- keeps costs low
- users can choose whatever model they trust
still validating this rn before building deeper
curious:
do you think “humanization” is actually a strong enough wedge here?
or does reddit just fundamentally hate ai-assisted writing no matter how good it gets?
