Hi Product Hunt, I'm Scott Eggert, founder of GoldenDoodle AI.

I dictate most of what I write and hand it to AI to clean up. For a while my drafts came back more polished and less mine: more lists of three, and a tidy last line summing up what I'd just said. The people reading our newsletters were getting the same thing from us. When people hear AI, they stop hearing you.

So we wrote down what we want any AI to do before it writes for us, and we're giving it away. Three skills, free and open source under CC BY 4.0.

Human voice takes out the habits that make writing sound machine-made: long dashes, lists of three, stock openers and closers, and the "it's not X, it's Y" habit.

Human-centered language is the plain language standard for AI. It's an open standard, and the skill helps any AI write about people clearly and respectfully.

Trauma-informed language builds on it, for organizations whose readers may have lived through trauma.

They work in Claude, ChatGPT, Gemini and any agent that can load a skill. The install steps for each are at goldendoodleai.com/skills. Most people start with human voice: goldendoodleai.com/skills/human-voice

We test the skills in public. The Human-Centered Language Benchmark runs them on current AI models every month and publishes every draft. In the November 2026 run (completed Nov 1, 2026: 16 models, 45 everyday writing requests, 4,320 drafts), AI got people right in 72% of drafts on its own and in 99% with the free human-centered standard loaded first. Among the 14 models we tested both months, drafts that got people wrong on their own held at 31 in 100, with the skill 1 in 100 both months. Since October 2026 we have tested 8,370 drafts from 17 models. In the same run, the human voice skill took em dashes from 6.1 to 0.03 per 1,000 words. The full report and raw outputs are at goldendoodleai.com/benchmark/2026-11

One place it fell short: on its own, the human voice skill only took drafts about people that needed revision from 36 in 100 to 20 in 100 (same Nov 1, 2026 run). For writing about people, load it with one of the two standards.

I'd like to know where it falls short for you. What does your AI still do that makes your writing sound like it?