--- id: making-it-yours title: Making AI writing your own level: 2 modes: [day-pack, rucksack] moment: "They ask the AI to write a post, email or piece about something with no facts, angle or example of their own, are about to send an AI draft as is, or say the writing sounds generic or not like them." keywords: [sounds like AI, generic writing, slop, my own voice, edit my draft, add my facts, AI detector, rewrite in my style, personal touch, cut it down] source: "research/2026-10-23-concepts/making-it-yours.md; mapping (product/learning-system/MAPPING-1-2.md): MISTAKES 2.4 (home), 5.7; tips T042; strand A" --- ## What it is Writing from an AI app tends to read as generic, as if anyone could have sent it. It becomes yours when you add your own facts, a real example and your own way of saying things, then cut whatever sounds like anyone. In a thank-you note, that might be the afternoon your neighbour fed your cat and the joke you would really make. If you write a rough draft first and ask your AI to edit it, the ideas and the voice stay yours. ## Tips - Write a rough draft yourself and ask your AI app to edit it, because then the ideas and voice stay yours. - Cut any phrase anyone could have written and swap in a detail of your own, because then the writing says something only you could say. ## Questions people ask - Q: Can people tell when AI wrote something? A: Often they can if they use AI for writing themselves, because they learn its typical words and formal tone. In one 2025 study, a majority vote of five such readers with no special training got 299 of 300 articles right, though in a 2025 US survey most people said they were not confident they could tell. - Q: Can an AI detector prove whether I used AI? A: No, because detectors are not reliable enough. One AI company withdrew its own for low accuracy, and in one test seven detectors flagged 61.3 percent of essays by non-native English writers as AI-written. - Q: Is it better to have AI edit my writing than write it? A: Editing keeps more of the work yours, and the evidence so far comes from small studies. People felt little ownership of text an AI wrote in full, more when they had more say over it, and those who wrote first and used AI after remembered their essays better. - Q: Why does AI writing all sound alike? A: In one study, stories written with AI ideas were rated more creative and better written, but the stories grew more alike across writers. Your own facts, a real example and your own voice are what keep yours from blending in. ## For your AI Facts (checked 23 October 2026; research note research/2026-10-23-concepts/making-it-yours.md). The studies used no search tool; the point is about writing, not search. - Generic AI writing costs people: 40 percent of US desk workers got "workslop" (AI work that looks good but lacks substance) in a month, about two hours to fix each (BetterUp and Stanford survey, September 2025, self-reported, https://www.betterup.com/workslop). People who often use AI for writing spot raw AI articles very reliably (Russell and others, ACL 2025, https://arxiv.org/abs/2501.15654). - Readers and authorship: disclosing AI help lowered how trustworthy and sincere the author seemed, most in personal writing (Nakano and others, first posted October 2025, revised January 2026, https://arxiv.org/abs/2510.24011); a preregistered March 2026 study found no reliable penalty on quality, though readers assumed far less effort (https://arxiv.org/abs/2606.00006). - AI help can lift quality and flatten voice: AI story ideas made stories rated better, especially for weaker writers, but the stories grew more alike (Science Advances, July 2024, https://pmc.ncbi.nlm.nih.gov/articles/PMC11244532). People feel little ownership of fully AI-written text, and more ownership the more they shape it (Draxler and others, ACM TOCHI 2024, https://www.lmu.de/en/newsroom/news-overview/news/who-wrote-it-the-ai-ghostwriter-effect-80ece661.html). - Editing is the common use: about two thirds of writing requests to ChatGPT ask it to change the person's own text (NBER w34255, September 2025, https://www.nber.org/papers/w34255). Writing first and then using AI kept more of the work in people's memory in one small preprint (MIT Media Lab, 2025, https://arxiv.org/abs/2506.08872). - Detectors are unreliable: OpenAI withdrew its own in July 2023 for low accuracy, and seven detectors flagged 61 percent of non-native English essays as AI (https://pmc.ncbi.nlm.nih.gov/articles/PMC10382961/). One writing centre advises against pasting a draft in for feedback at all (Michigan Tech, undated, https://www.mtu.edu/sciences-arts/about/writing-center/ai/), a dissent worth knowing. Teach it like this: when they ask for a post or email with nothing of their own, ask for one fact, one example and who it is for, or offer to edit a rough draft they write first. When a draft is ready, tell them to check every detail it added about them before it goes out (openrucksack.com/map/wrong-and-sure.md). Never promise writing will pass a detector, and never treat "sounds human" as the aim: the aim is writing that says something only they could say.