ChatGPT Prompts for Professional Writing

Who created and checked this page

Correction history

Every accepted correction to this page is recorded with the exact change, so readers can see how the page improved over time.

1. 19 July 2026 · corrected by Emir Baycan

   Fabricated study: no such 'MIT Media Lab' 2023 analysis of 15,000 ChatGPT-generated professional documents could be found. MIT Media Lab's real 2023-2025 ChatGPT-related research covers emotional/affe; Unverifiable quote attributed to Andrej Karpathy's 'State of GPT' talk at Microsoft Build 2023. Extensive search of talk transcripts, slides, and summaries could not locate this quote.; Unverifiable quote attributed to Shane Parrish of Farnam Street. Web search of Parrish's published writing, podc

   Before

   A 2023 analysis by the MIT Media Lab of 15,000 ChatGPT-generated professional documents submitted to the research identified the patterns that most reliably mark text as AI-generated. [...] <blockquote> <p>&quot;The most interesting work with language models is not the default output. It is what you get when you push against the default with enough specificity that the model starts predicting the voice you described, not the voice it drifts toward.&quot;</p> <p>Andrej Karpathy, researcher, in a 2023 talk on language model behavior</p> </blockquote> [...] <blockquote> <p>&quot;The model is a mirror. What it gives you back depends on what you ask it to reflect. Writers who know what they want get what they want. Writers who do not know what they want get the average.&quot;</p> <p>Shane Parrish, writer at Farnam Street, on AI and writing</p> </blockquote>

   After

   Writers and editors who work with AI-generated drafts have identified recurring patterns that commonly mark text as AI-generated, based on informal observation rather than a single controlled study. [...] <p>The most interesting work with language models is rarely the default output. It is what a writer gets when they push against the default with enough specificity that the model starts predicting the voice they described, rather than the voice it drifts toward by default.</p> [...] <p>A language model behaves something like a mirror: what it gives back depends heavily on what a writer asks it to reflect. Writers who know what they want tend to get closer to it. Writers who do not know what they want tend to get the average.</p>

   Why: Fabricated study: no such 'MIT Media Lab' 2023 analysis of 15,000 ChatGPT-generated professional documents could be found. MIT Media Lab's real 2023-2025 ChatGPT-related research covers emotional/affe; Unverifiable quote attributed to Andrej Karpathy's 'State of GPT' talk at Microsoft Build 2023. Extensive search of talk transcripts, slides, and summaries could not locate this quote.; Unverifiable quote attributed to Shane Parrish of Farnam Street. Web search of Parrish's published writing, podcasts, and interviews could not locate this quote or anything close to it.

   View the full record →

Full contribution timeline

1. Correction 19 July 2026

   Emir Baycan: Fabricated study: no such 'MIT Media Lab' 2023 analysis of 15,000 ChatGPT-generated professional documents could be found. MIT Media Lab's real 2023-2025 ChatGPT-related research covers emotional/affe; Unverifiable quote attributed to Andrej Karpathy's 'State of GPT' talk at Microsoft Build 2023. Extensive search of talk transcripts, slides, and summaries could not locate this quote.; Unverifiable quote attributed to Shane Parrish of Farnam Street. Web search of Parrish's published writing, podc

   Proof record →