Study: ChatGPT helped writers complete tasks, such as writing press releases or reports, 40% faster and produce work that assessors scored 18% higher in quality
That ChatGPT is a productivity booster is indisputable. The question is whether it leads to fewer jobs as companies need fewer people or whether it leads to more opportunities in those fields? … Twitter: Peter Gasston / @stopsatgreen : Popular opinion: ChatGPT is going to make us lazy and stupid. Scientific study: [image] Jarvis Ka / @jarviska : MIT research shows that ChatGPT is a total game-changer! It turbocharges good writers, making them 40% faster, while even weaker writers produce top-notch content! But here's the big question: Is ChatGPT a job killer or a career catalyst? https://www.technologyreview.com/ ... @dwaynephillips : A study of 453 persons who write much of the day shows, “The writers who chose to use ChatGPT took 40% less time to complete their tasks, and produced work that the assessors scored 18% higher in quality than that of the participants who didn't use it.” https://www.technologyreview.com/ ... Izak Oosthuizen / @oosthuizen_izak : Exciting news from MIT! A study by 2 economics graduate students shows that #ChatGPT can bridge writing gaps in the workplace. Less experienced employees lacking writing skills can produce work on par with their more skilled colleagues #AI #OpenAIChatGPT https://www.technologyreview.com/ ... Riku Arakawa / @hciphdstudent : My words are quoted 😁 The new Science paper about productivity effects of ChatGPT is a great piece of work answering many ongoing questions. I thank the authors for mentioning our CatAlyst paper from CHI'23! https://www.technologyreview.com/ ... @mrbumpyface : You sure? This is ChatGPT butchering Ecclesiastes the observation is made that success is not solely determined by swiftness or strength, nor is it exclusively bestowed upon the wise or the wealthy. https://www.technologyreview.com/ ... Glenn Gabe / @glenngabe : MIT Study: ChatGPT helped writers complete tasks, such as writing press releases or reports, 40% faster & produce work that was scored 18% higher in quality BUT, the article also explains that ChatGPT is good at presenting false information... https://www.technologyreview.com/ ... [image] S. Stoerger / @csoleil : “I think this is a really interesting result that demonstrates how human-AI cooperation works really well in this kind of task. When a human leverages #AI to refine their output, they can produce better content.” https://www.technologyreview.com/ ... #writing #employment #highered See also Mediagazer
Context & Ripple Effects
This study is the first hard measurement behind a question that has hung over generative AI since ChatGPT and its peers were framed as a step change in AI use cases: does the technology actually make knowledge work better, or just faster-looking? The answer here is both — professional writers finished press releases and reports 40% faster, and blind assessors scored the output 18% higher in quality.
The result lands mid-debate about whether ChatGPT augments or replaces writers, and it sits alongside earlier evidence that users self-correct the technology's weaknesses: in a Wharton professor's ChatGPT-required essay classes, students quickly learned to fact-check and spot the model's accuracy and bias problems.
First-order effects
- Professional writers doing routine business writing — press releases, reports — can now produce measurably more output per hour, shifting the unit of work from drafts written to drafts reviewed and edited.
- Employers hiring for these tasks get direct evidence that one writer plus ChatGPT outperforms one writer alone on both speed and assessed quality, changing what a writing role is worth.
Second-order effects
- Because the same models can produce clean, convincing text that repeats misleading narratives — as researchers documented earlier in 2023 — the productivity gain creates a matching demand for verification and editing labor, partially offsetting headcount savings.
- Training and education providers face pressure to teach AI-assisted drafting and fact-checking as core skills, following the pattern already visible in university classrooms where students had to learn the model's limits firsthand.
Third-order effects
- If speed-plus-quality gains hold across other white-collar task categories, entry-level writing and reporting roles thin out while review, editing, and oversight roles expand — a restructuring of the occupation rather than its elimination, though which side dominates remains genuinely unresolved.
- The finding strengthens the case for vendors building task-specific wrappers around general models, the direction OpenAI later took with custom GPTs as part of its gradual iterative deployment strategy.
The trend: Generative AI is moving from novelty to measured productivity infrastructure in knowledge work, with each rigorous study converting the augmentation-versus-replacement debate into concrete labor-market decisions.