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
People who use ChatGPT to help with writing tasks are more productive and produce higher-quality work than those who don't, a study found.
Context & Ripple Effects
This study lands mid-arc in ChatGPT's move from curiosity to workplace tool. Coverage from December 2022 documented early adopters already drafting business emails and class work with GPT-3-era bots, and by June 2023 outlets were reporting that LLM tools had replaced some marketing and social media writers outright. What was missing was hard measurement of whether assisted writing is actually better, not just cheaper.
The new numbers — 40% faster completion and an 18% higher assessor score on tasks like press releases and reports — give employers a defensible case for adoption, and give the displacement trend a productivity justification rather than a pure cost-cutting one.
First-order effects
- Writers doing routine professional prose (press releases, reports) get a measured speed-and-quality edge from ChatGPT assistance, strengthening the argument that unassisted drafting is now the slower, lower-scoring option.
- Employers weighing LLM rollout gain their first quantified productivity baseline, which makes the earlier cost-driven replacements of marketing and social media writers look less like an outlier bet.
Second-order effects
- Rival assistants face a benchmark problem: as later testing showed, [[a:847219|ChatGPT still led Gemini-powered Bard across most task categories but by a narrowing margin]], so vendors must compete on measurable task-level gains like these rather than demos.
- Pricing power shifts toward whoever can prove per-task value — the same logic that turns an assistant into a priced line item in knowledge-work budgets.
Third-order effects
- If assisted drafting becomes the default, the human role migrates toward prompting, judgment, and editing — a shift whose cognitive costs are only starting to be measured, as the later MIT finding that [[a:887009|ChatGPT-assisted SAT essay writing produced the lowest brain engagement of any method tested]] suggests.
- Sustained measured gains would push routine business writing toward a commoditized, tool-mediated layer of the labor market, with differentiation concentrated in oversight and domain expertise.
The trend: Generative writing assistants are being re-evaluated from novelty to measured productivity infrastructure, with quality and cognitive-cost studies now setting the terms of enterprise adoption.