Writing Drops to 21% of ChatGPT Requests, with Practical Guidance Taking the Lead

Data released by OpenAI on October 26, 2026, showed that writing—the main reason for opening ChatGPT in mid-2024—now accounts for about a fifth of requests. Writing tasks dropped from 34.1% of consumer requests in July 2024 to 21.4% in September 2026, according to the third release of "OpenAI Signals" data.

Practical guidance—which includes "how-to" advice, tutoring, ideas, and health and fitness questions—rose from 26.1% to 28.1%, remaining the largest category every month since March 2025. Coding, mathematics, and data analysis assistance dropped from 13.5% to 4.6% of requests, while image and other media requests more than quadrupled from 2.0% to 8.8%.

The share of requests categorized under work fell from 51.6% to 32.2%, while the share of users with female-leaning first names rose from 42.2% to 51.8%. Meanwhile, the share of the 18 to 24 age group dropped from 39.3% to 30.6%, and the share of those aged 65 and older tripled from 0.9% to 2.7%.

The data is based on an automated classification of one million consumer messages per month from July 2024 to September 2026, excluding enterprise messages, Codex data, and users who deleted their accounts or opted out of training data usage. No human reads the messages; rather, a large language model classifies them according to written instructions, with statistical noise added to protect privacy.

What do these terms mean?

Large Language Model: An artificial intelligence program trained on massive amounts of text to understand and generate language.

OpenAI Signals: A data initiative by OpenAI's economic research team that publishes categorized statistics on ChatGPT usage.

Differential Privacy: A technique that adds random noise to published statistics to prevent the identification of any individual request.