ChatGPT learns users’ writing styles to make AI text less robotic

OpenAI is adding a capability that allows ChatGPT to better mimic a user’s personal writing style, aiming to make AI-generated text sound less generic and more natural.

The feature is designed to help ChatGPT go beyond producing grammatically polished copy by reflecting the way an individual normally writes, including their preferred words, sentence structures, tone and other stylistic habits.

According to an India Today report, the capability is currently associated with ChatGPT Work and can draw insights from connected workplace apps. These include services such as Gmail, Google Drive, Slack and SharePoint, allowing ChatGPT to analyse users’ existing writing patterns.

For example, the system may identify whether a user typically writes in short sentences, repeatedly uses particular words or phrases, or follows a specific style when ending emails. It may also account for details such as capitalisation and other individual writing habits.

OpenAI also allows users to provide samples of their own writing and instruct ChatGPT to follow the same style. Previous emails, notes or other examples can be used to guide the system on vocabulary, sentence length, level of formality and overall tone.

OpenAI has yet to provide details on how broadly the feature will be rolled out to general ChatGPT users. The development comes amid growing criticism of repetitive, formulaic AI-generated writing, often described as “AI slop,” and could make AI-assisted content more closely resemble an individual writer’s voice.