Muhammad Amanullah
Date: October 8, 2026 The True Post (Web News) —OpenAI has announced a new technology related to text generated by ChatGPT that could make it possible to identify content produced by artificial intelligence.
According to the company, the technology will incorporate subtle and hidden statistical signals into text. These signals would not be noticeable to ordinary readers but could be detected by specialised automated systems.
Hidden Signals
According to OpenAI, the new text-watermarking technology will make minor changes during the process of selecting words in text generated by ChatGPT. The changes would be designed so that the text continues to appear normal to an ordinary reader, while specialised technology could determine whether the content was generated using artificial intelligence.
The company has named the technology textGrain. Its basic purpose is to create a statistical marker within content generated by ChatGPT that can later be detected automatically.
Identifying AI-Generated Content
The identification of content generated by artificial intelligence has become an important issue in the technology sector in recent years. Educational institutions, media organisations, businesses and other sectors increasingly face the question of how to determine whether a piece of writing was produced by a human or by artificial intelligence.
The new technology is intended to provide one possible solution to this challenge. Under the system, an indicator would be incorporated into the text that cannot be seen by the human eye but whose presence could be examined by specialised software.
European Union
OpenAI’s initiative is also being linked to new European Union regulations concerning artificial intelligence. Under the AI Act, companies providing generative AI services have responsibilities to take various technical measures to identify content generated by their systems.
The purpose of these regulations is to provide users with greater transparency about AI-generated content and encourage the development of technical tools capable of identifying such material.
Limited Application
According to OpenAI, at this stage the text-watermarking technology is being introduced for ChatGPT-generated text within the European Union. In contrast, some other AI companies have also discussed the use of similar technologies in different regions.
Because of this limited application, the extent to which the system will be available for ChatGPT-generated text in other parts of the world remains a separate issue of importance.
User Concerns
Following OpenAI’s announcement, some users have also raised objections to the technology. Critics argue that incorporating identification signals into every piece of AI-generated text could create difficulties for users in certain circumstances.
Some users have also expressed concerns that if such technology is not completely reliable, it could result in false identification. If human-written or modified AI-generated text were incorrectly classified as having been produced by artificial intelligence, it could lead to practical consequences.
Another Company Enters the Field
Before OpenAI, Anthropic had also announced plans to incorporate similar technology into its AI model Claude. This development indicates that AI companies are increasingly focusing not only on generating content through artificial intelligence but also on methods for identifying the content produced by these systems.
As the use of artificial intelligence continues to grow, the issue could become even more important, as AI is expected to play a greater role in news, educational materials, business documents and other forms of writing in the future.
How Will It Work?
According to information about textGrain, the technology adds a statistical signal to the process through which ChatGPT selects words. In other words, the appearance and readability of the text remain intact, while a specific statistical change is introduced into the overall pattern of word selection.
A detection system can later analyse the text to determine whether the relevant statistical pattern is present. If the required signal is detected, the system could help determine whether the text was generated by ChatGPT.
The Challenge Ahead
Alongside identifying AI-generated content, another major question is how effectively such technology will be able to identify text that has been modified, translated or rewritten in a different style.
Similarly, it will be practically important to determine whether the hidden signal remains intact after users make changes to the text.
OpenAI’s announcement of textGrain points to a new direction in the artificial intelligence industry, where the important question in the future may not only be who can create content, but also how that content can be identified.
With European regulations and initiatives by major AI companies, systems designed to identify AI-generated writing could become increasingly common in the coming years.
