Researchers presented participants with six pairs of images, two pairs of videos, and two pairs of poems, asking them to determine which content was created by artificial intelligence and which by a human. According to VTsIOM analysts, text proved to be the most difficult format: only 34% of respondents correctly identified the authorship in both pairs of four-line poems, 48% gave one correct answer, and 18% failed to identify either correctly.

Overall, Russian internet users correctly identified AI-generated content in six out of ten cases. Among the six pairs of images, only 4% of respondents correctly identified the authorship of all images. Around 70% distinguished between three and five AI-generated images and real ones, while 25% identified only one or two correctly. Just 1% did not give a single correct answer. For video content, 59% of respondents completed the task without errors.

According to an earlier VTsIOM survey conducted in October, every second Russian used neural networks at least once during 2025. AI tools were most often used for information search (63%), creating music, images, and text (38%), and data analysis and task automation (36%).

A study by U.S. SEO company Graphite showed that in May last year, the volume of AI-generated articles on the internet briefly exceeded the number of human-written texts, reaching 52% versus 48%. The figures later converged. The analysis was based on 65,000 web pages from the open Common Crawl database covering the period from 2020 to 2025. The sharp rise in AI-generated content began after the launch of ChatGPT in 2023, peaking in November 2024.

In May 2024, Anton Nemkin, a member of the State Duma committee on information policy, said that lawmakers were already working on a bill to label AI-generated content. He suggested that such labeling should use visual markers or watermarks, adding that companies like Sber and Yandex should be among the first to implement these technologies.