heise+: AI in music: When agents compose themselves

According to an analysis conducted by the company Pangram, LinkedIn has emerged as the social media platform with the highest proportion of text generated by artificial intelligence. The study found that more than 40 percent of longer posts on the platform were classified by LinkedIn's detection system as fully AI-generated. Despite representing only about a third of all analyzed posts, LinkedIn accounted for nearly two-thirds of all AI-detected content identified in the research. Pangram’s findings were based on the analysis of over one million posts voluntarily shared by users who installed the Pangram Chrome extension between April and July 2026. These included posts with at least 50 words from several platforms including LinkedIn, Reddit, X, Medium, and Substack. However, the reliability of such AI detection methods remains controversial. While Pangram claims its system has a very low false alarm rate, there is currently no independent verification of these results. One reason for the high prevalence of AI-generated content on LinkedIn appears to be the platform’s integration of generative AI into multiple aspects of user interaction. Unlike many other social networks, LinkedIn offers native support for AI assistance in various areas, such as helping users craft profile texts, generating content for company pages, or revising posts. This makes the barrier to entry for AI-generated text relatively low compared to other platforms. Additionally, psychological factors seem to play a role. Researchers Mona Birjandi and Derya Oktay have studied how generative AI facilitates professional self-presentation on LinkedIn. Their findings suggest that after the introduction of ChatGPT, the number of published posts increased significantly. According to the researchers, AI seems particularly effective in showcasing apparent competence and supporting personal branding. This type of content distinguishes LinkedIn from platforms like Reddit or X. In response to this situation, LinkedIn has announced plans to recognize and lower the visibility of AI-generated posts within its feed algorithmically. The nature of the content posted on LinkedIn likely contributes to the higher presence of AI-generated material. LinkedIn is characterized by longer, specialized posts often related to management, career development, projects, or industry trends. Such texts typically follow similar structures and maintain a professional tone, which large language models can generate or revise effectively. In contrast, platforms like Reddit or X see a lot of content created through direct interaction among users. Humor, spontaneous responses, insider knowledge, or references to previous comments play a larger role here. Although these forms of content can also be produced using AI, they tend to feel less natural when generated artificially. Furthermore, LinkedIn has made public the criteria used to evaluate posts in its feed. Factors beyond the number of likes or comments include dwell time—the amount of time users spend reading a post—as well as signals indicating professional relevance and the likelihood of engagement. As a result, it is relatively easy for users to tailor their content to align with these criteria. Large language models can produce texts that incorporate typical success indicators: structured paragraphs, concise headings, lists, or personal introductions. While it is not yet proven whether generative AI leads to greater reach, known ranking criteria still lead to systematically optimized content. The Pangram analysis also suggests that generative AI is preferred for longer texts across all platforms. Across all platforms, the evaluation categorized more than one in four posts with over 250 words as being generated by AI.