The professional networking platform LinkedIn has long been a space for career advice, industry insights, and self-promotion. But a new study suggests that a significant portion of the content posted there is not written by humans at all. According to research conducted by Pangram, a company behind a prominent AI text detection tool, 41% of longform LinkedIn posts are flagged as fully AI-generated, along with 30% of short-form content. The numbers are eye-popping, especially for a platform that prides itself on authenticity and professional credibility.

Pangram’s data is derived from its Chrome extension, which scans text as users browse the web and flags content likely produced by large language models like GPT-4. The extension aggregates anonymized data, giving Pangram a unique — if potentially biased — window into the prevalence of AI-generated writing across social media. The company acknowledges that its tool may have an interest in finding high rates of AI content (since it sells detection services), but independent experts have noted that the figures are plausible given the rapid adoption of generative AI in content creation.

The Scale of AI on LinkedIn

LinkedIn has become a hotspot for AI-written posts. The platform’s culture of sharing inspirational stories, career tips, and thought leadership lends itself well to AI-generated text, which can efficiently produce generic yet polished prose. Pangram’s study indicates that nearly half of all longform posts — those exceeding a certain word count — are entirely computer-written. Short-form posts, such as status updates, are not far behind at 30%. This saturation is far higher than on other platforms like X (formerly Twitter), where 29% of longform content is fully AI-generated, though the hybrid human-AI rate is even higher there: only 53.2% of X articles are flagged as fully human-authored. Normal X posts — the short tweets for which the platform is known — are only 9% AI-generated, likely because of their brevity and conversational nature.

Medium, a blogging platform often considered a social network, showed 31% of longform content as fully AI-generated. Reddit, where discussions are more community-driven and personal, reported 13% AI-generated longform content and just 3% for short comments. Substack, a newsletter platform, had the lowest longform AI rate at 10%, though its short-form content (e.g., notes) was slightly higher at 12%. The variation suggests that platforms with professional or thought-leadership cultures are more prone to AI-generated text, whereas platforms emphasizing authentic personal voices, like Substack or Reddit, retain more human-written content.

Why LinkedIn Is So AI-Saturated

LinkedIn’s high AI content rate can be traced to its user base and incentives. Many users post to build a personal brand, attract recruiters, or establish expertise. AI tools make it easy to generate lengthy posts about leadership, productivity, or industry trends without investing time in original writing. Some users even repurpose AI-generated articles from other sources, adding slight modifications to avoid detection. The platform’s algorithm tends to reward frequent posting, further encouraging the use of AI. A recent trend piece in the New York Times asked, “Was LinkedIn getting more interesting?” The article noted a surge in engagement and variety, but also raised the question of authenticity. If using AI counts as inauthentic behavior — and many would argue it does — then LinkedIn’s growing interestingness may come at a cost of trust.

Pangram’s study does not distinguish between AI used for drafting and AI used for full generation. Users might write a few sentences and then ask an AI to expand them, resulting in a human-AI hybrid. The tool’s flagging may not capture these nuances perfectly. Nonetheless, the sheer volume of fully AI-generated content suggests that many users are comfortable presenting AI work as their own, which could erode the value of the network. Recruiters and hiring managers who browse LinkedIn for authentic insights may find themselves sifting through boilerplate advice.

Comparisons Across Platforms

The study also highlights how different social media cultures shape AI usage. On X, the longform posts are often threads or articles shared via the platform’s note feature. The high hybrid rate — nearly half of all posts are partially AI — indicates that users may rely on AI for structure but add personal touches. On Reddit, the low AI rate reflects the community’s emphasis on original, often niche content; AI-generated posts are quickly downvoted or called out. Substack’s relatively low AI rate is likely due to its newsletter model, where readers subscribe for a specific human voice. Interestingly, Substack’s short-form notes show a higher AI rate, possibly because they are less scrutinized than full newsletters.

Pangram’s methodology has limitations. The Chrome extension may have a sample bias — users who install an AI detector might be more likely to browse AI-heavy content — and the detection algorithm may produce false positives. However, the consistency across platforms and the alignment with anecdotal observations make the findings credible. The study serves as a wake-up call about the scale of AI-generated content in our daily feeds.

For LinkedIn users, the question is: What does it mean to be authentic in a world where AI can mimic professional insights so convincingly? Some argue that using AI is simply a tool, like spell-check or grammar assistants. Others contend that it represents a breach of trust, especially in a professional context where personal experience and expertise are valued. The Pangram study does not answer that question, but it provides a crucial data point: AI is no longer a fringe phenomenon on social media — it is the new normal.

Source: Gizmodo News