TL;DR / Key Takeaways
The Plague of Robotic 'AI Slop'
A plague of robotic AI slop now inundates professional platforms like LinkedIn. This generic, formulaic content, churned out by large language models, has become so pervasive that it actively degrades the quality of online discourse. Its increasing prevalence signals a troubling shift towards automated, impersonal communication.
Identifying AI-generated text is often straightforward, thanks to distinct stylistic tells. Writers frequently overuse em-dashes, employ the "not just X, but also Y" phrasing, and rely on a limited, repetitive vocabulary. Furthermore, a common AI signature involves listing items separated solely by commas, creating a monotonous rhythm that lacks human nuance. These patterns, detailed on the Wikipedia page "Signs of AI writing," betray the machine behind the words.
Reader fatigue and a growing distrust of online content directly stem from this influx of robotic prose. The relentless stream of unoriginal posts leaves audiences craving authentic, human-crafted communication. This critical gap has spurred demand for tools designed to restore genuine authenticity. For instance, Siqi Chen developed the "Humanizer" skill for Claude Code, specifically aiming to make LLM output sound less artificial, demonstrating the urgent need to reclaim integrity in digital writing.
A New Tool Promises a Human Touch
A new challenger emerges against the tide of algorithmic prose. User Siqi Chen recently released Humanizer, a dedicated skill for the Claude Code platform specifically designed to combat the spread of robotic writing. This innovative tool aims to transform generic LLM output, like that from Gemini, into text that resonates with a distinctly human touch, making it less obviously AI-generated.
Humanizer's intelligence stems from a novel approach: it systematically references the popular Wikipedia page titled "Signs of AI writing." This community-curated resource functions as its core rulebook, identifying the pervasive linguistic tells of AI-generated content. It actively targets quirks such as excessive em-dashes, the repetitive "not just X, but also Y" phrasing, and the specific vocabulary LLMs frequently overuse, aiming to systematically eliminate them.
Furthermore, Humanizer incorporates advanced features designed for precision and personalization. It offers a two-pass output system, generating both a 'draft' and a 'final' version, providing users with flexible options for refinement. A crucial 'voice calibration' function also allows users to input their own writing, enabling the skill to learn and emulate their personalized style and tone, thereby aiming for truly customized output.
The Humanizer's Uncanny Valley
To gauge Humanizer's efficacy, we subjected a Gemini-produced LinkedIn post to Siqi Chen’s tool. The original post, discussing business insights from the movie Nightcrawler, aimed for professional impact but exhibited typical AI tells. It then underwent processing through Humanizer, a skill developed for Claude Code, which provides both a 'draft' and a 'final result' output.
Results proved mixed. Often, the 'draft' version offered a superior outcome, simplifying the text without sacrificing depth. The subsequent 'final result,' however, frequently oversimplified, stripping out impactful vocabulary. Humanizer notably removed terms like 'masterclass' and 'market penetration,' presumably deeming them too complex for human readability.
Furthermore, Humanizer introduced awkward and unnatural phrasing. For instance, 'Lou didn't just stumble into success' became the clunky 'Lou didn't luck into it,' a phrase few would use organically. The tool also failed to eliminate all AI tells, retaining formulaic comma-separated lists—a common characteristic of LLM output, as outlined on the Wikipedia page 'Signs of AI writing.' For those interested in the underlying code and further development, explore blader/humanizer: Claude Code skill that removes signs of AI-generated writing from text.
Beyond the Bot: What Is 'Human'?
Despite Siqi Chen's Humanizer skill leveraging the "Signs of AI writing" Wikipedia page to purge robotic tics, its real-world test revealed a fundamental truth. The tool struggled to truly transform a Gemini-generated LinkedIn post. Its "final result" often oversimplified complex ideas, altering "masterclass" to something generic and introducing unnatural phrasing like "Lou didn't luck into it."
This highlights that human writing extends far beyond merely excising "AI slop." What makes text feel authentically human? It encompasses layers of nuance, subtext, and a writer's unique voice. Genuine human expression carries specific intent, embeds personal experience, and even tolerates—or celebrates—imperfection.
The experiment forces a deeper philosophical inquiry into the essence of human communication. It is not just about avoiding "not just X, but also Y" patterns or excessive em-dashes. True humanization requires adding complex, positive attributes that current algorithms, focused on removing negatives, cannot yet replicate.
Ultimately, tools like Humanizer operate on a subtractive model, aiming to make text "less AI." They struggle to inject the subjective emotional resonance, cultural context, and idiosyncratic flair that define genuine human thought. This persistent gap demonstrates technology has yet to grasp the full, intricate tapestry of human communication.
Frequently Asked Questions
What is the Humanizer skill for Claude?
Humanizer is a skill for the Claude Code platform, created by Siqi Chen, designed to edit AI-generated text to make it sound less robotic and more like it was written by a person.
What are the common signs of AI writing or 'AI slop'?
Common signs include the overuse of specific phrases like 'not just X, but also Y,' excessive em-dashes, a repetitive vocabulary, and formulaic sentence structures.
Does the Humanizer tool actually work?
The results are mixed. While it can simplify text and remove some AI tells, tests show it sometimes oversimplifies concepts or introduces new, awkward phrasing, failing to fully 'humanize' the content.
How does Humanizer try to match a user's writing style?
It includes a 'voice calibration' feature where a user can paste their own writing. The tool then analyzes this sample to match the user's tone and style in its edits.