Ruben Hassid’s Guide to Detecting and Deterring AI-Generated Writing

In a recent LinkedIn post, Ruben Hassid delves into the increasingly complex landscape of AI-generated text, sharing his personal frustrations and offering practical strategies for identifying and mitigating its use. Hassid, whose own work centers on AI, expresses a strong aversion to AI writing, even as he teaches others about the technology. He highlights a common, now-obvious trick used to spot AI writing – the avoidance of em dashes – and points out the irony that perfectly structured human writing can now be mistaken for AI output.

Identifying the Hallmarks of AI Writing

Hassid outlines nine specific patterns he has observed as “dead giveaways” of AI-generated content. He details how AI often employs predictable sentence structures and phrasing that, while grammatically correct, lack human nuance and personality. One of his key observations is the tendency for AI to use a specific rhetorical device:

“That’s not X. That’s Y.”

Hassid advises that instead of using such constructions, a more human approach would be to state the point directly, like “They’re stalling.” He also points out the AI habit of combining short, punchy phrases:

“Fast. Simple.” “No fluff. Just answers.”

According to Hassid, a real person would typically choose only one of these to convey a message effectively. He further identifies other common AI tells, including overly visual metaphors without direct advice, self-congratulatory phrases, and comparisons to familiar concepts that the reader might not understand.

Strategies for Combating AI Writing Patterns

Beyond identifying the problem, Ruben Hassid proposes a multi-step approach to actively prevent AI from producing these tell-tale writing patterns. His primary recommendation involves a specific prompt instruction:

“Use ASD-STE100.”

As Hassid explains, this refers to a writing standard used for aircraft manuals, characterized by short words and sentences, which helps to eliminate 80% of undesirable AI writing characteristics. He notes that this prompt works across various AI models, including ChatGPT, Gemini, and Grok.

Leveraging Specialized Tools and Techniques

Hassid further suggests leveraging specialized skills and custom instructions within AI platforms. He has developed a comprehensive version of the ASD-STE100 standard, available for free, that can be uploaded as a “skill” in platforms like Claude. This skill, when activated with a command like `/ste`, helps enforce stricter writing guidelines.

He also advocates for a disciplined approach to AI usage, recommending that users “turn it off for anything with a heart.” This implies using strict writing standards only for functional content like guides, explanations, and formal documents, rather than creative or emotional writing.

To reinforce good writing habits, Hassid suggests several other techniques:

• Incorporating Zinsser’s four principles of quality writing (Simplicity, Brevity, Clarity, Humanity) into custom instructions.
• Creating a “forbidden” list of the nine AI writing patterns and having the AI check drafts against it, refining the list over time.
• Starting new chat sessions frequently, as prolonged conversations can lead to a degradation in AI writing quality and repetition.

In essence, Ruben Hassid provides a detailed, actionable framework for both recognizing and actively combating the stylistic quirks of AI-generated text, empowering users to maintain a more authentic and human voice in their communications.