Russell Brandom highlights in a recent TechCrunch article that Wikipedia’s guide to “Signs of AI writing” is one of the most helpful resources for spotting AI-generated text. As AI writing becomes more common among the general public, it is getting harder to tell it apart from human writing. He also notes that Wikipedia editors have been working on Project AI Cleanup since 2023 to handle AI-generated submissions.

I asked Claude AI to summarize the “Signs of AI writing,” then edited it to make it clearer and shorter:

• AI-generated text has spread across Wikipedia at a scale that led editors to launch Project AI Cleanup. From this effort came a simple public guide based on real examples from articles and drafts. The guide shares two key points. First, these signs do not always mean the text is AI-generated, since AI tools learn from human writing, including Wikipedia. Second, the list shares patterns, not strict rules. It also highlights an important point: automated detection tools are not very reliable.
• Inflating importance: AI writing often spends too much time explaining why a topic matters. It uses general phrases like “a pivotal moment” or “part of a broader movement.” It may also connect small details to bigger ideas in ways that feel forced or repetitive.
• Tailing clauses and superficial analysis: The guide points out a common pattern where sentences end with vague claims about importance. For example, phrases like “emphasizing the significance” or “reflecting the continued relevance” appear often. Once you notice this pattern, it becomes easy to spot.
• Promotional and marketing language: AI text often sounds like marketing copy. Descriptions are vague and overused. Places are “scenic,” views are “breathtaking,” and everything feels polished and modern. As Wikipedia editors note, it can read more like an ad than an encyclopedia entry.
• Negative parallel structures: AI writing often uses contrast phrases to make a point, such as “not only,” “but,” or “however.” This can make the text feel like it is correcting the reader or setting up a contrast that is not needed.
• Em dashes: While human writers use em dashes, AI tends to use them more often and in a patterned way. They often appear where a comma or simple break would work better, which can make the text feel forced.
• Formatting issues: Some problems are easy to spot, like too much bold text, broken formatting, or odd citation styles. These surface issues can point to deeper problems, so fixing only the visible parts may not be enough.
• A note on individual models: Each AI model has its own writing style. What is common in one model may not appear in another. These signs are best used together, not as a strict checklist.

Brandom notes that these patterns come from how AI models are trained and used. They can be hidden, but they are hard to remove completely. Wikipedia’s guide stands out because it is based on real examples at scale, not just theory. It offers a practical way to review, edit, and improve writing.

Blake Stackton sums it up simply: AI writing often slips into a promotional tone, even in B2B content, and that tone is hard to fix. It is more effective to guide the AI with clear, specific edits than to try to limit it too much.

Acknowledgement: This summary was generated using AI writing tools as part of my AI Explorations process. For more information, please refer to the original articles listed below.

Related Resource Links

• The best guide to spotting AI writing comes from Wikipedia | Tech Crunch
• Signs of AI writing | Wikipedia
• Don’t Write Like AI: 10 Takeaways from Wikipedia’s Signs of AI Writing | Blake Stockton