Em dash

2026-02-15 · Source: Simon Willison's Weblog · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Fundamental Awareness, quick

Summary

The author addresses accusations of using Large Language Models (LLMs) to generate blog content, asserting that their writing style generally lacks an "LLM smell." A specific exception is noted: the automated insertion of em dashes into posts. This functionality, implemented via a Python code snippet `s = s.replace(' - ', u'\u2014')`, has been part of the author's blog infrastructure since at least 2015. This predates the widespread public awareness and use of advanced LLMs, originating from a blog port from an older Django version to a new GitHub repository.

Key takeaway

For content creators concerned about their writing being mistaken for AI-generated text, review your publishing pipeline for automated stylistic modifications. Understanding the origins of such features, like the em dash insertion from 2015, can provide evidence against LLM accusations and maintain your authentic voice.

Key insights

Automated text processing features can inadvertently mimic LLM-generated content characteristics.

Principles

• Code history reveals feature origins.
• Style choices predate LLM prevalence.

Method

The method for inserting em dashes involves a simple string replacement: `s.replace(' - ', u'\u2014')` within the blog's templating system, dating back to a 2015 Django port.

In practice

• Review code for legacy text processing.
• Document stylistic automation.

Topics

• LLM Content Generation
• Blog Development
• Python String Manipulation
• Django Framework
• Em Dash Formatting

Code references

• simonw/simonwillisonblog

Best for: Software Engineer, General Interest

Related on AIssential

• Do we really need to detect LLM-generated text? — Masked word prediction by LLMs can distinguish AI-generated text from human text without explicit training.
• Analysing Differences in Persuasive Language in LLM-Generated Text: Uncovering Stereotypical Gender Patterns — LLMs exhibit pervasive gender-stereotypical persuasive language biases across models and languages, reflecting societal norms.
• Content for Content’s Sake — LLM-generated content is subtly altering human language and communication, eroding trust and overwhelming digital systems.
• I don't understand AI. How does it work? — LLMs predict text by mapping words to semantic vectors and learning statistical relationships from vast training data.
• AISLOPOPEDIA: The Complete Encyclopedia of AI-Generated Slop — AI-generated text often exhibits predictable, overused linguistic patterns that diminish originality and impact.
• It’s not just one thing — it’s another thing — A specific sentence structure is now a strong indicator of AI-generated corporate communications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.