Worried your writing sounds like AI? These tools can help

AI-generated writing has become ubiquitous in various forms of content, from LinkedIn posts and Substack articles to tweets and medical papers. As we consume more of this content, certain patterns and quirks start to stand out, prompting some writers to seek tools that can help identify AI-generated text. Companies like LinkedIn and Substack have introduced moderation efforts, while AI detection tools such as Pangram and Turnitin assist universities and organizations in spotting AI-generated content in essays and work documents.

However, even writers who don't use AI tools are increasingly concerned about inadvertent AI traces in their work. George Strakhov, founder of the AI automation consultancy move38 and creator of the AI style linter Slop Tells, notes that AI-generated writing tends to have a distinct style that writers can learn to detect. "What’s interesting is that these models do speak in a specific way, and you develop a sense of it—you develop the smell of the writing," Strakhov explains.

To assist writers in spotting AI-generated content, designers like Strakhov have created their own interfaces to catalog and highlight key aspects of phrasing, formatting styles, and specific words that make the writing sound robotic. Two notable tools are the LLM cliché highlighter and Slop Tells.

The LLM cliché highlighter, designed by programmer Simon Willison, tracks 38 different signs of AI writing, including rhetorical tics like "The punchline is," "Here’s the twist," and "Don’t take my word for it." Users input their text, and the tool highlights flagged sentences in light yellow, with darker yellow indicating matches.

It also displays a counter at the top of the page showing the total number of flagged sentences, matches, and chain items (patterns like "no X, no Y") found within the text. While this tool provides a clear visual representation of AI-saturated text, its scope is limited, and it doesn't adapt to the evolving landscape of large language models (LLMs).

In contrast, Slop Tells, designed by Strakhov, takes a more dynamic approach. It constantly reassesses AI tells as chatbots and people evolve, ensuring that the tool remains relevant. The tool uses three piles of information to inform its model: human writing before 2022, reverse-engineered prompts from the first pile, and prompts fed through current AI models.

This process also incorporates well-regarded pre-AI author work to avoid flagging qualities of good writing. Slop Tells highlights AI jargon in pasted sections of text and provides a sense of whether the jargon is currently active or faded in models, along with a link to a compendium of information on each example. This tool offers a more granular and informed look at AI tells in a writer's work.