Simon Willison Built an LLM Cliché Detector on Fable 5

On 17 July 2026, developer Simon Willison released LLM cliché highlighter, a free web tool that highlights about 11 stock phrases typical of neural-network-generated text.

How the tool came about

Simon Willison built LLM cliché highlighter after yet another run-in with an article stuffed with AI clichés like "no fluff, no filler, no jargon." He made the app by vibe-coding: he asked Anthropic's Fable 5 model to write all the code from a description of the task, rather than programming it by hand.

Willison is one of the most-cited writers on LLMs and the creator of the Datasette library and the `llm` command-line tool. On his blog, simonwillison.net, he regularly examines how large language models write, make mistakes, and give themselves away with characteristic turns of phrase.

"I'm tired of reading yet another article packed with the clichés of LLM-written text," —

Simon Willison, on simonwillison.net.

How does the tool work?

LLM cliché highlighter analyzes the pasted text and marks in yellow any fragments that match predefined patterns of machine writing. A screenshot from Willison's blog shows a set of 11 rules, each of which can be switched on and off individually.

• Published 17 July 2026 by Simon Willison
• Vibe-coded with Anthropic's Fable 5 model
• Highlights about 11 typical LLM cliché patterns
• Articles are loaded by URL via the r.jina.ai service
• A "Show just the highlights" mode displays only the highlighted passages

The highlighting has two levels: pale yellow marks the whole "suspicious" phrase (the flagged sentence), while a more saturated color marks the specific construction inside it (the pattern match). Example patterns include the constructions "Is real … and / not" and "Worth naming"; the demo phrase in the interface reads "That loss is real and it's worth naming," where the tool highlights two matches at once.

It also works on other people's articles: you paste an address into the URL field, the page is fetched through the r.jina.ai service, and the text is then run through the same rules. For a quick check, there is a button that loads a ready-made example, plus a mode that leaves only the highlights on screen.

Why this is useful

LLM cliché highlighter gives writers a quick way to check their own draft for the "smell of the machine" before publishing. The author pastes in the text, sees the highlighted stock phrases, and rewrites them. These are exactly the turns of phrase that modern models insert most often and that mark a text as generated.

The approach reflects a broader problem: as LLMs write more and more public content, their favorite turns of phrase become a recognizable marker. Tools like this turn the gut feeling that something "sounds like ChatGPT" into a concrete list of rules that can be applied mechanically rather than by eye.

What this means

LLM clichés are turning into a measurable checklist: what used to be a matter of an editor's taste can now be checked with one click. For writers, it is a cheap filter against machine style; for readers, it is a reminder that recognizable turns of phrase give generation away.

Frequently asked questions

What is LLM cliché highlighter?

LLM cliché highlighter is a free web tool by Simon Willison, published on 17 July 2026. It highlights about 11 patterns in a text that are characteristic of large language model writing.

How do I check an article for AI clichés?

Paste the text into the tool's field or enter the article's URL, and the page is fetched through the r.jina.ai service and run through the rule set. Matches with machine-writing patterns are highlighted in yellow, and the "Show just the highlights" mode leaves only those on screen.