AI Words Finder
Paste your text to spot words and phrases that show up a lot in AI writing, with a plainer swap for each one. Free, and it runs entirely in your browser.
These words are common in AI writing, but people use them too. A high count is not proof of AI.
Your text, highlighted
What we found, by category
Counts are based on simple word matching. They don't understand context, so a match can be perfectly natural in your sentence.
Want an overall estimate? Our full detector checks the whole text, not just word choice.
Try the AI detector →What are "AI words"?
Chatbots tend to reach for the same words and phrases again and again. Wikipedia's editors, who review a lot of AI-written drafts, keep a public field guide called Signs of AI writing that lists many of them. It includes words like "crucial," "pivotal," "underscore," "showcase," "tapestry," "testament," and "vibrant," phrases that inflate importance like "stands as a testament" or "plays a pivotal role," and leftovers from chat conversations like "I hope this helps."
This tool groups what it finds into four buckets:
- Overused words: single words that appear far more often in AI text than in everyday writing.
- Puffery and significance: phrases that tell you something is important instead of showing why, like "a key turning point" or "nestled in the heart of."
- Formulaic and chatbot phrases: stock transitions ("Moreover," "In conclusion") and things a chatbot says to the person it's talking to.
- Structural tells: patterns like "not just X, but Y," heavy use of em dashes, leftover Markdown symbols, and "despite these challenges" endings.
Why does AI write like this?
Language models predict likely next words based on huge amounts of text. Wikipedia's guide explains that this pulls them toward safe, generic, positive phrasing that could fit almost any topic, and away from the specific, unusual details a person with first-hand knowledge would include. The guide also notes that the favorite words change over time as models are updated. "Delve," for example, was heavily overused in 2023 and early 2024 and became much less common later.
How to use the results
Treat the highlights as an editing checklist, not a verdict. A few tips:
- Look for clusters. One "crucial" means nothing. A paragraph with five of these words is more likely to read as machine-made.
- Swap or cut. Many of these words can simply be deleted. "It's important to note that the bus is late" is just "The bus is late."
- Add specifics. Replacing "a vibrant community" with what actually happens there does more for your writing than any synonym.
- Don't judge others by it. Wikipedia's editors stress that not all text with these signs is AI-generated. Formal writers, students following a template, and people writing in a second language use many of these words naturally. Research has found AI detectors flag non-native English writers far more often (Liang et al., 2023, "GPT detectors are biased against non-native English writers," Patterns).
For a fuller picture, read our guide on the signs of AI-written text and why AI detectors aren't proof.
FAQ
Does using these words mean my text was written by AI?
No. These words and patterns show up more often in AI writing, but people use them too, especially in formal, academic, or business writing. A high count is a reason to take a second look, not proof.
Where does the word list come from?
It's adapted from Wikipedia's "Signs of AI writing" guide, written by Wikipedia editors (WikiProject AI Cleanup), plus phrases we've seen often in chatbot output. Our adapted list and swap suggestions are shared under CC BY-SA 4.0.
Is my text uploaded or saved?
No. The AI Words Finder runs entirely in your browser. Your text isn't sent to our server or stored, and the tool keeps working if you go offline after the page loads.
What does "per 100 words" mean?
It's how many matches appear for every 100 words of your text. It lets you compare a short email and a long essay fairly.
How is this different from the AI detector?
This tool only counts specific words and patterns. Our main detector also runs AI-detection models on the whole text and gives an overall estimate. Neither one is proof on its own.
Sources and license
- Wikipedia contributors (WikiProject AI Cleanup), "Wikipedia:Signs of AI writing." https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
- Liang, Yuksekgonul, Mao, Wu & Zou, "GPT detectors are biased against non-native English writers," Patterns 4(7), 2023. https://arxiv.org/abs/2304.02819
- Creative Commons, Attribution-ShareAlike 4.0 International. https://creativecommons.org/licenses/by-sa/4.0/
The word and pattern list used by this tool is adapted from Wikipedia's Signs of AI writing guide by WikiProject AI Cleanup and its contributors, used under CC BY-SA 4.0. Our adapted list and swap suggestions are shared under the same license.