Un•AI•ify looks for patterns commonly found in AI-generated writing created by models like ChatGPT, Grok, Claude, Gemini, and others. LLMs use these writing styles and rhetorical patterns because of their popularity in English-language content.
Why does Un•AI•ify focus on rhetoric?
Communication is inseparable from persuasion. This has been known as far back as Aristotle who said, "It is not sufficient to know what one ought to say, but one must also know how to say it" — itself an example of both It's not X. It's Y and But reverse!. (No surprise Aristotle knew something about rhetoric.) Since rhetoric is inseparable from communication, understanding how writing works to persuade and influence an audience, whether intentionally or incidentally, is fundamental to communication. Further, because modern digital communication exists in extreme competition, writers and content creators face immense pressure to capture attention. Is it any surprise that rhetoric and other means of winning attention pervades content online today? Communication that uses rhetoric successfully is more likely to be seen. No surprise, more and more content online uses and abuses rhetorical writing. This kind of content is created by people and AI.
How do you interpret the Rhetorical Score? Understand the score is an arbitrary calculation. Certain patterns weigh more heavily on the Rhetorical Score (It's not X. It's Y will increase the score more dramatically. Em dashes less). Now, think of the Rhetorical Score as how much an audience is likely to perceive they are being pushed or pulled by the content. The higher the score, the more likely an audience is to react to the writing, and this reaction could be positive or negative. If they are receptive to the message already, they are likely to react positively (think of this as "preaching to the choir"). If they approach the content with skepticism, perhaps trying to discern value quickly, they are more likely to react negatively. The underlying hypothesis of Un•AI•ify: People will dismiss communication when they perceive it is intended to persuade or when it is perceived to be low-value ("AI writing" could be seen as low-value as it's so easy to produce). Persuasion and low-value could be two parts of the same dynamic: If an idea is valuable, does it need to be sold or will it sell itself?
Regardless, people today have little choice in the matter. They must filter through the flood of content they are exposed to online in order to and discern what's valuable or not. By using "tells" (such as overused and rhetorical writing patterns), people can more quickly filter information. Expect this to continue. Subscribe to the Un•AI•ify blog for more in-depth discussion of modern communication dynamics.
TL;DR: Use the Rhetorical Score as a flag to gauge how likely your audience may see the content as rich in information or rich in persuasion. Ask: How will this content be received by the audience?
By raising awareness of the overuse of rhetoric and other Patterns, the astute reader and writer can produce better writing and a sharper, smarter mind.
If there are other patterns you would like Un•AI•ify to identify, reach out to [email protected].