Short, straight answers about what Bullshit Detection does — and, just as importantly, what it doesn’t.
It analyzes any text you paste in for rhetorical and stylistic patterns that are common in AI-generated writing and in attention-driven online content. It reports a count for each of seven patterns, an estimate of readability, and a single Rhetorical Score that gauges how hard the writing is working to persuade.
No. Bullshit Detection is a tool for awareness, not a verdict. It highlights patterns that AI models tend to reuse, but people write with these patterns too. Non-AI writing can be flagged, and AI writing can slip through. Treat every result as a flag worth a closer look — not proof.
Seven: “It’s not X. It’s Y” (negation then affirmation), “Lady doth protest” (excess intensifier adverbs), “But reverse!” (pivot words like but and however), clichés, buzzwords, triads (three items in a row), and dashes (em and en dashes). Each is explained in full on the Patterns page.
An intentionally arbitrary number that weighs the patterns found in your text. Some patterns count more than others. Read it as a rough measure of how strongly an audience is likely to feel pushed or pulled by the writing — higher means a stronger reaction is likely, positive or negative.
No. The analysis runs entirely in your browser. Your text is not sent to a server or saved by the tool.
No. The analyzer is free to use.
Because frequency and consistency matter more than the dash itself. Not every em dash means “AI wrote this,” and generative AI can write well. The tool simply counts your dashes so you can watch for overuse and keep your style consistent within a document.
Subscribe from the home page. Bullshit Detection is early, and subscribers get first access to new detectors along with writing about influence, persuasion, and the flood of AI content online.