BEHIND THE BENCH / A PRACTICAL GUIDE
Bullshit Detector
Make vague writing clearer and separate confident language from evidence.
When it helps
A proposal sounds impressive, but you still cannot tell what it does or what supports its claims.
What to bring
A headline, email, proposal or paragraph up to 5,000 characters. Use non-confidential text.
What happens
AI reviews the writing in Normal or Roast style. Normal is constructive; Roast adds sharper humor about the writing. Research Auto adds a public-source check when appropriate, On requests it, and Off reviews wording without web research. Every mode still uses AI.
What you leave with
Editorial findings, a plain-English interpretation and a suggested revision. Copy a section or the full review, or return to revise your source.
Illustrative editorial takeaway from the built-in sample
A short example
Input. Fictional sample: “Our innovative AI-powered platform leverages cutting-edge synergies to unlock transformational customer outcomes at scale.”
Result. Ask what the platform actually does, for whom, and with what evidence. Replace abstract promises with a specific description you can support.
Example only. AI wording can vary; this is not an interactive result.
Why it works this way
Vague writing is easier to improve when you can see the specific phrases causing trouble. Findings and a suggested revision sit together so you can compare the diagnosis with the proposed fix.
Your judgment stays in the loop
Verify that a revision preserves your meaning. A score or confident critique does not prove a claim false. Read cited sources and the research limitations when research is used.
Current practical limits
Public research can be incomplete or unavailable. Research Off is not a privacy mode. This is not a truth detector, plagiarism test or guarantee of accuracy. Sample text is fictional input for a real AI review, not a canned result.