AI Sycophancy Detector
Sycophancy is an AI telling you what you want to hear: it agrees, inflates and drops the counterargument. This check scores how much of that is in your draft.
Paste any AI-written or AI-assisted text. You get a Yes-Man Score from 0 to 100 and a quote for every overclaim, missing counterargument and vague superlative the model found, with a plainer version of each.
This usually takes 3-5 seconds.
Sycophancy Detector runs the moment you sign in, on the text you just pasted. Nothing to enter twice.
A saved check is kept for 24 hours.
You've used up today's free runs on this check.
Come back tomorrow for another free run.
Why Detect Sycophancy?
AI chatbots affirm users' views 49% more often than other humans do (Science, 2026). This means AI-assisted content systematically overclaims, omits counterarguments, and uses stronger language than the evidence supports. Your readers notice — even when you don't.
Paste your content
Any AI-generated or AI-assisted text — blog posts, newsletters, LinkedIn posts, marketing copy.
AI scans for sycophancy patterns
The model checks for overclaiming, missing counterarguments, vague superlatives, uncritical agreement, and excessive hedging.
Get your Yes-Man Score
A 0-100 score with specific flags showing exactly where your content flatters instead of informs, plus fixes for each issue.
Research published in Science (March 2026) found that people exposed to over-affirming AI came away more convinced they were right and less willing to consider alternatives. The Yes-Man Score helps you catch this pattern before publishing.
Want the mechanism behind it? Read what AI sycophancy is and why models do it
Had enough of AI that agrees with you? Make it argue the other side with the AI Red Team
Example Analysis
"New research proves intermittent fasting is the most effective weight loss method available today. Studies consistently show remarkable results, with participants experiencing transformative health improvements across the board."
This content reads like a press release, not a research summary. 'Proves' is too strong for any single body of research. 'Most effective' ignores that no weight loss method outperforms all others for all people. The superlatives are doing the work that evidence should be doing.
"New research proves intermittent fasting is the most effective weight loss method"
'Proves' implies scientific certainty. No single study 'proves' a weight loss method is 'the most effective.'
Frequently Asked Questions
Sycophancy is when AI tells you what you want to hear instead of what's true. It shows up as overclaiming (“proven” when evidence says “preliminary”), missing counterarguments, and vague superlatives (“revolutionary”, “groundbreaking”) without evidence. AI chatbots do this 49% more than humans.
Sycophantic content erodes trust. Your readers may not know the term, but they feel it — content that's 'too positive' or 'too smooth' triggers skepticism. The claims that sound most confident are often the least supported. Readers screenshot and share the wrong ones.
A single model catches the obvious patterns: inflated language, missing caveats, unsupported superlatives. Council scores the same text with 4 independent models, takes the median and drops any flag whose quote is not in your text. It scores tone; the claim and citation checks cover the facts.
We store your text and the result so your receipt link keeps working; the receipt is private unless you share it. We use enterprise API agreements, so your content is never used for model training.
This tool scores one piece of content with one model. On a paid plan, Council scores it with 4 and shows where they split. Whether the claims in it are true is a separate check, Review, which runs claims and citations together.
Powered by TrueStandard
Four independent models check every claim in your draft.