Three engines off the same pipeline. Check whether a quote is real, whether a paper has been retracted, or hand over a whole draft and get every unsupported claim found and cited. New: check whether an argument actually follows. Building something? The same engine is an API.
Half the famous quotes on the internet were never said by the famous person under them. Type the quote and who supposedly said it; we hunt the actual record.
Every other tool here asks whether something is true. This one asks whether the reasoning holds. Paste up to 100 words and get the named fallacies, the exact words that trip them, and the strongest honest version of the point being made.
True or not, it has nothing to do with the point.
The argument is left untouched while the person making it is discredited. Whether they are awful is a separate question from whether they are right.
This witness has been paid by the defendant, so weigh her testimony carefully.
When somebody's CREDIBILITY is the evidence, as with testimony or expert opinion, their interests are relevant. That is not an attack, it is assessing the source.
Hypocrisy is answered instead of the point. A smoker telling you to quit is still telling you something true.
You said last year that this rule was unworkable. What changed?
Pointing at an inconsistency to ask for an explanation is fair. It becomes a fallacy only when the inconsistency is treated as the refutation.
The origin of an idea is treated as its verdict. Bad people occasionally say true things and good institutions publish mistakes.
This study was funded by the company selling the drug, so I want independent replication before I believe it.
Funding is a real reason to want more evidence. Asking for replication is not the same as declaring the finding false.
Expertise is a shortcut, not a proof, and it only transfers inside the expert's actual field.
The consensus of climate scientists is that warming is human-caused.
Citing a broad expert consensus INSIDE the relevant field is normal evidence. The fallacy is one credential outside its lane, or a name used to end the discussion.
Headcount is not evidence. Most people were wrong about most things for most of history.
Nine of ten reviewers found the software crashed on launch.
Many people REPORTING an observation is data about that observation. The fallacy is treating belief as evidence, not treating reports as reports.
The feeling may be entirely appropriate and still not be a reason to accept the claim attached to it.
This policy will separate families, which is a cost worth weighing against the benefit.
Naming a real human cost is part of the argument. It turns fallacious when the feeling REPLACES the weighing rather than entering it.
Naturalness carries no information about safety, efficacy or morality. Hemlock is natural; insulin is not.
This ingredient has been eaten for centuries, so we have long observational evidence about its safety.
A long track record is genuine evidence. The claim rests on the OBSERVATION, not on the naturalness.
Age is not correctness. A practice can survive for centuries because nobody checked it.
This process has survived every reorganisation, which suggests it is solving something the alternatives did not.
Survival under pressure is weak evidence, and treating it as weak evidence is honest. The fallacy is treating age alone as the justification.
A new topic is introduced that feels related and quietly replaces the one that was losing.
Before we discuss the overrun, we need to agree what was in scope, because the number depends on it.
A genuine prerequisite is not a diversion. The test is whether the detour RETURNS to the original question.
The opposing position is restated in a form that is easier to knock down, then knocked down.
If I follow your premise, it implies X. Is that what you mean?
Drawing out an implication and CHECKING it is steelmanning. It becomes a straw man when the weaker version is asserted rather than offered for correction.
Smuggles in what it was supposed to prove.
The thing to be proved is already sitting in the evidence offered for it, so nothing has actually been argued.
Assume for the sake of argument that the premise holds. Then the conclusion follows.
Explicitly ASSUMING something and saying so is a valid move. The fallacy is hiding the assumption inside the premise.
Two claims prop each other up and neither is anchored to anything outside the pair.
These two findings are mutually reinforcing: each is supported independently, and they agree.
Independent support that happens to converge is corroboration. Circularity needs the support to run only through the loop.
A spectrum is presented as a switch, so rejecting one option looks like accepting the other.
Either the number is above the threshold or it is not.
Some choices really are binary. The fallacy is manufacturing a binary where the middle exists and matters.
Any answer concedes the hidden premise, so the question does the arguing before the answer starts.
You said the deadline slipped. What caused it?
A question resting on something ALREADY ESTABLISHED is not loaded. The presumption has to be unproven and contested.
A counterexample appears, so the definition quietly narrows until it no longer fits.
I mean chartered engineers specifically, which I should have said at the start.
Clarifying a definition you were always using is honest. It is a fallacy when the definition changes ONLY to survive the counterexample.
The bold claim is what is being pushed; the modest one is what gets defended when challenged, then the bold one resumes.
I overstated it. The defensible version is the narrower one, and that is what I will argue from now.
Conceding to the narrower claim and STAYING there is honest updating. The fallacy needs the bold version to come back.
Evidence is supplied, so a new and larger requirement appears in place of the one just satisfied.
That answers my first concern. It raises a second one I had not thought of, which is separate.
Genuinely new concerns arise as arguments develop. The fallacy is replacing the ORIGINAL standard rather than adding to it while acknowledging it was met.
Money and time already gone cannot be recovered by spending more. Only future costs and benefits are decision-relevant.
We are three years in, so most of the remaining work is integration, which is cheap. Finishing is now the better bet.
Past spend that changed the FUTURE cost is relevant. The fallacy is treating the unrecoverable spend itself as the reason.
The leap from some to all, or from part to whole.
A small or unrepresentative sample is treated as though it settled the general question.
In our three trials it failed every time, which is enough to stop and investigate before spending more.
A small sample can justify an ACTION without justifying a general claim. The scope of the conclusion is what matters.
A memorable case is given more weight than a large body of evidence pointing the other way.
Here is a case the model does not explain, which is worth understanding.
A single counterexample can be a real problem for a theory that claims to cover everything. It is fallacious when used to outweigh statistical evidence about a TENDENCY.
The evidence shown is real, and the selection is what does the lying.
I am excluding the pre-2010 studies because the measurement method changed, and here is what they showed anyway.
Stated, justified exclusion with the excluded data disclosed is method. Cherry picking hides the criterion and the remainder.
Properties do not always survive assembly. Every part being light does not make the machine light.
Every component is rated to 200 degrees, so the assembly tolerates 200 degrees, assuming no interaction effects.
Some properties DO aggregate, and saying which assumption you are relying on is the difference between an inference and a fallacy.
The reverse mistake. An average or an aggregate says little about any individual inside it.
The batch failed inspection, so every unit in it is suspect until tested.
Treating members as SUSPECT because of an aggregate is a reasonable precaution. Asserting each one has the property is the fallacy.
A chain of consequences is asserted without support for any link in it, and the far end does the persuading.
Three jurisdictions made this exception and each widened it within five years, so the risk is documented.
A slope with EVIDENCE for the mechanism is a forecast, not a fallacy. The fallacy is asserting the chain without it.
Mistakes a pattern for a cause.
Sequence is taken as causation. Something has to happen after everything, and most of it is coincidence.
Crime rose the month the policy took effect, which is worth investigating with a control area.
Timing is a legitimate REASON TO LOOK. It becomes a fallacy at the point it is treated as the finding.
Association is compatible with cause, reverse cause, a third factor driving both, or chance.
The association survives adjustment for the obvious confounders and there is a known mechanism, so cause is plausible.
Correlation plus mechanism plus ruling out confounders is how causal claims are actually built. The fallacy is skipping straight from the correlation.
A complicated outcome is pinned on the one factor that fits the argument being made.
Of the several causes, this one accounts for most of the variance.
Ranking causes by contribution is analysis. Collapsing them to one is the fallacy.
A pattern is found in the data first, then presented as though it had been predicted.
We found this pattern while exploring, and we are now testing it on a fresh dataset.
Exploration is how hypotheses are born. The fallacy is presenting the exploratory hit as a CONFIRMED result without the fresh test.
Independent events have no memory. A coin that landed heads five times is still an even bet.
Six reds in a row makes me suspect the wheel is biased.
Using a streak as evidence about the PROCESS is sound reasoning. The fallacy is using it to predict the next independent draw.
Gets the burden of proof the wrong way round.
Absence of a disproof is treated as a proof. It usually just means nobody has looked.
We looked hard in the places it would show up and found nothing, which is evidence it is not there.
Absence of evidence IS evidence of absence when you searched somewhere it would have appeared. The fallacy is claiming it without the search.
The person making the claim is the one who owes the evidence. Handing that job to the doubter reverses it.
I have given my evidence. If you think it is wrong, that objection needs support too.
Once a claim is SUPPORTED, a challenger does take on a burden. The fallacy is demanding disproof before offering anything.
Fairness of airtime is confused with equality of evidence, which makes a fringe position look like half the field.
The field is genuinely split, so here are both positions with their support.
Real disagreement deserves both sides. The fallacy is manufacturing symmetry where the WEIGHT of evidence is lopsided.
A term is used in one sense in the premise and another in the conclusion, so the argument only appears to connect.
I am using 'significant' in the statistical sense here, not the everyday one.
Flagging which sense you mean prevents the slide. Equivocation needs the meaning to change UNANNOUNCED and do work.
Papers get withdrawn after you cited them, and nobody sends you a memo. Paste a DOI; we check it against the open literature index, free, right now.
Claim detection across everything you paste or upload. Pro adds the argument critique. Citations land in the style you pick.