Why smart people fall for it

Confirmation bias is not laziness

Confirmation bias is the tendency to seek, notice and remember information that fits what you already think, and to hold contrary information to a higher standard. It does not feel like bias from the inside. It feels like being appropriately careful about a dubious claim and appropriately relaxed about an obvious one.

The asymmetry is the whole mechanism. You are not refusing to check. You are checking unevenly, and the unevenness is invisible because each individual judgement seems reasonable.

The practical counter is not to try harder to be neutral, which does not work. It is to change what triggers a check. Make the trigger the shape of the content rather than your agreement with it: an unlinked statistic, a screenshot, a percentage with no baseline, an emotional jolt. Those triggers fire whether or not you like the conclusion.

Motivated reasoning and why intelligence can make it worse

Motivated reasoning is confirmation bias with something at stake: an identity, a group, a decision you already defended out loud. The conclusion arrives first and the reasoning is assembled afterward to support it.

What is uncomfortable is that better reasoners are better at the assembling. Someone with more knowledge of a topic has more raw material for constructing a defence, and more skill in finding the flaw in a study they dislike. Expertise is a real asset and it is also a bigger toolkit for rationalisation.

The counter is procedural, not attitudinal. Decide what would change your mind before you look. If the honest answer is nothing, you are not evaluating a claim, you are defending a position, and no amount of reading will change the outcome.

The confidence gap, measured

The single most useful thing you can learn about yourself here is the distance between how sure you feel and how right you are. It is measurable and it is usually larger than people expect.

In the Veriff Deepfakes Report 2026, a survey of 1,000 US adults fielded in February 2026, average accuracy at identifying manipulated video scored 0.07 on a scale running from minus 1 to 1, where zero is random guessing. Respondents caught one manipulated clip 52 percent of the time and another only 30 percent of the time. Around half believed they could reliably spot manipulated media, and 79 percent said they worried about deepfake fraud.

So: high concern, high confidence, chance-level performance. That combination is the dangerous one, because confidence is what converts a belief into a share.

Our challenge asks for a confidence rating before every answer for exactly this reason. Seeing the two numbers side by side does something an explanation cannot.

Fluency, repetition and the illusion of truth

Things that are easy to process feel more true. A claim in clear type on a clean layout feels more credible than the same claim in a cluttered one. A statement you have heard four times feels more established than one you have heard once, even when all four times were the same original source recirculating.

Repetition inside a feed is not independent corroboration. Twenty accounts repeating one unsourced claim is one claim, not twenty. This is why the volume of agreement you see is nearly worthless as evidence, and why finding the original matters so much.

The bias blind spot

Most people rate themselves as less susceptible to bias than the average person, which is arithmetically impossible for most people. Believing you are unusually immune is itself one of the better predictors that you are not doing the checks.

None of this is a reason for despair or for treating your own judgement as worthless. It is a reason to externalise the work. A procedure you follow does not care how you feel about the conclusion, which is precisely its advantage over your intuition.