If attention is elsewhere, the change never happened.
People miss surprisingly large changes when they happen during a blink, a reload, a cut, or a distraction. Interfaces that change in place without showing what changed leave users looking at a new screen while believing it is the old one.
In 1890 William James observed that we take in far less of a scene than we feel we do. Film editors learned it in practice: continuity errors between shots, such as a glass that refills or a scarf that changes color, routinely slip past audiences. In the 1970s George McConkie showed that words changed during a reader’s eye movement often went unnoticed.
The term took hold in 1997. Ronald Rensink, Kevin O’Regan, and James Clark showed that large changes to a photograph went unseen when a brief blank separated the two versions, and Daniel Simons and Daniel Levin named and reviewed the effect. In a 1998 field study, Simons and Levin swapped a stranger asking for directions for a different person mid-conversation, and about half the people did not notice.
People notice a change only when their attention is on the thing that changed at the moment it changes. If the change happens during a blink, a reload, a cut, or a distraction, even a large difference can go unseen, while the viewer feels sure they would have caught it.
In the flicker task, a photo alternates with a slightly altered copy, separated by a brief blank. Even large changes can take many cycles to find, sometimes more than a minute. Changes to the parts people care about are found faster.
A full reload blanks the screen, which is the same disruption as the flicker. People who apply a filter or save a form can miss what changed. Update in place and keep the scroll position.
A brief highlight on an updated row, field, or total draws attention to where the change happened. Let it fade, so the screen settles back to normal.
Changes far from where the user is looking, such as a badge in the header or a total at the top of a long form, often go unseen. Put feedback next to the control that caused it.
In the mudsplash studies, small blobs that appeared at the moment of change, away from it, were enough to hide it without any blank. Simultaneous animations, ads, and pop-ups do the same.
A tile that turns red among dozens of others is easy to miss when nobody was watching it at that moment. Use alerts, persistent change markers, and a log of what changed.
Johansson and Hall asked people to pick the more attractive of two faces, then secretly handed them the other one. Only about a quarter noticed, and many explained a choice they had not made. A clear confirmation step catches the mismatch.
People overestimate how well they detect changes, a finding called change blindness blindness. Designers make the same mistake about their own screens, so test whether users noticed instead of asking whether they would.
AI rewrites, regenerates, and updates content in place. The more it changes at once, the easier it is to miss the one change that matters.
When AI rewrites a document or a block of code, a changed number or a deleted clause looks just like the rest. Show a diff, mark the changed spans, and let people accept changes one by one.
A regenerated or refined answer that replaces the previous one hides what is different. Keep earlier versions, and highlight what changed between them.