Learn · UX laws
The same principles that make interfaces easy to use can be bent to steer people. For each law: what it says, how it shows up in AI products, and how to spot when it is used against you.
01
The more choices people see, the longer they take to decide.
In AI products
An assistant ends a reply with five suggested follow-ups, or an agent offers many plans with one pre-selected.
How to spot it
Count the options after each reply. Is one highlighted so you take it without thinking?
Watch for
hook.continue
02
Big, close targets are faster to hit than small, distant ones.
In AI products
“Accept” is a large button under your cursor; “Decline” is a small grey link. “Stop generating” sits far from where you are reading.
How to spot it
Compare the size and position of accept and decline, continue and stop.
Watch for
consent.pressure
03
People expect your product to work like the ones they already use.
In AI products
Delete chat, memory settings and data controls live in unusual places, so people never find them.
How to spot it
Can you delete a chat or turn off memory where you would expect it, in two steps or fewer?
Watch for
exit.blocked
04
People can hold only a few items in working memory at once.
In AI products
Long answers bury the one warning that matters in paragraph six.
How to spot it
Is the caveat, limit or risk in the first lines of the reply?
Watch for
risk.unreferred
05
People judge an experience by its most intense moment and its end.
In AI products
Every chat ends with a warm, flattering summary, or a companion makes goodbyes feel guilty.
How to spot it
Read the last message of a session. Does it inform you, or try to make you stay?
Watch for
affect.bond
06
Responses under about 400ms keep people engaged.
In AI products
Fast streaming text feels confident, even when the answer is a guess.
How to spot it
When an answer arrives instantly, does it still show sources and how sure it is?
Watch for
certainty.false
07
People remember unfinished tasks more than finished ones.
In AI products
Replies end on a cliffhanger: “Want me to show you the rest?”
How to spot it
Does the reply finish the job, or leave something out to bring you back?
Watch for
hook.bait
08
People speed up as they get closer to a goal.
In AI products
Streaks, levels and “you’re 80% there” meters on chatbots and companions.
How to spot it
Is the progress meter tied to your goal, or to time spent in the app?
Watch for
hook.continue
09
Every system has complexity that someone has to handle.
In AI products
An agent hides complexity by making choices for you: the dates, the vendor, the price tier.
How to spot it
Does the agent show what it decided on your behalf before it acts?
Watch for
request.altered
10
The item that looks different is the one people remember and pick.
In AI products
A sponsored product appears as the highlighted “best pick” in an answer.
How to spot it
Which option looks different, and is it labelled as paid or as the maker’s own?
Watch for
ad.disguised
11
People trust attractive designs and forgive their problems.
In AI products
Clean headings, tables and confident formatting make a wrong answer look authoritative.
How to spot it
Does the formatting add authority without adding sources?
Watch for
certainty.false
Based on Laws of UX by Jon Yablonski (lawsofux.com) and the original research behind each law.