UX writing: how a few words change what people do

Button labels, form hints and error messages are some of the cheapest things to test. Four documented cases where wording alone moved behavior, and the rules that come out of them.

First published 2018-09-18Updated 2026-11-06

Short answer (2026): Wording changes alone have moved measurable behavior in published tests. Google saw 17% more engagement when a hotel search button changed from “Book a room” to “Check availability”. GOV.UK raised clicks on one button from 10% to 65% by changing two words. Large field experiments with text message reminders show changes of 1 to 2 percentage points from wording alone. Write for what the user is trying to do at that moment, in words they already use.

UX writing is the text inside a product: button labels, form hints, error messages, empty states, confirmations. Copywriting gets people to try a product. UX writing helps them get through it.

Because it’s only words, UX writing is one of the cheapest things to change and test. Here are four cases where the effect was measured.

Case 1: Google, “Check availability”

At Google I/O 2017, Google UX writer Maggie Stanphill described a change to hotel search results. The button said “Book a room”. The team found that was too committal for people who were still comparing hotels and dates. They changed it to “Check availability” and saw a 17% increase in engagement.

The talk didn’t give the baseline or define engagement, so treat the number as an example, not a benchmark. The reasoning is the useful part: the old label asked for a decision the user wasn’t ready to make. The new one matched what they were actually doing, which was looking.

Case 2: GOV.UK, “Finish” versus “Give feedback”

The last page of GOV.UK’s driving test booking service had a button that led to a short satisfaction survey. The team tried four labels over four releases and measured how many people clicked through:

The button always led to the same survey. “Give feedback” described what the service wanted. “Finish” described what the user wanted. The second got six times as many clicks.

Case 3: Vaccine reminders, “made available to you”

In 2021 a team of researchers ran two large experiments with UCLA Health patients, published in Nature. Sending a text reminder raised COVID-19 vaccination appointments by 6.07 percentage points, an 84% increase over people who got no message. Vaccinations rose by 3.57 points, or 26%.

The researchers also tested wording. Messages that told people the vaccine had “just been made available to you” did better than plain reminders: 1.51 more points for appointments and 1.09 more for vaccinations. Same channel, same timing, same offer. The only difference was framing the vaccine as already belonging to the reader.

Case 4: Flu shots, “waiting for you”

A 2022 megastudy in PNAS tested 22 different text messages encouraging flu shots, sent to 689,693 Walmart pharmacy customers. On average the messages raised vaccination by 2.0 percentage points, a 6.8% increase. The best one was a pair of texts sent three days apart, the second telling people a flu shot was waiting for them at Walmart. It raised vaccination by 2.9 points.

The two health studies point the same way. People respond to wording that makes the next step feel like theirs and already prepared, instead of a task someone else is asking them to do.

Rules that come out of this

Describe the user’s goal, not yours. “Finish” beat “Give feedback”. “Check availability” beat “Book a room”. Before you label a button, ask what the person thinks they’re doing at that moment.

Match the commitment to the stage. Early in a flow, use low-commitment words: see, check, compare, try. Save “buy”, “book” and “pay” for the step where that actually happens.

Use the words people already use. “Sign in” and “Sign up” look almost identical and confuse people, especially non-native speakers. “Log in” and “Create account” are harder to mix up. Avoid internal jargon. An error that says “IP address conflict” tells most people nothing. Say what happened and what to do next.

Pick one word and keep it. If every step in a flow says “Next”, don’t switch to “Continue” or “Proceed” halfway through. People wonder whether different words mean different results.

Answer the worry before it comes up. Airbnb’s booking request form tells guests what a host wants to read, so they don’t stall in front of an empty message box. Netflix has explained at sign-up that you can cancel online at any time. A sentence like that near the decision point often does more than a better headline.

Write error messages that fix the problem. State what went wrong in plain words, say how to fix it, and keep what the user already typed.

Test it. Wording is cheap to change, so it’s cheap to A/B test. Results don’t transfer reliably from one product to another, so the cases above are ideas to try, not answers.

Sources

1. Maggie Stanphill, "How Words Can Make Your Product Stand Out", Google I/O 2017. youtube.com
2. GOV.UK, "Book a practical driving test: how changing the wording on one button increased clickthroughs by 600%", Inside GOV.UK blog, 2014-04-23. insidegovuk.blog.gov.uk
3. Hengchen Dai and others, "Behavioural nudges increase COVID-19 vaccinations", Nature, 2021-08-02. pmc.ncbi.nlm.nih.gov
4. Katherine L. Milkman and others, "A 680,000-person megastudy of nudges to encourage vaccination in pharmacies", PNAS, 2022-02-01. pmc.ncbi.nlm.nih.gov
5. Prototypr, "Building Better Product with UX Writing", 2018, the article the original version of this page was based on. blog.prototypr.io