
Microcopy Examples: 9 Choices Compared
On this page
- 1. Specific button labels vs Submit
- 2. Visible labels and persistent hints vs placeholders
- 3. Specific errors with a fix vs generic errors
- 4. Empty states with a next action vs blank space
- 5. Confirmation messages vs a generic success message
- 6. Price and fee wording near the action
- 7. Reassurance beside the action vs an unsupported promise
- 8. Plain language vs jargon
- 9. AI chat replies that match policy
- Testing microcopy with limited traffic
UX
The best microcopy examples do not always use the fewest words. Microcopy is the short functional text that helps someone act: button labels, field labels and hints, errors, empty states, confirmations, price and reassurance lines, and chat replies. On a hypothetical quote form, changing Submit to Send my quote request tells the visitor what will happen.
NN/g recommends enough words to describe a command accurately, in no more than two to four words, while its placeholder research shows what can go wrong when a form saves space by dropping visible labels. Aim for the fewest words that answer the person's question at that moment. Microcopy is one part of a small company's user experience, alongside the layout and behavior of the path itself.
| Choice | When it helps | When it hurts | Evidence (source and date) | Effort to change | How to test |
|---|---|---|---|---|---|
Specific button vs Submit | Makes the action clear | Promises an outcome that does not happen | NN/g, 3 March 2019 | Low | Ask what happens on click |
| Visible label and hint vs placeholder only | Keeps instructions available | Hint is irrelevant or wrong | NN/g, 11 May 2014; W3C, 13 May 2024 | Medium | Fill and review the form |
| Specific error and fix vs generic error | Explains how to recover | Appears before entry is complete | NN/g, 14 May 2023; Baymard, 9 January 2024 | Medium | Submit bad input |
| Useful empty state vs blank page | Explains what to do first | Says empty while data loads | NN/g, 19 September 2021 | Low | Open a new account |
Detailed confirmation vs Done | Sets the next expectation | Promises an unsupported response time | W3C, WCAG 2.2 | Medium | Submit and check announcement |
| Price and fee wording near action | Makes known costs visible | Hides mandatory charges | Baymard, 22 September 2025 | Medium | Find the total before acting |
| Reassurance near action | Answers a real concern | Makes a claim that is false | FTC Act, Section 5 | Low | Check the policy and setting |
| Plain language vs jargon | Makes instructions easier to understand | Removes a term customers need | GOV.UK, Use clear language | Low | Ask users to explain the text |
| AI chat tone and policy accuracy | Gives clear answers to routine questions | Makes a confident, wrong promise | NN/g, 17 July 2016; Moffatt, 14 February 2024 | Medium | Compare replies with policy |
Effort ratings are Iron Goo's estimates for changing an existing path. They depend on who can edit the site and whether the form or chatbot needs code changes. Sources were read on 29 October 2026.
1. Specific button labels vs Submit
When it wins. A verb and object tell people what the click will do. For a hypothetical plumbing company whose form sends a request for a later quote, compare these labels:
Submit
Send my quote request
NN/g's command-label guidance, published by Anna Kaley on 3 March 2019, recommends verbs for commands that start actions or submit information. It favors specific verbs over generic ones such as Next and says to add a noun when the target is unclear. Calling Submit generic is Iron Goo's application of that rule: the NN/g article does not discuss that label.
The label should also reflect the state the system will move into. On a confirmation dialog, use a command such as Discard and offer Cancel, so the person can see what each choice does.
When it loses. Book now is misleading if the form only sends a request and no appointment is booked. A cute label can also make a routine action harder to identify. Ask a tester what will happen before they click; if their answer describes a different outcome, change the label or the action.
2. Visible labels and persistent hints vs placeholders
When it wins. Keep the field name visible outside the input, and keep any necessary instruction visible while the person types. For a hypothetical plumbing quote form that uses text messages to arrange visits:
| Before | After |
|---|---|
Placeholder inside the field: Phone | Visible label: Mobile number. Persistent hint: We text you the visit time. |
Katie Sherwin's NN/g research on placeholders, published 11 May 2014 and last reviewed 10 September 2018, explains several failure modes. When the placeholder disappears, people must remember the instruction and cannot easily review what each entry means. A placeholder can also be mistaken for a default value, and text that vanishes on focus can frustrate keyboard users.
The W3C WAI forms tutorial, updated 13 May 2024, says placeholders disappear on input, usually have lower contrast and do not replace labels. WCAG 2.2 Success Criterion 3.3.2, Labels or Instructions, Level A, states: “Labels or instructions are provided when content requires user input.” See the wider accessibility and inclusive design checks for work beyond labels.
When it loses. A persistent hint takes space, so use it for information the person needs to enter or review the answer. A placeholder can still show a format example alongside a visible label, provided the instruction needed to complete the field remains available after typing.

The revised form keeps its field labels and hint visible while the visitor types. Its button names the request being sent.
3. Specific errors with a fix vs generic errors
When it wins. An error should identify the problem in text and give a usable correction when one is known. For a hypothetical service company accepting ten-digit US phone numbers:
| Before | After |
|---|---|
Invalid input | Enter a 10-digit phone number, like 312-555-0100. |
The WCAG 2.2 Error Identification criterion, Success Criterion 3.3.1 at Level A, states:
If an input error is automatically detected, the item that is in error is identified and the error is described to the user in text.
Success Criterion 3.3.3, Error Suggestion, Level AA, requires known correction suggestions unless providing them would jeopardize the security or purpose of the content. NN/g advises putting the message near the field, using an indicator beyond color alone, and preserving the person's entries. NN/g's error guidance also recommends plain, nonjudgmental language and no jokes in routine errors.
When it loses. An error shown while someone is still entering a valid answer can interrupt the task. In Baymard's inline-validation research, published 9 January 2024, premature validation frustrated participants; 31% of sites in its ecommerce benchmark had no inline validation. Test timing as well as wording, and make a network failure explain how to retry without losing entered data.
4. Empty states with a next action vs blank space
When it wins. A first visit to an account can leave a saved quotes page with nothing to display. The empty state should say what is present now, offer a cue about the feature and provide a direct route to the useful action. Those are the three guidelines in Kate Kaplan's NN/g empty-state research, published 19 September 2021.
For a hypothetical plumbing company that lets customers save quotes:
| Before | After |
|---|---|
No data | You have no saved quotes yet. Request a quote to see it here. Button: Request a quote |
The revised message explains the blank page and gives the visitor a way forward. The button should lead to the actual request form.
When it loses. Do not display “no saved quotes” while the account is still loading or when the request for data has failed. Those are different system states and need different messages. Check the page with an empty account, during loading and with the data request unavailable.
5. Confirmation messages vs a generic success message
When it wins. After a request is sent, tell the person what happened and what happens next. Iron Goo's rule is to name who will contact them and by when, provided the company can support that promise. For a hypothetical plumbing company with a stated one-business-day response policy:
| Before | After |
|---|---|
Done | Your quote request was sent. Our office team will contact you within one business day to discuss the job. |
That message answers the immediate question: did the form work? It also makes the next step observable for both the visitor and the company. NN/g's first usability heuristic, visibility of system status, calls for keeping people informed about what is happening.
When it loses. A deadline in the message becomes misleading if the office cannot meet its stated policy. Use the response window the business actually offers, and update the confirmation when that policy changes.
WCAG 2.2 Status Messages, Success Criterion 4.1.3 at Level AA, requires status messages in markup to be programmatically determinable so assistive technology can present them without receiving focus. W3C gives a successful form submission message as an example. For measurement, fire generate_lead on the confirmation page after a successful submission, not when someone merely clicks the button.
6. Price and fee wording near the action
When it wins. Put known mandatory charges where a person can see them before requesting or booking the service. If the final price depends on an inspection, say what is known and what the quote will establish. For a hypothetical plumbing company with a fixed callout charge:
| Before | After |
|---|---|
Book a visit | Book a visit. The callout fee is $75; we will quote any repair separately before work begins. |
The example assumes both statements match the company's actual terms. If tax or another mandatory charge applies, include it in the disclosed total where it can be calculated.
Baymard's checkout abandonment research, last updated 22 September 2025, reports that 40% of US online shoppers who abandoned during checkout cited extra costs that were too high, including shipping, tax and fees. Another 12% could not see or calculate the total order cost up front. Those figures exclude the 42% who were just browsing or not ready to buy; the page does not state a sample size. They are ecommerce findings, not forecasts for a service company.
When it loses. A neat price line hurts when it omits a mandatory fee or implies a fixed total the company cannot offer. Iron Goo's inference is that a callout fee or price range withheld until a service call creates a similar uncertainty for the visitor.
The FTC Rule on Unfair or Deceptive Fees was announced on 17 December 2024 and took effect on 12 May 2025. It applies to live-event tickets and short-term lodging, requires the total including mandatory fees to be clear, conspicuous and the most prominent price, and does not ban fees. This is not legal advice.
7. Reassurance beside the action vs an unsupported promise
When it wins. A short reassurance line can answer a question at the moment it arises. For a hypothetical service company offering a free account that requires no payment details:
| Before | After |
|---|---|
Create account | Create account with No credit card required beside the button |
The same logic can apply to Cancel anytime or We never share your number when the offer and policy actually support those statements. Place the line near the decision it addresses, where the visitor can use it.
When it loses. A reassurance line is a factual claim, not decoration. Some popular microcopy articles recommend these phrases without making truth a condition; their reported lift percentages are single vendor anecdotes and do not establish a general result.
A reassurance or price line that is not true can create liability. In the United States, Section 5 of the FTC Act prohibits deceptive acts or practices. This is not legal advice.
Iron Goo's decision rule is to ship a reassurance line only when you can point to the setting or policy that makes it true and name the person who would notice if it stopped being true. Check both again when the offer changes.
8. Plain language vs jargon
When it wins. Use the words your customer uses for the task. A hypothetical cleaning company asking a visitor to request a service could make this change:
| Before | After |
|---|---|
Please initiate a service procurement inquiry. | Ask for a cleaning quote. |
GOV.UK's clear-language guidance advises writers to try to split sentences over 25 words and keep paragraphs to no more than five sentences. It recommends familiar words such as buy, help and about in place of purchase, assist and approximately. Its guidance says people with higher literacy, including specialists, prefer plain English, and reports that one in six adults in England have very poor literacy skills.
The US Plain Writing Act of 2010 applies to federal agencies. A private company can still use the same practical test: ask a customer to explain an instruction in their own words.
When it loses. A technical term can be the clearest choice when the customer already uses it. A trade buyer looking for a known part number needs that identifier on the page and in search, even if the surrounding instructions use ordinary language.
9. AI chat replies that match policy
When it wins. A chatbot reply is useful when its facts match the current policy and its tone fits the question. For a hypothetical service company whose policy says the office confirms appointment times:
| Before | After |
|---|---|
You're all booked! We'll be there tomorrow. | Your request was sent. Our office will contact you to confirm an appointment time. |
The second reply assumes the request was successfully sent and that the policy says the office will make contact. Iron Goo builds and runs website chatbots that handle routine inbound questions; the business still needs an accurate policy for those answers to follow.
Kate Moran's NN/g research describes four tone dimensions: formal or casual, serious or funny, respectful or irreverent, and matter-of-fact or enthusiastic. In a study of 50 US respondents, differences between paired samples were statistically significant but small, about half a point to one point on a five-point scale. Choose a tone deliberately, then check the factual answer first.
When it loses. A friendly reply cannot repair a wrong promise. In Moffatt v. Air Canada, 2024 BCCRT 149, decided on 14 February 2024, the British Columbia Civil Resolution Tribunal held the airline liable for negligent misrepresentation after its website chatbot gave incorrect bereavement-fare information. Air Canada argued it was not liable for information from its agents, including the chatbot; the tribunal said the airline was in effect suggesting the chatbot was a separate legal entity responsible for its own actions and called that “a remarkable submission.”
Compare answers about prices, cancellation and response times with the policy page after every policy change. Iron Goo's rule is to avoid jokes in replies about money or errors, where a clear answer and a next step matter more.

Compare the chat reply with the policy it is meant to represent. Check the actual promise as well as the tone.
Testing microcopy with limited traffic
A small site may not collect enough conversions to make most A/B tests practical. In a hypothetical example, 300 visits a week at a 3% conversion rate produce nine conversions a week: 300 × 0.03 = 9. Kohavi and Thomke report that only about 10% to 20% of experiments at Google and Bing generate positive results, so a plausible wording change should be treated as a hypothesis.
Start with five people who resemble your customers. Give each a task, such as requesting a plumbing quote, and watch where they hesitate or misread the next step. The small-team UX research guide explains how to run task tests and customer interviews.
Then run three submissions yourself: leave required fields empty, enter a phone number in the wrong format, and submit with the network off. Check that each message identifies the problem, gives a useful next step and preserves entered data. Use Microsoft Clarity's rage click and dead click filters on the form URL to find interactions worth inspecting, and use the conversion-path guide to find which step loses people before choosing what to rewrite.
Put your next steps in order
Build on what you have just read with a free ebook on semantic SEO and AI visibility. Get worked examples, decision rules, checklists and a do-this-week action in every chapter, plus a six-week plan for your site.


