How to Write Button Copy: The Few Words That Do the Work

I once watched a founder spend two full days agonizing over her homepage headline — every word weighed, every comma debated — and then ship the page with a button that said “Submit.” All that persuasion, funneled into a word that means nothing and promises less. Here’s the thing about how to write button copy: the button is where the decision actually happens, and the whole craft fits in one law. A button label is a promise about what happens next — so say what the click does, in the user’s own language, truthfully. Lead with the value when you can (“Get the checklist” over “Download”), match the label’s weight to the size of the ask, and never let a button lie about effort, cost, or consequences. That’s it. Everything else in this guide is that one law, applied.

Two to four words. That’s usually all you get, and those few words carry the whole page — because every headline, every testimonial, every lovingly crafted paragraph exists to walk someone to a button, and the button either finishes the thought or fumbles it. Okay, let’s make sure yours finishes it.

One scope note before we dive in, because this guide has a sibling. If you’re wondering about where the button sits, what color it is, how big it should be, or how to test the whole package, that’s the design-and-placement side of the craft, and I’ve covered it separately in my guide to how to optimize CTA buttons. This piece is about the words — purely the words — because in my experience the words are the part everyone skips and the part that does the most work.

How do you write button copy that keeps its promise?

Start here, because this is the law everything else hangs on: a button label is a contract. The user reads the label, forms an expectation, clicks, and then the interface either honors that expectation or breaks it. “Download the guide” followed by a guide downloading? Contract honored, trust deposited. “Download the guide” followed by a five-field form demanding a work email and a phone number? Contract broken — and the user just learned your buttons can’t be trusted, which taxes every click that comes after.

The quickest way to check any label against the contract is what I call the predict-the-next-screen test: could a first-time visitor, reading nothing but the label, accurately predict what appears after they click? If yes, the label is doing its job. If they’d have to guess — if the honest answer is “something happens, probably” — the label is what designers call mystery meat, and mystery meat is where good conversion rates go to die.

The three beige horsemen: “Submit,” “Continue,” “Learn More”

Let’s name the usual suspects, because you have at least one of them live right now. (We all do. No shame, only audits.)

• “Submit” is the bleakest word in interface design. Nobody wakes up wanting to submit. It describes the database’s experience of the click, not the human’s — the form gets submitted; the person gets a quote, books a call, or joins a list. Label the human’s outcome: “Get my quote,” “Book the call,” “Join the list.”
• “Continue” is a hallway with no sign on the door. Continue to what? The fix is the destination preview — “Continue to shipping,” “Continue to payment” — so the user always knows which room is next. One honest caveat, though: mid-flow, in a wizard where the next step genuinely is “more of the same form,” plain “Continue” can be the truthful label, and truth wins. Dress it up only when there’s a real destination to name.
• “Learn More” is the shrug of button copy. More what, about what? “See plans and pricing,” “Read the full case study,” “Watch the 2-minute demo” all pass the predict-the-next-screen test; “Learn More” never does. (It also creates an accessibility maze we’ll get to — this phrase fails twice.)

Notice none of these are evil — they’re just vague, and vague makes people hesitate, and hesitation is friction you can delete with a better verb. If you’ve already run a full conversion audit, you may have flagged vague buttons under the clarity lens; this is the chapter on fixing them.

“Start my trial” or “Start your trial”? The first-person question

You’ll run into advice insisting that first-person labels (“Start my trial,” “Get my checklist”) dramatically outperform second-person ones (“Start your trial”). The theory is sweet: first person reads like the user’s own inner voice completing the sentence “I want to…”. And sometimes it does seem to help. But I won’t hand you a made-up percentage, because the honest answer is: both work, it varies by audience, and if you have the traffic, test it on your own buttons. What matters more than which pronoun you pick is consistency — a page that mixes “Start my trial” on one button with “Claim your discount” on the next reads like it was written by a committee mid-argument. Pick a person, stay in it.

How to write button copy for every situation: the pattern library

Here’s the centerpiece: six patterns that cover nearly every button you’ll ever write, each with a before-and-after pair. All the examples are fictional — invented businesses, invented buttons — but the shapes are exactly what you’ll apply to your own pages.

Pattern 1: The value-forward label

Before: “Download.” After: “Get the checklist.”

“Download” names the mechanism; “Get the checklist” names the prize. People don’t click for file transfers, they click for the thing — so put the thing in the label. Imagine Lena (fictional), who runs a gardening blog and offers a planting calendar. “Download” tells her readers a file exists. “Get the spring planting calendar” tells them why they’d want it. Whenever the click delivers something, the label should hold what’s delivered.

Pattern 2: The specificity upgrade

Before: “Pricing.” After: “See plans and pricing.”

Specific beats generic every time it’s true — and “when it’s true” is the load-bearing phrase. “See plans and pricing” is a stronger label than “Pricing” because it previews exactly what the page contains: plans, compared, with prices. If your pricing page is actually a “contact sales” wall, then “See plans and pricing” is a lie in a nice outfit, and you’ve traded a click for a grudge. Upgrade the specificity only as far as the next screen can back it up.

Pattern 3: The effort-honesty label

Before: “Sign up.” After: “Start free — no card needed.”

This is the qualifier that converts by removing doubt rather than adding pressure. The hesitation behind a signup button is rarely about wanting the product; it’s “what is this going to cost me, in money or hassle?” If the honest answers are “nothing” and “none,” say so at the button — “Start free — no card needed,” “Try it free for 14 days.” But the rule is absolute: only when true. If your flow asks for a card on screen two, that label is a trap, not a pattern. (Free-trial pages live and die on this kind of honesty — the same principle that governs honest trial design governs the button that starts it.)

Pattern 4: The next-step label

Before: “Continue.” After: “Continue to shipping.”

Multi-step flows are where “Continue” breeds. The next-step label adds a destination preview, which does two quiet jobs at once: it tells the user what’s next, and it reassures them about what’s not next (you’re not being charged yet; this is just shipping). In a fictional checkout for Theo’s print shop: “Continue to shipping” → “Continue to payment” → “Review order.” Each label is a signpost, and a signposted flow feels shorter than an unsigned one even when the step count is identical.

Pattern 5: The low-commitment label

Before: “Buy the template.” After: “Preview the template.”

This is the commitment-matching principle: the label’s weight should match the ask’s weight. A visitor who’s three minutes into discovering you is not ready for “Buy” — but they might be ready for “Preview the template,” “See how it works,” or “Browse examples.” A low-commitment label isn’t weaker; it’s calibrated. It catches the not-yet-ready instead of scaring them off, and it keeps the heavy label (“Buy the template — $29”) for the moment the visitor has actually reached buying weight. One page can honestly hold both: the heavy ask for the ready, the light ask for the curious.

Pattern 6: The final-step gravity label

Before: “Confirm.” After: “Place order — $49.00.”

The last button before money moves deserves maximum clarity and zero surprise. “Place order — $49.00,” with the real, final total right in the label, is the no-surprise button: the user knows exactly what they’re agreeing to, exactly what it costs, at the exact moment of commitment. This is the pay-button honesty law, and it’s not just ethics — surprise at the final step is precisely the doubt that abandons carts. The same gravity applies to any consequential click: “Cancel subscription,” “Delete account,” “Send to 2,400 subscribers.” Big consequences get plain, complete, unambiguous labels.

The pattern library at a glance

Pattern	Before	After	Use when
Value-forward	Download	Get the checklist	The click delivers something
Specificity upgrade	Pricing	See plans and pricing	The next screen can back the detail up
Effort-honesty	Sign up	Start free — no card needed	The reassurance is 100% true
Next-step	Continue	Continue to shipping	Multi-step flows with real destinations
Low-commitment	Buy the template	Preview the template	The visitor isn’t ready for the big ask
Final-step gravity	Confirm	Place order — $49.00	Money or consequences are about to move

How do you match button copy to your brand’s tone?

Once the label is honest and specific, you get to ask how it sounds — and the answer is a register check, not a personality contest. Your brand voice gets about two words of runway on a button, so spend them deliberately.

Playful brands can play — in the right rooms. A quirky stationery shop can absolutely say “Let’s go” on a quiz or “Make it mine” on a product page; the label still says what happens, just in the brand’s accent. A bank should not “Let’s go” anyone into a mortgage application. The register check is simple: read the label aloud in the voice of the moment the user is in. Browsing a fun product? Light works. Moving money, sharing data, making a commitment? The voice should steady, not sparkle.

The panic-register ban: serious actions stay plain. Delete, cancel, unsubscribe, remove — these buttons get boring labels on purpose. “Delete account” — never “Say goodbye forever 😢,” never “Nuke it.” A user at a destructive action is often stressed, sometimes upset, and occasionally about to make a mistake; cute copy there reads as mockery and, worse, can blur what the button actually does. Plain labels on serious actions are a kindness and a safety feature in one.

The confirmshaming ban: opt-outs stay neutral. You’ve seen the pattern — a popup where declining requires clicking “No thanks, I hate saving money.” That’s confirmshaming, and here’s my whole take: it’s a trick that collects email addresses from people who now slightly resent you, which is a strange thing to pay for. The ethics line is bright and easy to hold: the decline option is labeled as neutrally as the accept option. “No thanks” or “Maybe later” — full stop. Persuade with the offer, never by humiliating the exit. (And no, SocialBlaze won’t optimize your popups for you — we’re a social media scheduling tool, not a CRO suite — but we do feel strongly about buttons that respect people.)

What should the words around the button say?

Button copy doesn’t work alone; it works in a tiny ensemble, and two supporting players matter most.

The clicktrigger: the small line that removes the last doubt

A clicktrigger is the short reassurance that sits under or beside a button — “Cancel anytime.” “Takes 2 minutes.” “No spam, ever.” It’s not decoration; it’s a targeted answer to the exact doubt a user feels at that button. The craft is diagnosis: figure out the hesitation (“will this trap me?” “will this take forever?” “will you flood my inbox?”) and answer precisely that, in five words or fewer. And the rule from pattern 3 applies with full force: true claims only. “Cancel anytime” next to a cancellation flow that requires a phone call is not a clicktrigger, it’s evidence. Clicktriggers are really a species of microcopy, and if you want the full craft — field labels, error messages, empty states, all the tiny words — my guide to how to write microcopy that converts goes deep on it.

Button-and-headline agreement: the mixed-signals audit

Here’s a dissonance that’s invisible to the team and loud to the visitor: the headline promises one thing and the button asks for another. A headline that says “Plan a month of posts in an afternoon” above a button that says “Request a demo” is a mixed signal — the headline sold self-serve speed, the button asks for a sales call. Read your headline and your button as one sentence: the headline is the promise, the button is the first step toward it, and they should feel like the same conversation. If they don’t, one of them is lying about what the page is for — figure out which, and fix that one.

How do you make button copy accessible?

Everything above gets better when you write for ears as well as eyes — and this layer is non-negotiable, not nice-to-have.

Labels must work out of context. Screen reader users commonly navigate by pulling up a list of a page’s buttons and links, stripped of all surrounding layout. Now picture that list for a page with seven “Learn More”s: Learn more. Learn more. Learn more. A maze with identical doors. The distinguishable-labels rule fixes it: every button’s label should identify its own action without any visual context — “Read the pricing FAQ,” “See the template gallery,” “Watch the setup demo.” Delightfully, this rule and the predict-the-next-screen test are the same discipline wearing two hats: specific labels serve everyone.

Actions are buttons; navigation is links. One honest paragraph of semantics, because it changes how assistive tech announces your words: a button performs an action (submits, saves, deletes, opens a dialog), while a link takes you somewhere. Screen readers tell users which is which, and users build expectations on it — “button” implies something will happen here; “link” implies travel. When a thing styled as a button is secretly a link (or vice versa), the label gets announced with the wrong contract attached. You don’t need to become an ARIA scholar today; just keep the honest distinction, and your labels will mean what they’re announced to mean.

Never let color or an icon carry meaning alone. A red button isn’t a warning to someone who can’t perceive the red; a trash-can icon with no text isn’t “delete” to a screen reader (and honestly, icons without words are mystery meat for everyone — quick, what does the circle-with-a-slash do?). Words first; color and icons reinforce.

Should you test your button copy?

Where your traffic allows it — genuinely, yes, and button copy is one of the best things you can test: it’s a single variable, it’s a two-minute change, and it sits at the exact moment of decision. The discipline is the same one every honest experiment inherits: change one thing (the label — not the label plus the color plus the headline), run it long enough to trust, and let your audience answer questions like the first-person one for your own page. I won’t tell you what lift to expect, because nobody honestly can — your audience, offer, and baseline are yours alone.

And if you’re running a low-traffic site that can’t feed a trustworthy split test? You’re not locked out of good buttons — you just skip the ceremony. Apply the law, apply the patterns, run the predict-the-next-screen test on every label, then measure honestly: note your button’s click-through or form-completion numbers before the change, change the label, and compare the same metric over a comparable period after. That’s not a controlled experiment and you shouldn’t pretend it is — but “the label is clearer, the numbers didn’t get worse, and support stopped asking what the button does” is a perfectly honest win.

What button copy mistakes should you avoid?

The anti-patterns, gathered in one place so you can sweep your site for them:

• Mystery meat. “Submit,” naked “Continue,” “Learn More,” icon-only buttons — any label that fails the predict-the-next-screen test.
• Trick labels. The “Confirm” that quietly subscribes you to a newsletter, the “Continue” that places an order, the “Download” that opens a form. A label that lies once poisons every label on your site.
• Confirmshaming. Any decline option engineered to embarrass. Opt-outs stay neutral; that’s the ethics line, and it doesn’t bend for Q4.
• Cute-on-serious. Playful copy on delete, cancel, payment, or legal actions. Save the sparkle for low-stakes rooms.
• The seven identical “Learn More”s. A clarity failure and an accessibility failure in one — every button distinguishable, out of context, always.
• Labels that lie about effort or cost. “Start free” with a card wall behind it, “Quick signup” with eleven fields, a final button that hides fees for the receipt to reveal. Effort-honesty and pay-button honesty aren’t optional patterns; they’re the contract.

Your button copy toolkit

Everything above, compressed into the four things worth copying into your notes app.

The label-audit workflow

• Inventory every button on your money pages — yes, every one, including the footer and the popups. A spreadsheet with three columns: current label, what the click actually does, pattern to apply.
• Run the predict-the-next-screen test on each label. Mark every label a stranger couldn’t predict from.
• Apply the patterns: value-forward for deliverables, specificity where the next screen supports it, effort-honesty where it’s true, next-step previews in flows, low-commitment labels for early-stage visitors, full gravity on final steps.
• Read the page as one voice: consistent person (my/your), register matched to stakes, button and headline telling one story.
• Check the out-of-context list: would every label make sense in a bare list of buttons? Fix duplicates and vagueness.
• Measure honestly — split test where traffic allows, before-and-after where it doesn’t.

The clicktrigger menu

Pick the line that answers the actual doubt at your button — and only if it’s true: “Cancel anytime.” / “Takes 2 minutes.” / “No card needed.” / “No spam, ever.” / “Free forever plan — no trial clock.” / “Your data stays yours.” One per button, five words or fewer, placed where the hesitation happens.

The ethics card

Four lines, non-negotiable: the label says what the click does. Reassurances are true or absent. Opt-outs are neutral. Final steps show the real cost. Everything else in button copy is craft; these four are character.

The two-minute rewrite drill

For any weak button: name what the user receives (value), name what screen comes next (contract), name their biggest doubt (clicktrigger), then write the shortest label that carries the first two and pair it with a line that answers the third. Most buttons are fixed in one pass.

FAQ: how to write button copy

How long should button copy be?

Two to four words is the sweet spot for most buttons — long enough to say what happens, short enough to scan. Final-step buttons can run longer when clarity demands it (“Place order — $49.00”), and that trade is always worth it: never sacrifice the contract to save a word.

Is “Submit” ever okay on a button?

Almost never — it describes the system’s experience, not the person’s. Label the outcome instead: “Get my quote,” “Send message,” “Book the call.” The rare defensible case is an internal tool where “submit” is genuinely the users’ own vocabulary; even there, a more specific verb usually exists.

Should button copy be in first person or second person?

Both can work, and no honest universal winner exists — it varies by audience and context. First person (“Start my trial”) reads like the user’s inner voice; second person (“Start your trial”) reads like the brand speaking. Test it on your own buttons if you have the traffic, and whichever you choose, keep it consistent across the page.

What is a clicktrigger?

A clicktrigger is a short reassurance placed under or beside a button — “Cancel anytime,” “Takes 2 minutes,” “No spam, ever” — that answers the specific doubt a user feels at that moment. It works only when the claim is completely true; a false clicktrigger damages trust faster than no clicktrigger at all.

Why is vague button copy an accessibility problem?

Screen reader users often navigate by pulling up a list of a page’s buttons and links out of visual context. Seven buttons that all say “Learn More” become indistinguishable in that list, making the page a maze. Every button label should identify its own action on its own — which also happens to make it better copy for everyone.