How to Write CTA Button Text: Clear and Accessible

Good call-to-action text does three jobs at once. It names the action, it tells the visitor what they get, and it still makes sense when read on its own, without the design around it. A button that says "Submit" or "Click here" usually fails the last test. Both the W3C and Google point the same way on this: link purpose should be clear from the label or its context, and Google specifically names generic phrases as poor anchor text.

It is written for marketers and content editors who decide button wording. Sentences that close a sponsored post are a different craft; for those, see sponsored post template: intro, body and CTA sentence examples.

Anatomy of a good label

Think of a CTA as a very short sentence with a verb, an object and, sometimes, a benefit. When all three work together, the label is understood even when nothing else on the page is visible, which is exactly how a screen reader user or a skimming visitor meets it.

Three parts: verb, object, benefit

• Verb: what the visitor does. Download, apply, compare, book.
• Object: what the verb acts on. Download the price list, start a free trial.
• Benefit: optional, what the visitor gets. Get a quote, email me the guide.

Not every label needs a benefit. The aim is to make the next screen predictable. "Send" names an action but not what is sent; "Request a quote" names both the action and the result.

Does the label match the next screen?

After drafting a label, ask one question: can the person who clicks predict what they will see next? If a button says "Get started" and the next page asks for payment, the label has set the wrong expectation. The label and the destination must agree. That agreement is also part of how a landing page keeps its promise; see landing page SEO and conversion balance for the page-level view.

How long a label can run

Short labels scan well, but labels that are too short lose meaning. Two to five words suits most buttons. That is a working guideline, not a rule from any specification, and neither W3C nor Google sets a word limit. Read the label aloud; if the meaning survives, it is long enough.

Example scenario: the mislabeled download

Example scenario: A software vendor's product page has a button reading "Learn more" that actually downloads a PDF catalog. Changing it to "Download the catalog (PDF)" tells the visitor the file type and the outcome, and it tells a screen reader user the same thing when they browse a list of buttons. Whether the change improves conversion is something only your own measurement can show.

Making link text work on its own

Many CTAs are links rather than button elements. For links there is written guidance from both an accessibility and a search angle, and the two fit together well.

WCAG 2.2: link purpose in context

The Understanding document for Success Criterion 2.4.4 (Opens in a new tab) sets a Level A requirement. The purpose of each link must be determinable from the link text alone, or from the link text together with its programmatically determined context, except where the purpose would be ambiguous to users in general. The document defines that context as information available from the markup, such as text in the same paragraph, list item or table cell, or the associated table header cell. It lists benefits: people with motion impairments can skip links they do not need, people with cognitive limitations are less likely to be disoriented by landing somewhere unexpected, and people with visual disabilities can learn a link's purpose from the surrounding text.

The document also notes that context should precede the link, because screen reader users read from top to bottom. It gives "Read more" links in a news summary as an acceptable pattern when the surrounding paragraph supplies the context, and it lists failure techniques, including context that appears only in unrelated content and an image-only link with no accessible name.

Google's advice on anchor text

Google's link best practices (Opens in a new tab) explain that it generally follows links built as an a element with an href. Links that rely on onclick handlers or an href on a span are not reliable. Anchor text belongs between the tags; for an empty link Google may fall back to the title attribute, and for image links it uses the image's alt text, so an empty alt leaves the link without text.

For the wording itself, Google advises anchor text that is descriptive, reasonably concise and relevant to both the current page and the linked page. It gives "list of cheese types" as a good example and calls "click here", "read more", "website" and "article" poor choices. Its test is simple: read the anchor text on its own, and if it makes no sense out of context, make it more specific. It also warns against overly long anchors, keyword stuffing and chains of adjacent links.

Where the two sources meet

The sources seem to disagree on "Read more", but they answer different questions. WCAG sets a floor for accessibility and accepts generic text when context carries the meaning. Google recommends better wording for readers and for search engines. The safest approach satisfies both: write link text that makes sense alone, then let the context reinforce it.

Button or link?

A CTA that takes the visitor to another page is a link and should be an a element with an href. An element that performs an action on the current page, such as submitting a form, should be a button. Keep appearance and behavior in line so keyboard and screen reader users get what they expect. A styled div that imitates a button gives up both behaviors.

Links made of images

When an image is the only content of a link, its alt text acts as the link text. An empty alt leaves the link nameless, which the W3C document lists as a failure. Describe where the link goes or what it does, not just what the picture shows.

Choosing verbs and keeping promises

The verb sets the tone of the label. Strong verbs describe a concrete action such as download, apply, compare or reserve. Weak ones are vague or passive: submit, continue, okay.

Vague labels and their specific replacements

Promises the next screen can keep

A benefit phrase works only as far as you can deliver it. "Free" is right when nothing is charged later; "instant" is right when it is instant. If the next screen contradicts the label, trust drops. Put real conditions, such as a campaign end date, in a short line beside the button rather than stretching the label.

Urgency that has a real deadline

Phrases like "Last chance" create pressure that is misleading when no real deadline exists. If a deadline is real, state it in plain text near the button. Keep the label itself to the action and the object.

One meaning in two languages

English and Turkish labels do not have to be word-for-word translations, but they should describe the same action and the same result. If one language says "Request a quote" and the other says "Learn more", the two pages are making different promises. Write each in natural phrasing for its own audience.

Voice: me, you or neutral

Few button debates last longer than the question of person: should a button speak as the visitor ("Create my account") or address the visitor ("Create your account")? Neither the W3C page nor Google's link guidance states a preference. Both care about whether the purpose is understandable, not about the pronoun, so the answer has to come from your own testing.

First, second and neutral forms

A rule for picking the voice

Decision rule: Let the button speak in the same voice as the rest of the page; a sudden change of person is more noticeable than either choice. The plain imperative is the safest starting point. Test first person against it later, and accept the result only from your own data, because no general finding applies to every audience.

Matching the audience

A formal business-to-business page might read naturally with "Request a proposal"; a consumer app might suit "Try it now". Look at how the rest of the page speaks to the reader and let the button follow. If you have a brand voice guide, check its sample button phrases first.

Several calls to action on one page

A page can carry several CTAs without trouble. Trouble starts when they all look equally important and say the same thing, because neither visitors nor assistive technology can tell them apart.

Choosing the primary action

Give each page one primary CTA and make it visually stronger than the others. Secondary actions, such as "See examples" beside "Compare plans", stay quieter. If two CTAs lead to the same destination, give them the same label; if they lead to different places, give them different labels.

Same destination, same label

Repeating the same CTA near the top, middle and bottom of a long page is fine, since a visitor should be able to act at any point. Identical labels for the same destination give a consistent experience. Identical labels for different destinations, such as several "Details" links, are indistinguishable in a list of links.

Cards that all say Details

Product or article cards that all end in "Details" are a common pattern. Adding the item name to the visible text or the accessible name helps people tell the cards apart in a links list. The W3C document treats text in the same list item, paragraph or table cell as programmatic context. A single link per card, rather than a title link plus a separate button to the same place, also reduces link clutter.

Link volume

Google notes that there is no ideal number of links on a page, but that if a page seems to have too many, it probably does. Apply that to CTAs. Surrounding every section with a button dilutes the emphasis you are trying to create.

Page-level checklist

• Does the page have one clear primary CTA?
• Do CTAs for the same destination use the same label?
• Can links to different destinations be told apart in a list?
• Do repeated "Details" labels carry the item name?
• Is every CTA a real a with href or a real button?

Supporting text and button states

The label should stay short. Small supporting text nearby can handle the doubts the label cannot.

A line of reassurance

A short line under the button can answer the most common worry, for example "No credit card required" or "Cancel any time". Write such lines only when they are literally true under your terms, because a wrong reassurance does more harm than a plain label. For price, duration or cancellation conditions, link to the page that explains them.

Sending and disabled states

When a form is being submitted, disable the button and change the label to something like "Sending" to prevent double clicks and to give feedback. If a button is disabled for another reason, say why nearby; a faded button alone does not explain itself. Define default, hover, focus-visible, active and disabled states, and make keyboard focus visible on every device.

Matching visible text and accessible name

The visible label and the accessible name should match. If the visible text is "Request a quote" and an aria-label says "Submit", a person using voice control cannot activate the button by speaking the words they see. When the visible text is enough, add no aria-label; when extra context is needed, use a longer name that contains the visible words.

Supporting text checklist

• Does the supporting line state a condition that is true?
• Does it agree with your terms and policies?
• Is there a status text while sending?
• Is the reason for a disabled state explained?
• Do visible text and accessible name match?

What each page type needs

The same rule produces different wording on different pages, because the page makes a different promise. The table gives starting hypotheses, all of which should be tested on your own traffic.

Purchase flows

In a purchase flow, each button should describe the move from the current step to the next. "Place order" is right on the last step; used earlier, it makes visitors think payment is imminent. Check each step's button against the screen it opens. If a button can be pressed twice and trigger a double action, protect it on the server too, but that is an engineering concern, not a wording one.

Prompts inside articles

A CTA inside a blog post should relate to what the reader is reading at that moment. A link that continues the topic, such as "See the measurement plan for this method", and a sales prompt such as "Buy now" in the same section weaken each other. Keep persuasive prompts for the end and let the body links inform.

Labels for files

For files, state the type and subject. The W3C document's book-list example shows separate "PDF" and "mp3" links beside each title so the title is not repeated three times and each link purpose remains clear.

Testing variants without hurting search

No single wording is best everywhere, so test directions rather than slogans. Change one variable per test, and let your own results decide.

From hypothesis to measure

Write a hypothesis first: "Naming the result will increase the number of visitors who reach the order page." Decide the success measure in advance, such as a click plus completion of the next step. Do not look at clicks alone. If clicks rise while completion falls, the label may be setting a false expectation.

Staying within Google's testing guidance

Google's A/B testing guidance for search (Opens in a new tab) comes down to four points. Do not show Googlebot different content from what users see; the page calls this cloaking and says it violates spam policies even during a test. Where test variants have their own URLs, point rel="canonical" to the original. If you redirect, use a 302 rather than a 301. And run the test only as long as needed, then remove the variants and scripts; the page warns that unnecessarily long tests, especially ones that send a single variant to a large share of users, could be treated as an attempt to deceive search engines. Changing a button label usually needs no separate URL at all. For a broader method, see SEO split testing without cloaking, and for display ads, banner ad A/B testing sample plan.

Interpreting the outcome

Do not declare a winner before enough data has accumulated, and note outside factors such as traffic source and device. A good result from a small sample may be chance. When you do find a winner, test it on other pages before copying it everywhere.

Rewriting a CTA: a working session

To turn this into practice, run a four-step exercise on your most important page.

1. List every label. Write down each button and link text on the page.
2. Add the destination. Note what each one opens.
3. Read each alone. Read every label out of context and mark those that make no sense, which is Google's own test.
4. Rewrite and compare. Rewrite the marked ones as verb, object and benefit, then compare the versions.

Example scenario: the campaign page rewrite

Example scenario: A course site uses "Start now", "Learn more" and "Click" on a campaign page, each leading somewhere different. The team rewrites them as "Sign up for a free trial lesson", "Browse the syllabus" and "Compare pricing plans". Each now says where it goes. The team measures not only clicks but also how many visitors finish the next step, because a label that raises clicks and sends people to the wrong place is not a success.

Agreeing on criteria

Button text passes through many hands. Cut the debate by agreeing on written criteria: is there a verb, does it match the next screen, does it make sense alone, do both languages say the same thing. When two candidates meet the criteria, decide with a test rather than an argument.

Keeping labels current

Labels age as campaigns, product names and prices change. Every few months, reread the buttons on your busiest pages and remove promises that no longer hold, such as a "first month free" offer that has ended. An outdated promise is one of the most damaging details on a page.

Final review

Use this as a final review before publishing.

For button text

• The label contains an action verb.
• It says what will happen.
• It matches the next screen.
• The person form agrees with the page voice.
• There is no inflated promise or false urgency.
• Turkish and English labels describe the same action.

For CTAs that are links

• The link text makes sense when read alone.
• It is a real a with an href.
• Image links carry meaningful alt text.
• Same destination, same label; different destination, different label.
• Context comes before the link.

For tests

• One variable is changed.
• The success measure was written down first.
• Googlebot and visitors see the same content.
• Variants are removed when the test ends.

Mistakes that keep recurring

• Sentences ending in "click here". Attach the link to the meaningful words instead.
• Using an a to submit a form or a button to change pages.
• The same "Details" label on every card.
• Judging by click rate and ignoring final conversion.
• Applying version-specific details without checking current documentation. Pages from the W3C and Google can change, so read the current text first.

Auditing a page that already exists

To review an existing page, export all button and link labels to a table with the label, destination, position and whether it is primary. Flag generic ones such as "Continue", "More" or "Click", rewrite each with the verb-object-benefit pattern and compare the new label with the screen it opens. Start with the page that gets the most traffic.

Frequently Asked Questions

How long should CTA button text be?

Neither W3C nor Google sets a word limit. Two to five words suits most buttons in practice, but the real test is whether the label makes sense on its own.

Is "Read more" acceptable for accessibility?

WCAG 2.4.4 allows link purpose to come from programmatic context, so a generic label can pass when the surrounding paragraph supplies the meaning. Google, however, lists "read more" among poor anchor texts. Writing the link text so it makes sense alone satisfies both.

Should button text be first person or second person?

The W3C and Google pages state no preference. Match the voice of the rest of the page and compare the options with your own measurements.

Does A/B testing a CTA hurt search rankings?

If you follow Google's guidance it should not: no cloaking, canonical on test URLs, 302 redirects and no longer duration than needed. This is not a guarantee.

Can a page repeat the same CTA label several times?

Yes, if the links go to the same destination. Using the same label for different destinations makes links indistinguishable in a list of links.

What alt text should an image CTA have?

Describe the purpose of the link. Google uses the alt text of an image link as its anchor text, and an empty alt leaves the link with no text.

Related guides: What Is Conversion Rate Optimization (CRO)? First Steps, Social Proof and Reviews on Landing Pages, Lead and Contact Form Design for Conversion.

Conclusion

Strong CTA text pairs an action verb with a clear object and, where it is honest, a benefit, and it stays understandable when read alone. The W3C asks that link purpose be clear from text or programmatic context, while Google prefers descriptive anchor text over generic phrases. Neither sets a rule on first or second person, so decide with your own data. Keep one primary CTA per page, use matching labels for the same destination and different ones for different destinations, and test one variable at a time without cloaking.

If you need text for your pages, including the wording of buttons and links, you can look at the content writing service on Akua Adforce; scope and terms are listed on the product page.