Tooltip Copy for the Long Game: Writing Hints That Still Make Sense Two Years and Ten Languages Later

    November 20, 2026
    12 min read
    By Callan Whitmore
    Share this article

    Teams usually treat tooltip copy as the overflow tray for an interface: whatever did not fit on the button, the label or the empty state goes into a small dark box that appears on hover. That is the habit that produces bad tooltips. A tooltip that explains a confusing label is a patch on a label that should have been rewritten, and the patch is invisible to every person on a touch screen, every keyboard user on a disabled control and every user who never thinks to hover.

    Good tooltip copy starts from the opposite assumption: most people will never read it. The interface has to work without it. The tooltip adds one thing the visible text cannot carry, in a line short enough to read before the pointer moves on. The rest of this guide is built on that assumption.

    Most tooltips patch a label that should have been rewritten

    Before writing a word, decide whether the element needs a tooltip at all, and which kind. Four patterns get called "tooltips" in design reviews, and they have different rules.

    Tooltip: a short text label that appears on hover or keyboard focus and disappears when the pointer or focus leaves. It cannot contain interactive content. In the WAI-ARIA Authoring Practices it is linked to its trigger with aria-describedby, so screen readers announce it as a description of the control.

    Toggletip: a small bubble that opens on click or tap and stays until dismissed. The term was popularised by Heydon Pickering in "Inclusive Components". It works on touch devices and can be announced through a live region.

    Popover or rich tooltip: a larger surface with a title, body text and sometimes a link or button. Material Design 3 calls this a rich tooltip and separates it from the plain tooltip that holds one short line.

    Coachmark: an onboarding hint that points at a feature without being triggered by the user. It is marketing for a feature, not help for a control.

    Copy rules differ by pattern. A plain tooltip holds a fragment or a single sentence. A toggletip can hold two or three sentences. A popover can have structure. Writing popover copy into a plain tooltip is the most common reason tooltips feel cramped.

    Then ask whether the problem should be solved somewhere else. Work through these in order:

    1. Is the trigger an icon without a visible label? If yes, a plain tooltip with the control's name is appropriate, and the name should also be the accessible name.
    2. Could the visible label be rewritten so the tooltip is not needed? If yes, rewrite the label. "Sync" with a tooltip saying "Send local changes to the server" is often better as a button that says "Upload changes".
    3. Is the information needed to complete the task, such as a format rule for a password? If yes, put it in visible helper text, not a tooltip. Required information that hides behind hover gets missed.
    4. Does the explanation need more than one sentence or a link? If yes, use a toggletip or popover.
    5. Is the hint aimed at teaching a new feature once? If yes, it is a coachmark or a changelog item, and it should not live permanently on the control.
    6. None of the above? Then a short tooltip that adds one fact the label cannot carry is justified.

    Following that sequence on an existing product tends to remove a large share of tooltips. That is the point. Every tooltip that survives is one where hover genuinely is the right moment.

    Icon labels deserve a note. The Nielsen Norman Group has argued for years that icons without text labels are hard to interpret, and that tooltips on icons are a fallback, not a fix. In a dense toolbar where text labels do not fit, a tooltip with the control's name is still the standard pattern, and the copy there is simple: the name of the action, in the same words used in menus and documentation.

    What a tooltip line has to do in under two seconds

    A hover tooltip competes with pointer movement. People hover while moving toward something else, and the tooltip has a short window before the pointer leaves. That window sets a word budget.

    Adult silent reading speed for English non-fiction averages about 238 words per minute, according to a 2019 meta-analysis by Marc Brysbaert. That is roughly 4 words per second for continuous text, and slower for a first glance at an unfamiliar UI string because the eye has to find the box first. A working budget:

    words allowed = (hover window in seconds - 0.5 s to locate the box) x 4 words per second
    
    Example: 2.0 s hover window
    words allowed = (2.0 - 0.5) x 4 = 6 words
    
    Character check for plain tooltips:
    characters ~ words x 6  (average English word plus space)
    6 words ~ 36 characters, which fits one line of a 240 px tooltip at 12 to 14 px type
    

    Treat the 2 second window as an assumption to test, not a law. Users who hover deliberately on an unfamiliar icon will wait longer. But a tooltip that needs 15 words is asking for deliberate reading, which is the job of a toggletip.

    Within that budget, the line has to add information. Four patterns cover most good tooltips:

    The name of an unlabelled control: "Bold", "Duplicate frame", "Copy link". Use the verb form people see elsewhere in the product.

    The consequence the label hides: on a "Publish" button, "Visible to everyone with the link". The label says the action; the tooltip says what changes in the world.

    The constraint behind a state: on a greyed export option, "Available on files under 50 MB". This explains why, so the user can do something about it.

    The keyboard shortcut: "Undo (Ctrl+Z)". Short, and useful to exactly the people who hover.

    What not to write is just as specific. Do not repeat the label: a button saying "Settings" with a tooltip saying "Settings" adds nothing but noise for screen reader users, who hear the same word twice.

    Do not open with "Click here to" or "This button lets you": the user is already looking at the button. Do not put marketing copy in a tooltip ("Supercharge your workflow"). And do not use the tooltip to apologise for a design decision ("Due to technical limitations..."); give the constraint and the workaround.

    A short before and after set, from a fictional file manager, shows the difference:

    • "This button allows you to share the file" becomes "Anyone with the link can view".
    • "Info" on an info icon next to "Storage used" becomes "Includes files in trash".
    • "Click to sync your files with the cloud" becomes "Last synced 3 min ago".

    Each rewrite moves from describing the control to telling the user something they did not know.

    Eight steps from trigger to shipped string

    Treat tooltip copy as content with a process, especially in a product that has hundreds of them. This sequence works for a single new tooltip and scales to a full audit.

    1. Write down the trigger element, its visible label and its state. A tooltip on an enabled "Export" button and one on a disabled "Export" button are different strings with different jobs.
    2. Run the six questions from the first section. Record the outcome: tooltip, toggletip, helper text, label change or nothing.
    3. State the one fact the tooltip adds, in a note to yourself, before writing the copy. If you cannot name a single fact, the tooltip is not needed.
    4. Draft the line within the word budget. Write three versions and pick the one that starts with the most important word.
    5. Check vocabulary against the product's terminology list. If menus say "Duplicate", the tooltip does not say "Copy".
    6. Read it aloud with the visible label in front of it, the way a screen reader will announce it. "Share button, Anyone with the link can view" should sound like one coherent thought.
    7. Add a context note and a screenshot for translators, and set a maximum length for the translated string. If your design system's tooltip has a maximum width, give translators the character count for that width.
    8. Log the string in a tooltip inventory with its trigger, owner and the date it was written, so it can be reviewed when the feature changes.

    Step 3 does most of the work. Designers often draft a tooltip and then look for what it says. Reversing the order forces a decision about value before any words are spent.

    Step 7 is where long-term quality is won or lost. Localisation teams commonly plan for German and French UI strings to run about 20 to 35 percent longer than English, and that rule of thumb is enough to break a 6 word tooltip that sat at the edge of a single line. A tooltip that wraps to three lines in German is still readable. One that truncates with an ellipsis is not, and truncated tooltips are a common bug in localised design systems.

    A worked illustration with numbers: a design system's plain tooltip has a maximum width of 240 px and 12 px type, which fits about 38 characters per line at typical letter widths. The English string "Includes files in trash" is 23 characters.

    A 30 percent expansion takes it to 30 characters, still one line. "Anyone with the link can view this file" is 39 characters in English and about 51 after expansion, so it wraps to two lines in German. That is acceptable if the component allows two lines. If the component is fixed at one line, shorten the English to "Anyone with the link can view" (29 characters) before it goes to translation, not after.

    Tooltips over two years: inventory, accessibility and drift

    Tooltip copy decays faster than other UI text because nobody sees it during routine reviews. Screenshots in a design review show the default state, not the hover state. A feature gets renamed in navigation and menus, and the tooltip on its icon keeps the old name for a year. A limit changes from 50 MB to 100 MB, and the tooltip on the disabled export button still says 50.

    The defence is boring and effective: keep an inventory. A spreadsheet or a field in your content management tool with every tooltip string, its trigger, the feature it belongs to and its owner. When a feature changes, the product manager or content designer filters the inventory by feature and reviews every hover string. Teams that store UI strings in a localisation platform such as Lokalise, Phrase or Crowdin can tag tooltip keys and run the same filter there.

    Accessibility requirements also shape the copy over time. WCAG 2.1 added success criterion 1.4.13, Content on Hover or Focus, which requires that content appearing on hover or focus is dismissible without moving the pointer, hoverable so the user can move the pointer onto it, and persistent until the user dismisses it or the trigger loses focus.

    That is a component requirement rather than a copy rule, but it changes what copy is reasonable: a tooltip that disappears in 1.5 seconds on a timer fails the criterion, so you cannot rely on a timer to hide a long message.

    Keyboard and screen reader behaviour matter for wording. With aria-describedby, the tooltip text is announced after the control's name and role. Starting the tooltip with the control name ("Share: anyone with the link can view") makes the screen reader say "Share" twice. Start with the new information instead.

    Touch screens are the hardest limit. Hover does not exist on a phone, and long press is a poor substitute because most users never try it. Any tooltip that carries information a mobile user needs has to move into visible text or a toggletip on small screens. Responsive products often end up with two strings for the same help: a 6 word tooltip on desktop and a visible helper line on mobile. Keep both in the inventory, linked to each other.

    There is a fair counterargument to all this process. Small teams with 40 tooltips in total can review them by opening the app, and a formal inventory feels like bureaucracy. That holds until the product is translated into a second language or the team adds its second designer. Past either point, the cost of an inventory is an hour a quarter, and the cost of not having one is stale copy shipped to every market.

    Measurement is possible, though rarely done. Some analytics tools record hover events, and tooltip triggers can be tracked like any other interaction. A tooltip that is hovered on 40 percent of sessions on a key screen is a sign that the visible label is unclear; the long term fix is the label, not a better tooltip.

    What designers ask about tooltip copy

    Should tooltip copy end with a period? Pick one rule per product and apply it everywhere. A common convention is no period for fragments such as "Export as CSV" and a period for full sentences, so a tooltip with two sentences always gets both periods.

    Can a tooltip contain a link? Not in a hover tooltip. Under the WAI-ARIA pattern a tooltip cannot receive focus, so a link inside it is unreachable by keyboard. Use a toggletip or popover that opens on click and can hold focus.

    How do I write tooltips for disabled buttons? Disabled buttons usually cannot receive focus, so a tooltip attached to them is invisible to keyboard users. Keep the button focusable with aria-disabled, or show the reason as inline text next to the control.

    Do tooltips need to be translated separately from the UI? They go through the same pipeline but need their own context notes for translators, because a string such as "Sync" with no screenshot can be translated as a noun or a verb and both will be wrong somewhere.

    Contextual help is moving away from the hover box. Products increasingly answer "what does this do?" through inline helper text, command palettes and in-app assistants that respond to a typed question. That trend will shrink the tooltip back to its oldest job: naming an icon and showing its shortcut. Teams that already write tooltips as one fact in six words will barely notice the shift.

    Related Articles