Copy that looks fine on screen can become gibberish when spoken. Screen readers and voice assistants drop bare symbols, expand emoji into their full names, and stumble over run-together acronyms and ellipses. This checker scans pasted UI strings for the patterns that reliably sound wrong when read aloud and suggests a spoken-language rewrite for each.
How it works
Each line is run through a set of rules that detect known read-aloud failure modes:
- Bare symbols —
&,/,>,<,~,|,@used as words, which many engines skip or mispronounce. - Math notation —
<=,>=,x,±, and ranges like5-10that are read as a literal minus. - Run-together acronyms / all-caps — strings like
CSVJSONor shouting that gets spelled out or mangled. - Ellipsis in labels — trailing
...or…, announced as “dot dot dot”. - Emoji — any emoji, each expanded to its verbose spoken name.
For every match the tool returns the affected string, the reason, and a concrete suggested alternative (for example, replacing & with the word “and”).
Example
The label Save & exit → triggers two flags: the bare & (read inconsistently) and the bare arrow → (often skipped). The suggestion is Save and exit, with the visual arrow kept as a decorative, aria-hidden icon rather than part of the accessible name.
Tips
- Keep the accessible name (visible text or
aria-label) free of symbols and emoji; move decoration intoaria-hiddenicons. - Spell out functional symbols: “and”, “per”, “to”, “greater than”.
- Replace status ellipses with real state (
aria-busy, anaria-liveregion) rather than punctuation. - All checks run locally — paste unreleased copy without it leaving your browser.