Free Spell Checker

Paste text and get back a version with every misspelling repaired — not just non-words like 'definately' or 'seperate', but the real-word errors a standard red underline waves straight through: homophone swaps such as 'peak your interest' for 'pique your interest', or 'casual' typed where 'causal' was meant.

Spelling carries outsized weight in how readers judge you; they'll forgive a clunky phrase long before they forgive 'accomodate'. This tool scans both letter patterns and sentence context, fixing what's wrong while leaving vocabulary you chose on purpose exactly where it stands.

0 chars · free · no signup

What This Tool Catches

Homophone confusion
Context-aware catches for sound-alikes that dictionary lookups can't judge: their/there/they're, principal/principle, complement/compliment, stationary/stationery, and discreet/discrete — decided by meaning in the surrounding sentence, not by letter pattern alone.
Doubled and dropped letters
The silent killers of credibility: accomodate → accommodate, embarass → embarrass, occurence → occurrence, plus transpositions like beleive and calender that slip past hurried proofreading.
Real-word typos
Every keyboard's worst output — form/from, quiet/quite, then/than, loose/lose — where the misspelling is itself a valid word. These survive every conventional checker because there's no red squiggle to catch them; only reading the meaning does.
UK vs US variant mixing
Flags documents that drift between colour and color, organise and organize, centre and center mid-stream — a consistency problem editors spot late and readers notice immediately.
Suffix and inflection errors
Applies the rules people half-remember: -y to -ies before plurals (company → companies), doubled final consonants (runing → running), dropped e before -ing (hopeing → hoping), and correct -ible/-able endings.
Proper noun protection
Brand names, place names, surnames, and domain-specific terms get flagged as questions instead of auto-'corrected' — so 'Zendesk' never becomes 'Zen desk' and medical terminology survives intact.

How It Works

1Paste the text you want scanned — an email, blog post, application form, product description.
2Click Check Spelling and receive the corrected version.
3Review corrections grouped by type: non-word typos, real-word confusions, suffix errors, and British/American inconsistencies.
4Copy back everything, or reject fixes on names and technical terms the tool flagged as questions rather than changed outright.
5Keep a note of your recurring error types from the grouped report — most people misspell the same dozen words for years without noticing.

Honest Limitations

A deterministic dictionary checker has one great virtue — it never hallucinates a correction — and one great weakness: it cannot tell 'their' from 'there', because both are words. Browser extensions add live underlining but want an install and increasingly an account. This AI-based checker reads context, which is the only way real-word typos and homophone traps get caught at all; the honest trade-off is that it could occasionally second-guess an unusual proper noun, which is why ambiguous items arrive as questions rather than silent rewrites. For maximum rigor on critical documents, run this first, then skim the flagged list yourself.

Context is the difference between checking letters and checking language

Conventional checkers compare strings against a wordlist; this one reads the sentence first. That's why 'The principle reason' gets corrected when 'the principal reason' is meant — both spellings are legitimate, but only one fits — and why UK/US drift gets caught even when every individual word passes a dictionary test.

Corrections come back sorted by error category rather than as an undifferentiated red mess. Seeing three suffix errors and two homophone swaps tells you something useful about how you draft; a flat list of twelve fixes tells you nothing except that you need coffee.

Before

The principle advantage of the new rota is it's definately faster, and staff seem less fatigued accross long shifts.

After

The principal advantage of the new rota is it's definitely faster, and staff seem less fatigued across long shifts. [3 fixes: principle → principal (homophone, context: 'the ___ advantage'); definately → definitely; accross → across]

Sample corrections returned

Pro Tips From Heavy Users

Proofread proper nouns separately from prose — read them backwards or in isolation so your brain stops auto-completing familiar names.
After autocorrect touches your text, hunt homophone trios first; they're the errors autocorrect actively creates, not just misses.
Keep a personal top-ten list of your repeat offenders from the grouped report and drill those ten instead of re-checking everything forever.
Choose one English variant per document before drafting — retrofitting UK↔US consistency later is where most mix-ups sneak through.
Never trust any spell checker with names, dosages, or account numbers; verify those against the source document by eye.

Frequently Asked Questions

What is a real-word typo and why do normal checkers miss it?

It's when you type a valid but wrong word — 'from' for 'form', 'quite' for 'quiet'. Wordlists can't flag it since the spelling exists; catching it requires understanding what the sentence means.

Will it change my company or character names?

No — unfamiliar capitalized terms are surfaced as questions for you to confirm, never silently rewritten into dictionary words.

Which English variant does it enforce?

Whichever dominates your text. If it detects British spelling overall, isolated American variants get flagged as inconsistencies rather than converted wholesale.

Can it check French, Spanish, or other languages?

English only at present. Other languages produce unreliable suggestions here and deserve a dedicated native-language checker.

Does it also fix grammar while it's at it?

Spelling strictly — agreement, punctuation, and phrasing are left for the grammar and punctuation tools so each fix stays explainable and reviewable.

Is there a length cap per check?

Multi-thousand-word documents work fine; anything longer may be processed in chunks, with variant detection kept consistent across all of them.