Voice Check
Paste your writing. See which phrases sound like everyone else.
This is not an AI detector. People write these phrases too; AI just writes them a lot. Use it to find the sentences that sound like everyone else.
Your text never leaves your browser.
Result
Your text, marked
Select a highlight with a click or the Tab key to see why it was flagged.
Whole-text signals
Every flag
Write where your voice stays yours
Galascribe is a writing platform built to keep your own voice on the page.
Open GalascribeHow this works
Everything runs in your browser. The page matches your text against a list of phrases and patterns, then works out a score.
| Part | Points lost |
|---|---|
| Flagged phrases | 8 points for every flagged phrase per 100 words, up to 60 |
| Em-dash density | Only when there are 2 or more and at least 0.8 per 100 words: 5 points per dash per 100 words, up to 10 |
| Lists of three | Only when there are 2 or more and they appear in at least 1 of every 5 sentences: 3 points each, up to 10 |
| Even sentence length | 15 points when there are 6 or more sentences and their lengths vary by less than 30% of the average |
The score is 100 minus the points lost, kept between 0 and 100. "X% sounds like you" is that number. It is a rough guide to how many familiar phrases you used, nothing more.
Common tells of generic writing
Generic writing is not always written by a machine. It is writing that could belong to anyone. A few habits cause most of it.
Stock vocabulary. Big, abstract words such as "delve", "tapestry", "landscape", "realm", "seamless" and "robust" appear in many drafts. They sound grand but hold no detail. Phrases like "a testament to" and "in today's fast-paced world" do the same job.
Stock structures. "It's not just X, it's Y" and "Not X. Not Y. Just Z." set up a claim only to correct it. A question that opens a paragraph, or a line such as "Here's the thing", puts a pause before the point.
Filler and hedges. "It's important to note", "in conclusion" and "at the end of the day" announce a point instead of making it.
Chatbot residue. Lines such as "Great question" or "I hope this helps" are meant for a chat window, not for a reader.
Patterns across a text. Many dashes, many lists of exactly three items, and sentences that all run to the same length make the rhythm feel flat. So do stacked intensifiers such as "very", "really" and "truly".
None of these is an error. Each is a sign that a phrase was chosen by habit. The fix is usually to say something only you could say: a name, a number, a small detail from your own week.
Phrase lists here draw on the Wikipedia guide Signs of AI writing, used under CC BY-SA.
Questions
Is Voice Check an AI detector?
No. It looks for phrases and patterns that show up often in generic writing, whoever wrote it. People use these phrases too. It cannot tell you who or what wrote a text, and it should not be used as proof of that.
Does my text get uploaded or stored?
No. The analysis runs in your browser. Your text is never sent anywhere and nothing is stored. Like the rest of the site, this page can count visits with Google Analytics if you accept the choice it offers, and that count never includes your text.
How is the score worked out?
The score starts at 100. It loses points for each flagged phrase per 100 words, for heavy em-dash use, for many lists of three, and for sentences that are all about the same length. The How this works section shows the exact numbers.
Why was my own writing flagged?
Because people write these phrases too. A flag means the phrase is common, not that it is wrong. Keep it if you chose it on purpose. Change it if you reached for it out of habit.
How do I make my writing sound more like me?
Replace general claims with specific ones: a name, a number, a place, something that happened to you. Cut sentences that announce a point instead of making it. Read the draft aloud and change anything you would not say.
Voice Check is free. Rules are a starting list and will miss things and flag fine sentences.