Free tool · runs in your browser

AI writing checker

Paste a post, a reply or an email. The phrases that make it sound AI-written are marked in place. Fix those, and most readers stop noticing. Nothing you type leaves this page.

161 / 280

In the extension, any draft with obvious model phrasing is blocked, and the ones that pass are ranked so you see the most human first.

The tells it looks for

These are the patterns readers pick up on first. Any one of them is fine once in a while. It’s the pile-up that gives text away.

TellWhat it looks likeWhat to write instead
Praise opener“Great question!”, “Love this!”, “So true.”Start with your answer. If you agree, say what you’d add.
Formula“The key to growing is…”, “Here’s the thing:”, “Let’s dive in.”Skip the announcement. Say the thing.
Not X, it’s Y“It’s not just a tool, it’s a mindset.”Say what it is. Leave out the version nobody believed.
Tidy moral“At the end of the day…”, “Let that sink in.”, “You got this!”End on your last real point and leave it there.
Buzzworddelve, leverage, elevate, unlock, game changer, seamless, landscapeUse the word you’d say out loud to a friend.
Engagement bait“Thoughts?”, “Agree?”, “Drop a comment below.”Ask a real question, or none at all. Good posts draw replies on their own.
Hype emoji🚀 ✨ 🔥 👇 💯 at the end of linesDrop them, or keep the one that carries meaning.
Em dash (soft)Several — in a short postSwap most of them for a full stop or a comma.

How to make AI writing sound human

Fixing the tells is the first pass. The second pass adds what generated text usually lacks: a person behind it. These edits make the biggest difference, listed in the order worth doing them.

1. Lead with the point

Models warm up. They praise the question, restate it, then answer. People who know the answer start with it. Delete the first sentence of most AI drafts and they improve right away.

2. Put something only you know into it

A number from your own work, the tool you actually use, the mistake you made last month. Generated text is built from what’s typical, so one specific detail does more than any amount of rewording.

3. Take a side

AI writing hedges: “both have their strengths”, “it depends on your needs”. Say which one you’d pick and why. If you wouldn’t pick either, say that.

4. Break the rhythm

Generated paragraphs are built from sentences of the same length, often in threes. Read the text out loud. Where it sounds like a list, make one sentence very short. Or join two.

5. Cut the ending

The last line of an AI draft is usually a summary, a moral or a question for the reader. Most of the time, the text is stronger if it stops one sentence earlier.

6. Write the way you talk on X

If you never use capital letters in replies, a perfectly capitalised reply looks borrowed. Match your own habits: your punctuation, your usual length, the words you reach for.

PostInk does this on every draft it writes for X. It learns your style from posts you wrote yourself, discards drafts with these tells, and shows you the one that reads most like you. If replies are what you need most, the AI reply generator page shows real examples.

Questions

Is this an AI detector?

No. An AI detector guesses whether a model wrote a text, and those guesses are often wrong about human writers. This checker does something narrower, and more useful when you’re editing. It marks the specific phrases that make writing sound generated, so you can change them.

Is the text I paste sent anywhere?

No. The check runs in your browser against a short list of patterns. Nothing you type is uploaded, stored or logged.

If it finds no tells, does that mean the text is human?

It means the text avoids the most common patterns, and those are what readers notice first. Writing can still sound flat for other reasons: no point of view, nothing specific, every sentence the same length. The guide below covers those too.

Is this the same check PostInk uses?

It’s a shorter version. PostInk’s own check runs inside the extension on every draft. It looks for more patterns and also scores how human each draft reads, so the best one is the one you see.

Are em dashes really a sign of AI writing?

On their own, no. Plenty of good writers use them. They’re marked as a soft tell because generated text uses them far more often than people do. Three in a short post are worth a second look.

Keep reading

Stop fixing drafts by hand

PostInk runs a fuller version of this check on every draft it writes for X, in your voice. You only see the ones that pass.

$3.99 once, after 7 days free