Tweet Rewriter
for the draft that isn't landing.
Paste what you were going to post. You get three rewrites — the point up front, the same idea shorter, and a concrete version — each one ready to copy. Free, no signup, and nothing is stored.
Most drafts fail in the first line.
A post gets about one second. Someone is moving their thumb, your first line arrives, and they decide before they have read the second one. Almost every draft that underperforms lost it there — not because the idea was weak, but because the idea was third in line behind a wind-up, a hedge and a bit of context nobody asked for.
That is what this page is actually fixing. It is not a thesaurus and it is not a grammar pass. It takes the thing you were trying to say, finds where you buried it, and puts it in front.
The three rewrites, and when to take each one.
- Sharper. Roughly the length you wrote, with the point moved to the front and the throat-clearing gone. Take this one when you like your draft and it just is not landing.
- Shorter. The same idea with every word that was not load-bearing removed, often down to one line. Take this one when the draft rambled, or when you are posting into a fast feed.
- More concrete. Leads with the specific thing already in your draft — the number, the name, the moment it actually happened — and cuts the abstraction wrapped around it. Take this one when your post is true but could have been written by anyone.
If all three come back looking alike, that is usually a signal about the draft rather than the tool: there was one idea and one way to say it, and you were already close.
The rules it writes under.
These are the same rules Penwell drafts under for paying customers, not a lighter set for the free page. They exist because AI-written posts have a smell, and the smell is made of specific, listable things.
- No em-dashes, emoji or hashtags. The three loudest tells. Forbidden in the instructions and stripped again afterwards, because models break that rule even when told not to.
- No unlock, transform, elevate, revolutionary. Marketing verbs make a claim sound smaller, not bigger. Same for exclamation marks.
- No invented facts. It cannot add a number, a name, a date or a result that was not in your draft. This is the rule that matters most: a rewrite that sounds better because it made something up is a post you have to delete.
- Never speaks as an AI. Anything that identifies its own writer as a model gets caught and dropped rather than shown to you. A post that outs your account as bot-written is the one mistake with no recovery.
- One read, no second pass. Simple words, short sentences, line breaks where the thought actually turns. If a line needs reading twice it gets rewritten.
What it is not.
It is not writing in your voice. It has never seen your account, so it writes in a house voice: plain, specific, undecorated. That is a good default and a poor impersonation, and pretending otherwise would be the easiest lie on this page. Learning how a particular person writes takes their actual posts, which is the part that happens after you connect an account.
It is also not a volume machine. Rewriting one post you already cared enough to write is a different activity from generating forty you did not. The second one is why so much of this app now reads like a press release.
Frequently asked questions.
Is the tweet rewriter free?
Yes. No account, no card, no email. Each rewrite is a real call to a large model and costs us real money, so there is a daily limit per visitor — fifteen, which is more posts than almost anyone writes in a day. Nothing you paste is stored or shown to anyone.
Why three rewrites instead of one?
Because the three things wrong with a draft are almost never the same thing. Sometimes the point is buried, sometimes it is twice as long as it needs to be, and sometimes it is true but too vague to stop anyone scrolling. Picking which of those applies before you have seen any of them is guesswork. Three costs the same single call, so you read all three and take the one that fits.
Will it change what I actually said?
No. Your claim, your stance and your facts carry over exactly. If your draft says three weeks, all three rewrites say three weeks. It is also blocked from inventing a number, a name, a date or an outcome that was not in what you pasted — a vague draft gets sharper wording, never a fabricated detail to make it sound better.
Does it add em-dashes, emoji or hashtags?
Never. Those three are the loudest tells that a post came out of a chatbot, and a rewrite carrying any of them is worse than the draft you started with. The instructions forbid them and a second pass strips them anyway, because models break that rule even when told not to. Em-dashes become commas so the sentence keeps its pause instead of collapsing into run-on.
Can I use the rewrites as they are?
That is the whole point. Every rewrite comes back as a finished post: no notes on what changed, no bracketed alternatives, no placeholders to fill in. Each one shows its character count against X's 280 so you can see it fits before you copy it.
Does it write in my voice?
Not on this page. It has never seen your account, so it writes in Penwell's house voice — plain, specific, no decoration. That is a good default and a bad impersonation. Voice matching needs your actual posts to learn from, which is what happens after you connect your X.
Does it work on threads or long posts?
It rewrites one post at a time and aims at the 280-character shape, so a thread is better handled a post at a time. You can paste up to 2,000 characters of draft — useful when the draft is long and the job is cutting it down — but what comes back is a single post, not a thread.
Is anything I paste saved?
No. The draft is sent to the model, the rewrites come back, and nothing is written to a database. Refresh the page and it is gone. We keep the token count for the bill, not the text.
One post is easy. It's the next two hundred that need help.
Continue with X