Your post is written. But is it any good?
Paste your draft, pick the network: you get a detailed score, the list of what actually goes wrong, three spare hooks and a rewrite. In about twenty seconds, before you publish.
Your diagnosis will appear here
Overall score, score per dimension, named flaws — and the fixed version if you leave the AI box ticked.
In short — A post is judged on four things: the hook (the one line 100% of your audience will see), the structure (nobody reads a wall of text), the style (short sentences, zero jargon) and the voice (someone is speaking, with specifics). This tool scores those four dimensions from 0 to 100, names the exact flaws in your text and suggests fixes. Free, no account, for LinkedIn, X, Instagram, Threads and Facebook.
By Sébastien de Bollivier · Updated on
The topic is almost never the problem. The problem is that the first line stops nobody, the text lands as one block, and nothing shows a human lived it. All three are visible before you publish — you just need someone to name them. That is all this tool does: it reads your draft the way a busy reader will, and tells you where you lose them.
How it works
Three moves, about twenty seconds.
1. Paste your draft
Your text as-is, line breaks and emoji included. Nothing to connect, no account to create.
2. Pick the network
LinkedIn, X, Instagram, Threads or Facebook: the useful length and the preview size differ, so the score does too.
3. Fix what gets named
Every dimension gets its score and its precise flaws. The AI adds a verdict, three spare hooks and a full rewrite.
Same subject. Two texts.
Both posts say the same thing. The scores below are the ones the tool actually returns — paste them to check.
I am thrilled to announce that our company is pursuing its digital transformation with an innovative solution that brings real value-add to our whole ecosystem, in a proactive and holistic approach. #innovation #digital #transformation #leadership #business
I lost $12,000 on a single client. Not because he never paid. Because I said yes to everything for 6 months. What I changed since: — scope in writing before the first call — 40% upfront — any out-of-scope request gets a quote Zero unpaid invoices in 14 months. Do you invoice before or after?
What the tool looks at
Four weighted dimensions. The hook carries the most.
Hook — 35% of the score
Does your first line fit inside the preview before the "see more" fold? Does it open on an empty formula ("I am thrilled to announce")? Does it carry tension — a number, a question, a contrast? It is the only line everyone sees.
Structure — 25%
Length fitted to the network, and room to breathe. A 900-character block with no blank line gets scrolled past, whatever it says.
Style — 20%
Average sentence length, corporate jargon, hashtag and emoji overload, shouting caps. These are the markers that read as "press release" instead of "someone is talking to me".
Voice — 20%
Are you speaking in the first person? Is there anything specific — a number, a date, a real case? And an honest exit rather than engagement bait, which caps the score outright.
Frequently asked questions
Is this just ChatGPT with a logo?
No. The score and the named flaws come from a deterministic analysis: same rules for everyone, same text means same score, no model call at all. The AI only comes in as a second layer, for the verdict, the hooks and the rewrite — and if it is unavailable, the score still shows.
Is my text stored or reused?
No. The draft is analyzed on the fly to produce the answer, then forgotten: nothing is written to a database, nothing trains a model, and there is no account to attach it to.
Does a high score guarantee a good post?
No, and nobody can promise that. A high score means the mechanical flaws that lose readers are gone. Whether the topic matters to YOUR audience is not something a form can measure.
Why does the score change per network?
Because the rules change: 280 characters on X versus 3,000 on LinkedIn, a 125-character preview on Instagram versus 210 on LinkedIn, and hashtag tolerance running from 2 to 10. The same text is not worth the same everywhere.
How many analyses can I run?
As many as it takes to rework a draft, with no account and no card. A per-IP limit exists only to stop automated abuse — normal human use never reaches it.
What if I disagree with a remark?
Good: this is a tool, not a judge. The rules are deliberately explicit (hook length, density, jargon) so you can overrule them knowingly. A post that breaks a rule on purpose is often the best one.
When X changes its algorithm, you'll know
I re-read the source code X published whenever it moves. If a weight actually changes, you get an email with the diff. No weekly newsletter, no filler.
An address, nothing else. One-click unsubscribe, any time.
Analyzing is good. Publishing is better.
Purrplan schedules your posts across eleven networks, from the web, your phone or your AI agent. Free 7-day trial.
Try Purrplan