Clean hidden text.Highlight AI writing patterns.
Two separate checks: exact cleanup for known hidden Unicode markers, plus an explainable review of visible writing patterns. Neither is proof of AI authorship.
Exact Unicode checks. Nothing is uploaded or saved.
Writing pattern review
Visible style clues inside the LinkedIn preview—not an authorship verdict
Your analysis appears here
Paste a post or load the example to reveal hidden formatting and writing patterns.
Cleaning removes artifacts. Voice makes the post yours.
Postiv learns from your real posts, then helps you rewrite without the generic tone.
What it actually checks
Three different problems, three honest answers
“AI watermark” is used for several unrelated technologies. Combining them into one magic pass or fail result is misleading. This cleaner keeps the layers separate.
Known Unicode markers
We inspect exact code points in a published ruleset, name each match, remove unwanted characters, and count the result.
Visible writing patterns
We highlight observable structures associated with generic AI prose. Each flag includes evidence and a practical edit, never an authorship verdict.
Claude embedded marks
Anthropic says detection support for users and third parties is coming. Until it publishes the mechanism, a public regex cannot verify the mark.
Current Claude marking status is based on Anthropic's official documentation, last checked August 2026.
How the detector works
A field guide turned into transparent checks
The audit is based on observable categories documented by Wikipedia's AI Cleanup editors. Their guide explicitly says these are potential signs, not proof. It also warns against relying on automated detectors alone.
Read the source field guideSpecificity vs abstraction
Broad promises are flagged when a concrete action, example, or result would be stronger.
Repeated templates
Repeated sentence openings, tidy triads, and recurring contrast structures.
Unsupported grand claims
Turning points, larger movements, and lasting impact without supporting evidence.
Overly balanced rhythm
Several similarly sized sentences can make a draft feel generated or over-edited.
Missing personal detail
Long, abstract drafts with no observation, number, or quoted example get a document-level note.
Repeated summaries
Packaged conclusions that restate the post instead of adding a consequence or next step.
Frequently asked questions
AI watermarks, detectors, and LinkedIn
Can this detect or remove Claude’s new text watermark?+
Not yet. Anthropic says supported new Claude models embed an imperceptible, machine-readable text watermark and that users and third parties will be able to detect it. As of August 2026, Anthropic has not published the detector, API, technical specification, or verification keys. This tool can verify and remove the known hidden Unicode markers in its ruleset, but it does not claim to detect Claude’s unpublished watermark.
Does LinkedIn detect AI-written posts?+
LinkedIn has not published a detector that labels ordinary text posts as AI-written. Readers can still recognize generic patterns, repeated formats, vague claims, and chatbot artifacts. The pattern audit helps you find those visible issues before they weaken trust or make the post sound like everyone else.
Does a highlighted pattern prove that AI wrote the post?+
No. The highlights identify editable style clues such as abstract claims, repeated sentence templates, unsupported grand claims, unusually even sentence rhythm, missing concrete detail, and repetitive summaries. Humans can write this way and AI can avoid these patterns, so the audit is writing feedback rather than an authorship detector.
What hidden characters does the cleaner remove?+
The cleaner removes zero-width spaces, soft hyphens, word joiners, byte-order marks, bidirectional controls, direction marks, Unicode tag characters, and similar copy-paste artifacts. It also converts unusual spacing characters into regular spaces and trims broken trailing whitespace. It preserves the zero-width joiner used inside many emoji sequences.
Will the cleaner rewrite my LinkedIn post?+
No. The free cleaning step only changes hidden characters and formatting that do not carry your meaning. The audit gives specific rewriting suggestions, but it does not silently replace your words. If you want a full rewrite, Postiv can learn from your existing posts and write in your voice after you choose to sign up.
Is my LinkedIn post uploaded or stored?+
No. The cleaner and pattern audit run inside your browser. The text is not sent to Postiv, stored in a database, or used for model training. Copying or refreshing the page is under your control.
Can a clean result prove that a post was written by a human?+
No. Humans use common AI-associated patterns, and AI can produce writing without them. Wikipedia itself warns that these patterns are descriptive clues, not proof. A clean writing review only means the tool found fewer listed patterns; it does not verify human authorship or the absence of a vendor watermark.
How should I make an AI-assisted LinkedIn post sound more like me?+
Replace generic praise with details only you know, include a real decision or observation, vary your sentence rhythm, and remove conclusions that merely restate the post. The strongest option is to compare the draft with posts you wrote yourself and restore your usual vocabulary, opinions, and level of directness.
Privacy note: the free cleaner runs entirely in your browser. Postiv only receives text if you later choose to use an account feature that requires generation.
Go beyond cleanup
Make every post sound unmistakably like you
Postiv learns from your real writing, then helps you turn rough ideas into LinkedIn posts in your voice.