Does your draft read as machine-generated?
Paste a README, a changelog, a PR description, a blog post, an investor update or a LinkedIn post. Your browser
finds every pattern that makes prose read as machine-generated - words like delve,
seamless and pivotal, em dashes, hedge stacks, "it's not X, it's Y", chatbot sign-offs, vague
"experts believe", leaked citation tokens and utm_source=chatgpt.com links - measures the rhythm of
every sentence, and marks each hit in your text. Quotations and code are never flagged. Free, before you sign in,
and nothing is uploaded. Then the audit lane answers every flag and finds what a pattern cannot,
and the rewrite lane fixes it without adding a first person, a dash or a fact you never wrote -
and the page rescans the rewrite.
The companies and people in the examples are invented. Three are audits (the investor update also carries its rewrite, one click after the audit), two are rewrites; each comes with a saved model reply, so you can see the whole page without signing in or spending a credit.
What this does, and what it does not
The check is pattern matching, not a detector: no model runs in it and it classifies nothing. Published audits found machine-text detectors flagging most essays by non-native English writers, so every flag here is a writing-quality signal with its evidence quoted, weighted by severity (P0 credibility killers, P1 obvious smell, P2 polish) and adjusted for the context you write in - a LinkedIn post may carry two dashes, a technical post may say "robust", a changelog may be terse. Rhythm counts most: a draft with every flagged word replaced and every sentence still 15 to 25 words long reads the same. Tier 1B words such as "in order to" are marked as clarity edits, never as authorship evidence.
The paid lanes see your draft (the middle is cut past 24,000 characters, and the page says so) and the page's flags. Every quote they give is checked against your draft, and a rewrite is rescanned in full. A draft stays in this browser unless you run a lane. Derived from the agent skill @wshobson/avoid-ai-writing by Seth Hobson (wshobson/agents, MIT), whose catalogue and word tiers come from conorbronsdon/avoid-ai-writing (MIT). Sources are listed in llms.txt.