Read this first
What the Claude Copyleaks checker is, and is not
This tool has no connection to Copyleaks. We do not run their detector, we have no partnership with them, and we have no way of seeing or estimating what their scan would report about your text. If you came here hoping for a preview of their result, there is not one, here or anywhere.
What you can do here is get an independent written reading of a draft before anyone else reads it. Paste the text, press Analyze, and findings come back as sentences. It runs on our server, 5 runs per rolling 24 hours, up to 12,000 characters each, with an optional instruction of up to 300 characters.
The loop
How to check a draft in three steps
Five runs a day is enough for a working session if you aim them rather than spraying them.
- Step 1
- Paste one passage, not the whole document. A section gives findings you can act on in one sitting.
- Step 2
- Put your actual worry in the instruction box. "Does this section repeat the one before it?" beats an unguided request every time.
- Step 3
- Copy the findings out beside the draft and edit. Nothing is stored here, so the result pane is the only copy until you move it.
Know which you need
Plagiarism checking and AI detection are different jobs
Copyleaks sells both, and the pages that list alternatives to it treat them as one thing. They are not, and this route does neither in the sense those products mean. Here is the split, so you can go to the right place for each.
- Similarity to existing work
- Needs a corpus to compare against. There is none on this route: no index, no matching, no source list, no similarity report.
- Machine-written signals
- Copyleaks documents per-section human-or-AI classification and sentence highlighting. Nothing here labels a span or returns a verdict.
- How the writing reads
- Repetition, flat openings, sections that do not deliver what their heading promised. This is the part this tool is for.
- Whether you followed the policy
- Not a tool question at all. It is a matter of your institution's or client's rules, and disclosure under them.
Why findings, not a number
Scores move with settings, drafts do not
A published review of the Copyleaks checker describes the output as "a likelihood score and sensitivity options", with adjustable sensitivity for different use cases. Read that carefully and it means the same passage can land differently depending on how a reviewer has configured the scan.
You cannot work against that. You can work against a sentence telling you that your three middle paragraphs open identically, because you can look and see whether it is true. That is the whole design argument for producing findings rather than a percentage, and it is why there is nothing here to optimise toward.
The same logic applies to the numbers around the edges. Copyleaks documents support for 30 or more languages in its detection API. We publish no language count at all, because nothing in this tool establishes one, and a figure we cannot stand behind is worse than saying nothing.
In practice
Using the Copyleaks checker before a scan
Do the useful things in order. Write the draft, run a section through here, fix what you agree with, then keep the wreckage: earlier drafts, notes, the version history in your editor, the tabs you actually read. That material is the only durable answer to a question about how something was written, and it has to exist before the question is asked.
Then read the policy that applies to you and disclose under it if you used AI. That is the advice the detector vendors themselves give, and it is better than the alternative, which is editing blind against a result you will never see.
Access
What you get without paying
The review we read sums up the Copyleaks side as "limited free access for testing", with "Full AI detection and plagiarism checking require a paid plan" and pricing aimed at professional and organisational use rather than occasional ones.
This side is 5 runs per rolling 24 hours, no account, no card, no stored history and no tier above it. The window is rolling rather than daily, so runs spent this afternoon are back tomorrow afternoon. That is the whole offer and there is nothing to be upsold into.
Where it stops
What this Copyleaks checker does not do
It does not produce a Copyleaks result, an estimate of one, or any percentage at all. No sensitivity control, no highlighted sentences, no per-section labels, no authorship claim about any model or person.
No plagiarism index, no file upload, no bulk queue, no API and no saved history. No accuracy figure either, because we have measured none, and quoting someone else's benchmark would describe a different tool than the one you are using.