About GPTCleanupApp tools

The premise

What GPTCleanupApp is

GPTCleanupApp is a set of small, single-job text apps. You open the one that matches the task, paste your text, run it, and take the result away. There's no project to create, no workspace to configure and no account anywhere on the site.

The apps split into four kinds of work:

Cleaning

Stripping invisible characters, extra spaces, markdown, HTML and accents out of pasted text.

Converting

Case changes, JSON formatting, URL encoding, unit conversion, splitting and counting.

Rewriting

Humanizers, paraphrasers, polishers and generators that call a language model to produce new wording.

Checking

Analyzers that read a document and report what they can observe in it, as signals rather than verdicts.

The set

The 14 hidden characters the cleaners handle

Every cleaning app on this site works off one shared list, and that list has fourteen entries. Competing pages name four or five and leave the rest implied, which is how a page ends up promising coverage it doesn't have.

Eleven are deleted outright: the zero-width space, the zero-width non-joiner, the zero-width joiner, the word joiner, the byte-order mark, the soft hyphen, and five directional controls. Three look exactly like an ordinary space, the no-break space, the narrow no-break space and the hair space, and those get swapped for a real space instead of deleted, because the word still needs a gap there.

Anything outside those fourteen is left alone, and the relevant pages name what they skip.

Workflow

How to use an app here in three steps

Nothing to install, nothing to set up.

Step 1

Open the app whose name matches your job. Check the line that says whether it runs locally or on a server.

Step 2

Paste your text and run it. Local apps return the result instantly, without a network request.

Step 3

Copy the output. Nothing is kept once you leave the page.

Two engines

Which apps run locally and which use a server

This decides what happens to your text, so it's worth checking before you paste anything sensitive.

Local apps

Cleaners, character removers, formatters, converters and the watermark scanners. The result is computed in your browser. Nothing is uploaded, and there's no run limit and no length cap.

Server apps

Rewriters, humanizers, generators and the analyzers. These need a language model, so your text is sent for the run. They accept between 1 and 12,000 characters and allow 5 free runs per rolling 24 hours.

The reliable tell

A limit exists to meter a server. An app with no daily allowance is one that runs on your own device.

How to check it yourself

Open your browser's network panel and run a local app. No request goes out while it works.

Scope

Which AI models these text apps work with

Any of them, because the cleaning apps match codepoints rather than models. A zero-width space is the same character whether it came from ChatGPT, Claude, Gemini, a PDF export, a word processor or a CMS editor. There's no supported-model list here because there's nothing for a model list to change.

That cuts the other way too. A hidden character doesn't prove a model wrote something. Publishing tools have been inserting these characters for decades, so a finding tells you how the text was handled, not who produced it.

The limits

What these apps will never claim

Every item here is a deliberate omission, not an oversight.

Never undetectable

No app here claims to make text undetectable, and none promises a result from Turnitin, GPTZero, Copyleaks or Originality.ai. Those systems belong to other companies and change without notice.

Never proof of authorship

The checkers report signals they can observe. A signal is evidence about the text, not proof of who wrote it.

Never an invented number

No accuracy percentages, no user counts, no testimonials, and no claim about how many hidden characters a typical AI reply contains. Numbers on these pages come out of the code.

Never an implied partnership

Product names appear because those are the words people search with. Naming one isn't a claim of affiliation in either direction.

The problem

What hidden characters actually break

This is the defect the site was built around, and it's specific. A character that renders as nothing still occupies a slot in the string, so anything comparing text character by character sees two values that a reader sees as identical.

A find-and-replace misses. A URL slug splits into two words. A spreadsheet lookup fails against a value that looks right on screen. A line of code throws a syntax error with no visible cause. None of it announces itself, which is exactly why it's worth a tool rather than a careful read.

Who uses it

Who these text apps are built for

Developers cleaning pasted text before it reaches a codebase, a config file or a commit. Writers and editors tidying a draft or getting a paragraph out of a register it got stuck in. Publishing and SEO teams checking titles and slugs where one invisible character splits a keyword.

Students and academics reviewing their own work, within whatever rules their institution sets, since those rules govern and nothing here changes them. And anyone who pasted text out of a chat window and found it behaving strangely for no visible reason.