TextCloaker Review: Preservation, Price, Cancellation

TextCloaker may be worth a controlled free test if the job is to keep visible wording unchanged. Its homepage says not one word changes and that every word, space, and character stays exactly as written. I did not run the tool, so this review cannot confirm that promise or predict a detector result.

Research stayed on the signed-out homepage, Terms of Service, Privacy Policy, Refund Policy, developer page, and API root on August 28, 2026. Every interactive path stayed unused: the text field, file input, cloak and detector controls, registration, checkout, terms acceptance, and payment. The `/app` route led back to the homepage.

The free card offers three Light cloaks per day without a card. That is enough to run a non-sensitive preservation check before deciding whether a paid cadence makes sense. Resolve the cancellation contract first: the homepage says `Cancel anytime`, while the written policy requires email notice and lists minimum commitments, a final payment, and early termination fees.

Test the preservation promise and settle the exit cost before treating either one as known.

The homepage uses two unusually broad statements: `without changing a single word you wrote` and `Every word, space, and character stays exactly as you wrote it.` Those claims cover more than meaning or tone. Read literally, they reach wording, whitespace, and characters.

The page also offers Light, Regular, and Deep modes. A mode label does not explain how the service can affect detection while keeping the promised text layers unchanged, and the public result claims remain vendor marketing until a shared test is run.

For a normal paraphraser, a quick read may reveal the edits. Here, visible sameness is the advertised behavior, so a visual check alone is too weak. A fair test needs to compare what the page renders, what copy and paste returns, and which Unicode characters are actually present.

"Looks the same" and "contains the same characters" are separate checks.

Four rendered public pages were audited as a disclosure set: the homepage, Terms, Privacy Policy, and Refund Policy. With the homepage's `WEEKLY` pricing tab selected, each page was read from `document.body.innerText` without whitespace normalization; the four raw lengths were measured separately and summed to 40,588 characters on August 28, 2026.

The sensitivity controls were live: `TextCloaker` appeared 42 times across the same four pages, and `Service` appeared 33 times on the Privacy Policy alone. The zeroes therefore support a narrow conclusion: those rendered pages do not publish a preservation contract for the listed structures.

This is not a claim about the backend. It does not prove the service must damage a citation, table, formula, footnote, tracked change, or cross-reference. It only means a reader should not treat preservation of those structures as documented without a successful file test or a written answer from support. The count covers rendered page text; it does not transcribe every frame of the homepage video.

Missing public specification means "verify," not "assume failure."

Use invented, non-sensitive content that makes small changes easy to spot. Keep the original untouched and place this sentence above the table: `Rivera (2025, p. 12) wrote, "Batch A stayed below 4.75 ms." The audit covered 128 rows, found 17 exceptions, and recorded 13.28%.`

The sample is a test fixture, not a claim about TextCloaker or detector performance.

Run one mode at a time and save each result separately. Then compare the original and result in this order:

1. Check the quotation marks, `Rivera`, `2025`, `p. 12`, every digit, decimal point, and percent sign.
2. Check the table headers, row order, column order, alignment, and all eight data cells.
3. Compare visible wording and whitespace side by side at high zoom.
4. Paste each version into separate UTF-8 plain-text files. Test exact equality, then inspect any code-point difference with a character-aware editor; a visual match is not enough.
5. Copy the result into a disposable local document, save it, close it, reopen it, and repeat the visible and plain-text checks.
6. If testing batch upload, reopen the returned `.docx` and inspect paragraph order, table structure, citations, styles, and any tracked changes or cross-references in the sample.

This protocol predicts no result. It is designed to tell you what happened to one authorized sample.

The homepage advertises batch upload for `.txt` and `.docx`. That expands the acceptance test from text equality to document round-trip behavior. A `.docx` file is a package of paragraphs, tables, styles, relationships, and other structures; unchanged visible sentences do not establish that the package came back intact.

Use a disposable file with two paragraphs, the sample table, one citation, and deliberately varied styles. Record the original file size and application used, but do not treat matching file size as proof of matching content.

After downloading, open the file in the program used for the final work. Copy the table back into a fresh local document and inspect it again. This catches a result that looks correct in one preview but breaks during the next editing step.

Judge the returned file after reopening and editing it, not from a download confirmation alone.

The homepage displayed one free allowance and three Premium billing cadences on August 28, 2026. The annual card converts its billed total to a monthly display, but the charge shown is annual.

Do not compare `$4.17/month` with the monthly plan as though both cards charge monthly. For an annual purchase, the immediate amount shown is $49.99. Prices and card wording can change, so verify the live payable total before purchase.

The free allowance is the lower-risk place to run the text fixture. It does not answer how Premium performs, but it can show whether the Light result meets the preservation checks that matter to your document.

Separate the displayed monthly equivalent from the amount and cadence actually charged.

The Premium card says `Cancel anytime`. The Refund Policy, effective July 31, 2026, describes a more specific exit process: cancellation by email only, at least 30 days' notice, and one final payment after cancellation. It also says sales, subscriptions, credits, and renewals are non-refundable except where law requires otherwise.

Those conditions materially affect a short trial. The public wording does not let this review calculate one universal exit total because the amount depends on the chosen cadence, timing, minimum commitment, final payment, and any applicable fee or local law.

Before paying, ask support to state in writing the earliest effective cancellation date, whether another renewal will be taken, whether the minimum commitment is satisfied, and the total amount due for the selected plan. Save the response and cancellation email. This article only reports published terms and offers no legal opinion.

"Cancel anytime" tells you when a request may be sent. The written policy governs what the request may cost.

The Terms say submitted text is transmitted to TextCloaker's servers. They also say text is not permanently stored beyond what is necessary to provide the service, with cloak history given as an example. The homepage says the last 15 cloaked texts are saved automatically.

That wording supports a server-processed workflow, not an offline-only one. Do not submit confidential, regulated, unpublished, or institution-restricted material until the storage, access, deletion, and permitted-use terms fit the document.

The Terms also require users to follow law and institutional policies. A successful technical result would not override a course, employer, client, or publication rule.

Decide whether the text is allowed to leave your device before testing what the service does to it.

TextCloaker's strongest public differentiator is the promise of unchanged text, not a conventional rewrite. That may suit a user who wants to test a non-sensitive passage without accepting visible wording changes. It is a weaker documented fit when document structures or a predictable cancellation path are essential.

A pass on one sample does not establish every mode, file type, language, or document. Keep the tested mode and file type attached to the result rather than turning one run into a product-wide conclusion.

The right question is whether the exact workflow you need passes its own acceptance test.

The commercially interested publisher of this comparison is HumanPen, which operates the other workflow. Its anonymous product page says it sends DOCX, PPTX, or TXT back in the incoming format. Word and PowerPoint outputs remain editable. A user can mark passages for processing while material outside the confirmed scope stays untouched.

This comparison is structural. It does not rank output quality or detector performance, and neither column proves what will happen to a particular file. The publisher's workflow may be more relevant when only selected passages should change and the rest of an editable document needs to remain in place.

Choose between an unchanged-text promise and a scoped document rewrite based on what the final file must allow you to inspect and edit.

TextCloaker makes a clear and testable public promise: words, spaces, and characters stay as written. The homepage also gives a no-card way to test three Light cloaks per day. Those are useful facts, but they are not observed performance.

For evidence-heavy work, the missing piece is a public structure contract. The four-page rendered-text audit found no preservation wording for citations, footnotes, tables, formulas, tracked changes, or cross-references. A disposable acceptance file can turn that uncertainty into a result for one workflow.

The purchase decision has a second gate. The homepage's `Cancel anytime` line needs to be read together with the Refund Policy's email-only cancellation, 30-day notice, final payment, minimum commitments, and early termination fees. Resolve those terms in writing before paying.

Try the free fixture first. Pay only after both the file test and the exit terms are acceptable.

Does TextCloaker change the words in my text?

Its homepage says it does not change a single word and that every word, space, and character stays exactly as written. No result was generated for this review, so treat that as a vendor promise and test it with non-sensitive text.

Can TextCloaker preserve citations, tables, and footnotes?

The four checked rendered pages did not publish a preservation contract for citations, tables, footnotes, formulas, tracked changes, or cross-references. That does not prove the backend cannot preserve them. Use a disposable `.docx` fixture and inspect the returned structure.

Is TextCloaker free?

The free card showed three Light cloaks per day, a daily reset, and no card requirement on August 28, 2026.

How much is TextCloaker Premium?

The homepage showed $2.99 per week, $6.99 per month, or $49.99 billed annually and displayed as $4.17 per month. Check the current payable amount before purchase.

Can I cancel TextCloaker anytime?

The homepage uses that phrase. The Refund Policy adds email-only cancellation, at least 30 days' notice, one final payment, minimum commitments, and early termination fees. Ask support for the effective date and total exit cost for your selected plan before paying.

Does TextCloaker store my text?

The Terms say text is sent to servers and is not permanently stored beyond what is necessary to provide the service, with cloak history as an example. The homepage says the last 15 cloaked texts are saved automatically.

• TextCloaker homepage
• Terms of Service
• Privacy Policy
• Refund Policy
• Developer page
• API root
• Editable document workflow

Source review date: August 28, 2026. Prices, allowances, policies, modes, and public disclosures can change. The developer page and API root were observed read-only; no request or access flow was tested.