editGPT AI Rewriter Review: Word Limits and DOCX Test
editGPT publishes a clear Word editing contract: import a document, edit with AI, and export a DOCX with track changes. The useful test is whether your existing revisions, citations, numbers, table structure, and editable file state survive that round trip.
HumanPen Team
· 10 min read
Is editGPT AI Rewriter worth trying?
editGPT is worth a controlled trial if you want AI edits delivered through a Word-style track-changes workflow. Its public page states that Word documents can be imported and exported as DOCX with track changes. It does not turn that statement into proof that every citation, table, field, or pre-existing revision will survive your file. Test those items before using it on a real deliverable.
The scope stops at what the public pages say. I checked the signed-out homepage, Free AI Rewriter route, Privacy Policy, and Terms and Conditions on August 28, 2026. Every text box and file input remained empty. I never opened a detector, checkout, or any action that would accept terms or charge an account.
The distinction matters because an import button, an editable return, and preservation of document objects are three different claims. editGPT makes the first two Word claims in public. The third still belongs in your acceptance test.
Short answer: treat track changes as the review interface, then verify the document objects underneath it.
What is editGPT AI Rewriter actually positioned to do?
The editGPT homepage calls the product an AI editor for proofreading, editing, and improving writing. On the checked date, the Free AI Rewriter route rendered the same core editing, Word, pricing, and privacy copy under a different page title.
| Rendered route | Normalized main-text characters | `track changes` | `docx` | literal substring `rewrite` |
|---|---|---|---|---|
| Homepage | 3,951 | 6 | 4 | 1 |
| Free AI Rewriter | 3,781 | 6 | 4 | 1 |
For that inventory, I read `document.querySelector('main').innerText`, collapsed each whitespace run to one space, converted a copy to lowercase, and counted case-insensitive literal substrings. On both routes, the same counter found zero instances of `humanizer`, `bypass`, `undetectable`, and `turnitin`. The nonzero Word terms are the sensitivity controls. The page title sits outside the counted `main` element.
This is a page-positioning result, not a backend capability claim. The public contract centers on editing and revision review.
How do editGPT prices and limits fit together?
The pricing area has Monthly and Annual views. I switched both controls in the signed-out page and waited for the selected state and prices to change. The monthly word allowance and per-document ceiling are separate units.
| Plan | Monthly billing | Annual view | Words per month | Words per document | Other limit shown |
|---|---|---|---|---|---|
| Pro | $15/month | $8/month; $96 billed annually | 500,000 | 200,000 | Not stated |
| Elite | $25/month | $15/month; $180 billed annually | 3,000,000 | 200,000 | Not stated |
| Business | $100/month | $60/month; $720 billed annually | Unlimited | No separate figure displayed | Unlimited requests; 4 team seats |
The Business card says `Unlimited words / month` and `Unlimited requests`. It does not label a single document or request as unlimited. For Pro and Elite, the 200,000-word document ceiling remains much smaller than the monthly pool.
Record billing cadence, monthly pool, and per-document ceiling in separate columns. They answer different questions.
What does the Word contract establish?
The homepage says users can import and export Microsoft Word documents while maintaining track changes. A second sentence narrows the route: import a Word document, edit it with AI, and export it back to `.docx` with track changes.
The same feature area says formatting support includes bold and italics, followed by the open-ended phrase `and more`. That wording is useful but not a specification for tables, citations, footnotes, fields, comments, or existing revision history.
| Document element | Public statement checked | What the sample still needs to prove |
|---|---|---|
| Return file | Export back to `.docx` | File opens in Word and remains editable |
| AI edits | Returned with track changes | Insertions and deletions are reviewable and attached to the intended text |
| Existing revisions | Covered only by the broader track-changes wording | Pending revisions from before import remain distinct and usable |
| Basic rich text | Bold and italics are named | Styles remain attached to the correct spans |
| Citation and number | No result claimed in this review | Exact strings and claim-to-citation placement survive |
| Table and Word fields | No result claimed in this review | Cell geometry, formulas, captions, links, and fields remain functional |
Importing and exporting a DOCX is a stronger public commitment than a paste-only editor offers. It is still a workflow statement. Only the returned artifact can establish object-level preservation for your file.
`DOCX out` is the start of the test. `The document still works` is the finish.
How can you run a reproducible DOCX preservation test?
Use a disposable, non-sensitive DOCX that you are allowed to process. Put one real citation, one measured number with a unit, one small table, and two existing Word revision marks into the file. Keep an untouched original.
Before uploading, make a ledger that can be checked without judging writing style:
| Checkpoint | Record before the run | Pass condition after the run |
|---|---|---|
| Real citation | Full citation string and the last six words of the supported claim | String is unchanged and still follows the same claim |
| Number | Value, sign, decimal, unit, date, and role in the sentence | Every token and its meaning are unchanged |
| Table | Header text, row and column counts, and two cells identified by coordinates | Geometry and both sentinel cells match after reopening |
| Existing revisions | One pending insertion and one pending deletion, with anchor text and author label | Both remain distinct; new AI edits do not replace or merge them |
| Basic formatting | One bold span and one italic span | Each style stays on the same words |
| Returned file | Input suffix, file size, Word version, and editable state | Return is a DOCX, opens without repair, and remains editable |
Use one document, one prompt, and one plan state. Save the input, returned file, time, settings, and visible delivery route together. If editGPT offers choices during the run, record them before changing anything.
- Open the untouched original in Word and capture the ledger.
- Import the copy, make one bounded editing request, and export the result.
- Reopen the returned DOCX in Word rather than judging an in-browser preview.
- Compare the citation and number character by character.
- Inspect the table by coordinates and recalculate any subtotal from its cells.
- Open Word's Reviewing pane and separate the two old revisions from the new edits.
- Save, close, and reopen the returned file once more to confirm editability.
Record each failure by object and step. One passing file supports only that document shape, request, account state, and product version. It does not establish a detector result, so keep detector scores out of this preservation test.
A preservation test should leave you with two files and a ledger, not one overall impression.
What does editGPT say about submitted documents?
The Privacy Policy gives unusually direct document statements. These are vendor policy commitments, not an independent security audit.
| Question | Policy statement checked on August 28, 2026 |
|---|---|
| Who can access saved documents and submitted text? | The policy says only the user can access them |
| Why are submitted text and files used? | Solely to deliver the service directly to the user |
| Are files or account data shared with third parties? | The policy says they are not shared |
| Are documents used for training? | The policy says documents, files, and submitted text are not used for training |
| What happens when a file is deleted? | The policy says it is permanently removed from the servers |
| What happens when an account is deleted? | Associated documents and account information are permanently removed |
The policy also says the service collects account information, usage information, log information, IP address, browser type, and operating-system data. If a document is confidential, unpublished, regulated, or institution-restricted, the written policy and your own permission both need to fit the job.
A clear deletion sentence is useful. It does not replace checking whether you are allowed to upload the document at all.
Who is the editGPT Word workflow a fit for?
The clearest use case is reviewing AI edits inside Word instead of copying a finished paragraph out of a browser. Risk rises as more Word objects need to keep working.
| Need | What to verify before relying on it |
|---|---|
| Review every AI edit | Returned DOCX exposes usable insertions and deletions in Word |
| Keep existing revision history | Old revisions remain distinct from new editGPT revisions |
| Citation-heavy document | Citation string, locator, link, and supported claim stay connected |
| Tables or calculated values | Shape, sentinel cells, formulas, and totals survive reopening |
| Long document | Plan's monthly pool and per-document ceiling both cover the job |
| Team workflow | Four-seat Business card fits the actual number of reviewers |
| Sensitive material | Policy, institution, client, and legal permissions all allow upload |
A clean prose edit cannot compensate for a broken field, moved citation, merged revision, or damaged table. Make those objects the acceptance gate when they matter to delivery.
The right sample is a small version of the document you must hand in, not a paragraph chosen because it is easy.
When does the publisher's workflow fit better?
HumanPen publishes this page and may benefit financially if readers choose its service. Its signed-out page says Word and PowerPoint files remain editable when returned. It names three incoming formats, TXT, PPTX, and DOCX. Users can confirm selected passages, or let an existing report from Turnitin or iThenticate mark the passages to process.
| Decision point | editGPT public page | Publisher's public page |
|---|---|---|
| Main review surface | Word DOCX with track changes | Editable return that matches the input format |
| Scope control | Review AI edits through the track-changes interface | Confirm selected or report-matched passages before rewriting |
| File types named | Word import and DOCX export | DOCX, PPTX, and TXT |
| Evidence in this article | First-party contract; no completed run | First-party contract; no completed run |
This comparison is about file workflow and review scope. It gives no writing-quality ranking and draws no conclusion about detector-score movement for either service. Before trusting a high-stakes file, apply the same object ledger to the route you plan to use.
Pick the workflow whose returned artifact matches the document you must deliver.
What is the verdict on editGPT AI Rewriter?
editGPT publishes a useful and specific Word promise. Import, AI editing, DOCX export, and track changes form a workflow that a reader can understand before signing up.
The limits are also legible once their units are separated. Pro and Elite combine a large monthly pool with a 200,000-word document ceiling. Business uses `unlimited` for the monthly word pool and request count, without turning that label into a per-request word statement.
The buying gate is the returned DOCX. A sample that preserves the real citation, numeric role, table coordinates, existing revisions, and editable file state gives evidence for that exact workflow. Anything broader would need more tests.
The public contract is clear enough to test. The file decides whether it passes.
Frequently asked questions
Does editGPT export a Word document with track changes?
The homepage says users can import a Word document, edit with AI, and export back to `.docx` with track changes. This review did not run the export, so the returned file still needs to be opened and checked in Word.
What is the editGPT per-document limit?
The Pro and Elite cards each show 200,000 words per document. The Business card shows unlimited words per month and unlimited requests, but no separate per-document figure in the checked view.
How much does editGPT cost?
With month-to-month billing selected, Pro displayed $15, Elite $25, and Business $100. Choosing Annual changed the displayed monthly equivalents to $8, $15, and $60, with yearly charges of $96, $180, and $720.
Does editGPT preserve citations and tables?
This review found a Word return statement and named support for bold and italics. It did not run a file, so it makes no preservation result claim for citations or tables. Use a non-sensitive DOCX and compare both objects against a before-run ledger.
Does editGPT use documents to train AI models?
Its Privacy Policy says documents, files, and submitted text are not used for training or purposes other than providing the direct service. That is a vendor policy statement, not an independent audit.
Does this review show whether editGPT changes an AI detector score?
No. The browser session sent neither prose nor a document, and it never opened a detector. The review covers the public Word workflow, limits, privacy wording, and a preservation method.
Official pages checked
Source review date: August 28, 2026. Pages, prices, limits, and policies can change. The session stayed signed out, left all inputs empty, and did not enter any detector, checkout, subscription, cancellation, or payment flow.
KEEP READING