Tools Article Shortener
Article Shortener

Reduce your article's word count to the journal's limit

Send the DOCX or PDF, choose how deep the cuts go, and get the file back with every cut marked as a tracked change to accept or reject in Word.

Version
2026.08
Last updated
Available since
Create free account How it works inside

Free credits · No credit card · Instant access What changed recently

What this solves

Your manuscript is over the journal's word limit and every paragraph looks necessary. The Article Shortener reads the body prose of your DOCX or PDF, condenses what is repetitive or wordy, and returns a Word file where every cut is a tracked change. Quotes, references, citations, numbers, tables, captions and headings are left untouched. You accept or reject each suggestion.

How it works

  1. Send the file

    Upload a DOCX or PDF of up to 10 MB and 20,000 words. If the document already carries pending tracked changes, accept or reject them in Word first and send it again.

  2. Choose the intensity and the target

    Pick the AI model and the cut intensity (light, moderate or aggressive). Optionally set a target in words or in pages: with a target, the review runs up to three passes to get there. The screen shows the exact credit cost before you start.

  3. Review in Word

    Follow the progress on screen or wait for the e-mail. Download the DOCX and open it in Word: each cut appears as a tracked change from the author XploreData, to accept or reject one by one. The original file stays available for download too.

Method

What happens to your file, step by step, and where the result stops being a decision made by a model.

How it works inside

A DOCX (or a PDF converted first with PyMuPDF) is classified by rule, not by model: only body prose is editable, while headings, captions, block quotes, everything after the references heading, and any paragraph carrying links, fields, equations, notes, comments or images stay protected. Editable paragraphs go to the chosen model in batches of at most 12 paragraphs or 9000 characters. Every returned paragraph passes a deterministic guard before being written back as Word tracked changes built with lxml.

Where AI is used, and where it is not

The model only rewrites paragraph text, and only the paragraphs the rule-based classifier released. It never decides what is protected, never touches tables or references, and never writes the file. What survives is decided by code: a revision is rejected when it is empty, longer than the original, under 30 percent of the original length, or missing any citation pattern or number present in the original. Rejected paragraphs keep the author's wording.

What it accepts

  • One DOCX or PDF up to 10 MB (PDF is converted to DOCX before anything else). Files that already contain pending tracked changes are rejected.
  • Up to 20000 words, counted Word-style per body paragraph including tables and excluding text boxes. A document with no editable prose paragraph is rejected.
  • Intensity: light, moderate or aggressive. Optional target in words or in pages; the page target is converted into words in the same proportion as the original file, and it is refused if the file does not declare a page count.
  • Model chosen in the AI model field (OpenAI, Gemini or Claude), which sets part of the price: 15 base credits plus 10 to 35 for the model, plus 0, 5 or 10 for intensity, plus 5 (or 10 on models priced 20 and above) when there is a target.
  • Two concurrent jobs per account; a third upload is refused until one finishes.

What you get back

  • DOCX with Word tracked changes (insertions and deletions authored as XploreData), so every cut can be accepted or rejected one by one in Word.
  • HTML diff preview of the revised paragraphs, capped at the first 150.
  • Job metrics: words before and after, paragraphs revised, paragraphs rejected grouped by reason (the revision came back empty, grew, cut below 30 percent of the original length, lost a citation, lost a number, or the paragraph's layout could not receive tracked changes), paragraphs whose batch failed, number of review passes and duration.
  • The original file stays downloadable, and when the target was in pages the metadata carries an estimated page count derived proportionally from the word count.
  • Files are deleted 30 days after the job ends; the cleanup runs the next time you open the tool, and the job metrics remain.

What this tool does not do

  • It only cuts. It does not summarize, does not rewrite for style and does not add anything: a revision that grows, empties or drops below 30 percent of the paragraph's length is discarded, as is one that is not shorter in words than what it replaces.
  • It does not touch protected content: references, headings, captions, block quotes, tables, footnotes, endnotes, headers, footers, text boxes and any paragraph containing links, fields, equations, comments, bookmarks or images.
  • It does not guarantee the target. Extra passes stop after three, or as soon as a pass changes nothing; the metadata records plainly whether the target was reached.
  • It does not check that the meaning survived. The guard verifies length, citation patterns like (Author, 2020) and [12], and every number and percentage; it does not verify fidelity, which is why the result ships as reviewable tracked changes.
  • It does not accept the changes for you, and paragraphs that contain tabs, manual line breaks, drawings or embedded objects are skipped at the tracked-changes step and counted as layout rejections.
Plan
Pro
Cost
25 to 70, depending on size and model

What you get

  • It only cuts body prose

    Headings, the references section, direct quotes, captions, tables, footnotes and paragraphs that carry links, fields or formulas are classified as protected and never go to the AI.

  • Every revision is checked before it enters the file

    The rewritten paragraph is compared with the original. If a citation such as (Silva, 2020) or [12] disappeared, if a number changed, or if the cut went below 30 percent of the original length, the revision is discarded and the paragraph stays as it was.

  • Native Word tracked changes

    The delivery is a single DOCX with real tracked changes signed XploreData. Nothing is applied silently: you decide cut by cut in Word, and you can preview the changes in the browser before downloading.

  • Target in words or in pages

    Type the journal limit. With a target set, the review runs up to three passes and escalates to aggressive; the result shows how far it got, how much is left, and offers a new run at aggressive intensity.

Changelog

Every line below is a change that actually shipped, dated by the day it went out.

  1. Latest
    • Credits are now charged right before the revised document becomes available for download. If the charge does not go through, the job ends with an explanation on screen, nothing is charged and nothing becomes downloadable.
    • Retrying a document after a failed finish no longer risks a second charge for the same job.
Show 5 earlier updates
    • The pages of the tool now follow the light and dark themes of the site.
    • Header, title and title animation match the rest of the tools.
    • The primary button and the model and intensity selectors follow the standard style, including a visible arrow on the dropdowns.
    • The reduction target can now be set in pages, not only in words. The page target is converted into an equivalent word target using the pagination of your own document.
    • The result compares before and after in words and in pages, with the page figure clearly marked as an estimate that may vary when opened in Word.
    • Credit costs recalibrated: the Shortener now has its own price per AI model instead of reusing the SmartChat table.
    • The tool only shows up in the catalog when it is enabled.
    • You can cancel a job while it runs, with no charge, and the waiting screen shows the real progress (pass X of Y, batch N of M). Each account processes up to two documents at the same time.
    • You can delete a submission with its files, see every submission on one page, run the same document again at aggressive intensity and download the original back. Files are deleted automatically 30 days after completion and the metrics stay.
    • The result page states plainly that the cuts are AI suggestions and that a human review of the final text is essential.
    • The history shows the file name, the upload area accepts click or drag, and the error messages are translated properly.
    • Processing moved to the background: you get a progress page with stage, elapsed time and estimate, you can close it and come back later, and an e-mail arrives when the document is ready. Jobs stuck for more than 45 minutes are stopped without charging.
    • PDF is accepted as input besides DOCX.
    • You can preview the changes in the browser, paragraph by paragraph, before downloading the file.
    • The word target became a real objective: the review runs extra passes to reach it and the result shows a meter with how far it got.
    • First version of the Article Shortener: send a DOCX, choose the AI model and the cut intensity, and get one file back with each cut as a native Word tracked change.
    • Paragraph classifier: only body prose is editable. Headings, the references section, direct quotes, captions, tables, footnotes and paragraphs with links or formulas are protected.
    • Deterministic check on every revision: anything that lost a citation, changed a number, grew instead of shrinking, or cut a paragraph below 30 percent of its length is discarded.
    • Shipped behind an internal launch switch that is off in production, so the tool is not open to users yet.

Questions and answers

Does it change my citations, numbers or references?

No. Inline citations like (Silva, 2020) or [12], the references section, direct quotes, tables, captions and headings are classified as protected and are never sent for revision. On top of that, any revision that lost a citation or a number is discarded automatically and the paragraph stays exactly as you wrote it.

Will it always reach the journal's word limit?

Not always, and the tool says so. Only body prose can be cut and no paragraph is reduced below 30 percent of its length, so a text heavy on references, tables and quotes has a floor it cannot go under. When you set a target we calculate that floor before charging anything and refuse an impossible target, showing the reachable minimum. If the review ends above the target, the result shows the gap instead of pretending it got there.

How many credits does it cost?

Between 25 and 70 credits per document. The final number depends on the AI model you pick, the intensity, and whether you set a target; the exact figure appears on screen before you start. The charge only happens when the revised document is ready, so cancelled jobs, failed jobs and jobs stopped by the time limit are not charged. If the AI finds nothing worth cutting, the job completes with zero changes and the analysis is charged. The tool requires the Pro plan or higher.

Is it available already, and what are the limits?

Not yet. The Article Shortener is built and tested but still behind a feature flag that is off in production, so it does not appear in the catalog until launch. When it opens: the file goes up to 10 MB and 20,000 words, each account can process two documents at the same time, and the files are deleted 30 days after completion (the metrics of each job remain in your history). The cuts are AI-generated suggestions, so a human review of the final text is essential.

How to cite this tool

Used it in your research? Here is the reference, already filled in with the version you are looking at and today's access date.

ABNT (NBR 6023)

LESSA, P. W. B. Article Shortener. Versão 2026.08. [S. l.]: Xplore Dados, 2026. Disponível em: https://xploredados.com/en/tool/article-shortener. Acesso em: 21 ago. 2026.

APA 7

Lessa, P. W. B. (2026). Article Shortener (Version 2026.08) [Computer software]. Xplore Dados. https://xploredados.com/en/tool/article-shortener

BibTeX

@software{xploredados_article_shortener_2026,
  author  = {Lessa, Patrick Wendell Barbosa},
  title   = {Article Shortener},
  organization = {Xplore Dados},
  version = {2026.08},
  year    = {2026},
  url     = {https://xploredados.com/en/tool/article-shortener},
  urldate = {2026-08-21}
}

Start with the free credits

Create an account and test the tools before deciding on a plan.

Create free account