A document count begins with extraction, not arithmetic
A DOCX can keep its main body, notes, headers, footers, and floating text in different regions. A PDF may contain selectable text, page images, or both. An HTML file mixes readable copy with markup and scripts. Before a counter can measure words, it must decide which of those parts become plain text.
Words Calculator makes that boundary visible. Upload one supported file, choose optional regions for a Word document, and inspect a report with words, characters, sentences, paragraphs, lines, vocabulary, repeated phrases, timing, and plain-text evidence. The Extraction tab records what entered the result and what stayed out.
What the file parser actually reads
“The whole document” is not one technical boundary. This table describes the scope used by the live tool, not a promise of visual layout reproduction.
| Format | Counted boundary | What to verify |
|---|---|---|
| Selectable page text | Review reading order; run OCR first for image-only scans. | |
| DOC / DOCX | Body plus the optional regions you select | Choose whether notes, page furniture, and text boxes belong in scope. |
| ODT | Document content XML | Layout and non-text media are outside the count. |
| RTF | Visible text after formatting controls are removed | Use the preview to confirm unusual fields or embedded objects. |
| HTML | Readable body text | Scripts, styles, embeds, and markup are removed. |
| Markdown / TXT | Complete normalized text source | Markdown link destinations and source wording remain part of the file. |
PDF is a text-layer question
A successful PDF report reads selectable text and can return the file's page total. A scan made only from page images has no searchable words until optical character recognition creates a text layer. OCR output should be checked because recognition can misread letters, punctuation, columns, and tables.
Word scope is a user decision
DOC and DOCX reports always include the body. Turn on notes, headers and footers, or text boxes only when those regions belong in your assignment, contract, editorial brief, or translation quote. If detectable text remains excluded, the report records that decision.
How to count a document without losing the audit trail
The number is most useful when someone else can reconstruct its scope later.
- 1
Choose the right file version
Upload the exact draft, manuscript, transcript, or source file attached to the decision. Remove password protection from a safe copy; do not substitute a visually similar version.
- 2
Define the Word regions
For DOC or DOCX, decide whether footnotes, endnotes, headers, footers, and text boxes belong in scope before extraction. The body remains the stable default.
- 3
Read the preview before the charts
Search the extracted text for missing pages, repeated furniture, broken column order, unexpected link destinations, or notes that should have been included.
- 4
Save the measurement with its scope
Download the report beside the source version. It preserves the file facts, counted regions, main totals, useful language signals, warnings, and counting-method version.
Why two document word counters can disagree
Different totals do not automatically mean one counter failed. First compare scope, then extraction, then word rules. A program may include notes or text boxes by default while another counts only the body. One PDF parser may follow tags while another infers layout. Counters may also handle apostrophes, joined compounds, numerals, emoji, or scripts without spaces differently.
When a limit affects a grade, invoice, filing, or acceptance test, use the rule named by the receiving organization. Keep this report as an independent measurement, and record the file version and selected regions rather than presenting a bare total without its method.
Four jobs where extraction scope matters
Measure the submitted file, then follow the institution's rule for references, appendices, notes, captions, and headers. A downloadable report records which Word regions were included.
Inspect body length, sentence and paragraph shape, recurring language, and reading time while keeping notes separate from the prose an editor is pricing or revising.
Decide whether page furniture and text boxes are translatable, verify the extraction, then save the source count with the quote. PDF columns deserve extra preview review.
Use speaking time as a planning estimate after confirming that stage directions, speaker labels, notes, and headers entered the intended scope.
Page totals are format facts, not guesses. A parsed PDF supplies a fixed page count. Flowing documents can repaginate when fonts, margins, paper size, printer metrics, or software versions change, so the report does not invent a DOCX or ODT page total.
A bounded server task, not a document library
Document parsing cannot happen until you press Analyze. The service accepts one file up to 10 MB, stops if extraction exceeds two million characters, keeps the preview to 8,000 characters for a responsive interface, and marks the response no-store. Source bytes are discarded after the structured result is prepared; document contents are not intentionally written to application logs.
Verify the format limitations
These references were rechecked on July 25, 2026.
- Adobe Acrobat — recognize text in scanned PDFsWhy image-only scans require OCR and why recognized text should be reviewed.Open source
- Adobe Acrobat — PDF reading orderWhy columns, tables, and page regions can extract in a different sequence.Open source
- Microsoft Support — footnotes and endnotesHow Word stores supporting notes outside the main body location.Open source
Questions that can change the total
Every answer describes the live tool's actual behavior.
Which document formats can this word counter read?
It accepts text-based PDF, DOCX, DOC, ODT, RTF, HTML, Markdown, and TXT files up to 10 MB. Browser-side parsers inspect the file's structure rather than trusting the extension alone. Password-protected documents, unsupported archives, and image-only PDFs stop with a specific explanation instead of returning a misleading zero.
Can the Document Word Counter read a scanned PDF?
Only when the scan already has a selectable OCR text layer. A scan made entirely from page images contains no searchable text for this tool to count. Run OCR in a trusted PDF application, review the recognized wording for errors, save a searchable copy, and open that copy for local analysis.
Why can this result differ from Microsoft Word's word count?
The counting scope and token rules may differ. This tool counts the Word document body by default and lets you opt into footnotes, endnotes, headers, footers, and text boxes. Microsoft Word settings, tracked or hidden content, fields, punctuation, and language boundaries can produce another total. Use the Extraction tab to document this report's scope.
Are footnotes, endnotes, headers, footers, and text boxes counted?
For DOC and DOCX files, the body is always counted and the three supplemental groups are optional. If the parser finds text in an available region that you left off, the report names that exclusion. Other formats use one documented extraction boundary because they do not expose those Word regions through the same parser.
Why does PDF text sometimes appear in an unusual order?
A PDF stores positioned page content, and its logical reading order may not match what the eye sees. Columns, tables, running headers, side notes, and footnotes are common trouble spots, especially in untagged files. The Text preview lets you inspect the extracted sequence before relying on sentence, paragraph, phrase, or readability results.
Does the tool calculate page count for every file?
No. It reports the page total embedded in a successfully parsed PDF. DOCX, DOC, ODT, RTF, HTML, Markdown, and TXT do not have a reliable fixed page count without reproducing the original fonts, margins, paper size, printer metrics, and layout engine, so the report says PDF only instead of inventing an estimate.
Is the whole document visible in the text preview?
The preview displays up to 8,000 extracted characters to keep the interface responsive. The word, character, sentence, paragraph, timing, and vocabulary results use the complete extracted text, up to the service's two-million-character safety limit. The Extraction tab states whether the preview is complete or capped.
Can I download the document word count results?
Yes. Copy or download a plain-text report containing the file facts, selected scope, primary counts, timing, structure, vocabulary summary, extraction warnings, and counting-method version. You can also download the extracted preview separately, which is useful as an audit note but does not recreate the original document's formatting.
Does my document leave the browser or get used for training?
No. The current document parser, extraction, counting, preview, and report run in your browser after you press Analyze. Source bytes and extracted text are not posted to Words Calculator, stored in an account, or used as model-training material. Browser memory is released when the page state is cleared or closed.