Writprint
Forensic Authorship Attribution

Identify the writer
behind any document.

When you have writing samples from a finite set of suspected authors, Writprint identifies which one wrote the anonymous text — and shows you the stylistic evidence that points the way.

Built for academic integrity offices, workplace investigators, and anyone who needs to know who put the words on the page.

Method
I.

Bring your suspects.

Paste between two and ten writing samples — emails, essays, Slack messages — from people you have authorization to investigate. Two hundred words apiece is a useful floor.

II.

Submit the document.

Paste the anonymous text in question. The sharper the contrast between candidate styles, the sharper the result.

III.

Read the report.

A ranked attribution with calibrated probabilities, the specific stylistic features that point each way, and the verifiable evidence behind every ranking.

Case Files
A-001

Academic integrity

Determine whether a submitted essay matches a student's known body of writing — class assignments, prior submissions, in-class samples — when ghostwriting or contract cheating is suspected.

B-002

Workplace investigations

Identify the author of an anonymous harassment email, leaked memo, or whistleblower note when the candidate set is bounded by access — direct reports, members of a Slack channel, attendees of a meeting.

C-003

Internal leak detection

Trace anonymous statements made to the press, on Glassdoor, or on internal forums back to a known set of staff with material knowledge.

Lineage
“In 2013, a forensic linguist identified J. K. Rowling as the author of The Cuckoo's Calling, published under a pseudonym, using the same statistical signatures that distinguish one writer's prose from another's.”
— Patrick Juola, Scientific American, 2013

Writprint applies the forensic stylometry techniques computational linguists have used in landmark attribution cases for forty years, augmented with state-of-the-art language models tuned for short-sample accuracy. Every ranking is backed by the specific, verifiable stylistic features the analyzer used to reach it.

Replication
Trial No. 001 · Federalist

We replicated the canonical
stylometric benchmark.

In 1964, Mosteller and Wallace attributed the historically disputed Federalist papers to Madison through statistical analysis of function words. We ran our own analyzer on the same problem. It agreed — in thirty seconds, with verifiable evidence.

Read the trial →
F-006
Federalist No. 6

Sanity — known Hamilton paper held out as target.

Expected: Hamilton · Predicted: Hamilton · p = 0.88
F-014
Federalist No. 14

Sanity — known Madison paper held out as target.

Expected: Madison · Predicted: Madison · p = 0.82
F-049
Federalist No. 49

Historically disputed. Mosteller–Wallace consensus: Madison.

Expected: Madison · Predicted: Madison · p = 0.82
F-051
Federalist No. 51

Historically disputed. Contains "If men were angels, no government would be necessary."

Expected: Madison · Predicted: Madison · p = 0.88
Trial No. 002 · Adversarial

The analyzer did not falsely attribute AI-generated text to a human.

On a generic AI essay placed against two human candidates, the analyzer correctly returned inconclusive. On an AI essay deliberately imitating one of the candidates, it flagged pastiche as the first caveat.

Read the trial →
Trial No. 003 · Multi-candidate

Discriminative power held across five candidates.

Held-out passages from Twain, Poe, and Emerson were placed against a five-author closed set (Twain, James, Emerson, Thoreau, Poe). Each was attributed to the correct author with high confidence and substantial gaps to the runner-up.

Read the trial →
Pilot
$299 · five cases · one-time

Pilot — five cases.

Five investigations, full reports, run them on whatever matters. About sixty dollars per case — set against the five to twenty‑five thousand dollars a forensic linguist charges for one. No subscription. If the work pays off, we'll talk about a seat license.

Start a pilot →
Privacy

Your case stays your case.

Submitted text is never used for model training and is held only in your private workspace, behind row-level access controls. We do not share with third parties beyond the model provider required to perform the analysis. You can delete an investigation at any time.

Begin

Bring us a case.

Sign in, paste your candidates and target, get the report. Direct line to the founder at hello@writprint.com.