PaperTrail™ Academic

You've read thousands of student essays.
Trust that instinct — then document it.

Six tools that give teachers structured, repeatable, documentable insight into how student writing was produced — from process monitoring in the classroom to a spoken defense of the finished essay. The evidence speaks clearly. The judgment stays yours.

No account required to start No student data stored Works inside Google Classroom documents Built by a teacher
✏️
Inspect  Free
How was this written? Sessions, pastes, struggle moments — and a printable Process View report.
Learn more →
🔍
StyleMatch™  $1.99/mo
Does this sound like them? Seven stylometric metrics compare submitted work to a known sample.
Learn more →
🔬
Verify  1 Token
Is the writing internally consistent? AI-assisted close reading. Three report types.
Learn more →
📚
Verify Citations  From 2 Tokens
Did the sources really say that? Every citation checked against the source itself — live web search, both sides quoted.
Learn more →
🎙️
Oral Defense  2 Tokens
Can they account for it? Students defend their own essay out loud — short spoken answers, no account needed.
Learn more →
🖊️
Write  $49/yr
Timed in-class writing sessions with a full process log.
Learn more →

A structured, repeatable, documentable data point for the hardest conversation in teaching.

The PaperTrail Suite

Six tools. One question: Did they write this?

Click any card to see a full preview and feature breakdown.

"Something about this essay doesn't add up."

Reconstruct the full revision history of any Google Doc — every session, paste, and pause — into a clear, documentable timeline.

  • Full revision timeline & paste log with novelty scores
  • Session history — when and how long they worked
  • ⚡ Struggle moment detection — regions of heavy revision
  • Revision playback — watch it being written
  • 📊 Process View — printable observational report (PDF)
The Social Contract — Final Draft
📄
Inspect
Summary
Pastes
Flags
47
Revisions
23m
Active
3
Pastes
📅 Timeline
📋 Paste #2 — 847 chars 82% Novel
"The social contract, as Rousseau described, represents a fundamental agreement…"
⚡ High-Density Revision · Para 2
412 chars rewritten · 8 revisions · 4m 12s · Early-draft position
Revision #31 diff
the argument remains unclear to scholars
the philosophical implications of this debate
have been studied in modern contexts.
Session 2 · Nov 14, 9:38 PM
18m · 1,240 chars · 2 pastes

Everything you need to understand a writing process

  • Full revision timeline — every edit laid out chronologically
  • Paste detection & paste log — flagged with timestamp, character count, and novelty score
  • Session history — date and time the student opened and worked on the document
  • ⚡ Struggle moment detection — regions where the writer cycled through deletions and rewrites in the same place
  • Revision playback — watch the document being written from the first character
  • 📊 Process View report — a printable, observational PDF summarising sessions, paste activity, struggle moments, and five research-cited writing-process metrics. Form fields for student/assignment/course are filled in the browser before printing — no transmission.
  • Activity flags — unusual patterns automatically surfaced
Free
No account required. Installs in seconds.
⬇ Add to Chrome — Free
Inspect reads Google Classroom document revision metadata. Requires Editor or Owner access to the document. No document content is stored or transmitted.

"This doesn't sound like them."

Compare a known writing sample against the submitted work using seven research-based stylometric metrics. The only PaperTrail tool that compares two pieces of writing against each other.

  • Seven stylometric metrics — function words, discourse markers, punctuation fingerprint, sentence rhythm, vocabulary, and more
  • Controlled-vs-submitted comparison — the structural feature Inspect cannot offer
  • Divergence ratings per metric: Similar / Notable / Divergent
  • Printable PDF report with research citations and per-metric interpretation
  • Runs entirely in your browser — no text ever transmitted
🔍 StyleMatch — Authorship Consistency Report
Controlled: 847 words · Submitted: 1,204 words
J. Martinez · Nov 14, 2025
FK Grade Level
92% Similar
Type-Token Ratio
87% Similar
Avg Sentence Length
81% Similar
Punctuation
82% Similar
Sent. Rhythm (SD)
68% Notable
Discourse Markers
43% Divergent
Function Words
3% Divergent
⚠ Analytical data point — not a finding of fact. Educator judgment required. Not suitable as sole basis for any disciplinary action.

Seven metrics. Research-backed. Entirely private.

The question StyleMatch answers: does this submitted piece of writing carry the same stylistic fingerprint as a known sample by the same student? Inspect describes the writing process; StyleMatch compares two pieces of writing.

  • Grammatical Function Words — Burrows' Delta over 30 core function words; among the most stable authorship signals in computational stylistics
  • Discourse Marker Profile — Burrows' Delta over 29 connectives & stance markers; reflects reasoning habits
  • Punctuation Fingerprint — comma, semicolon, dash, colon, !, ? rates per 1,000 characters
  • Avg Sentence Length — characteristic length that persists across writing tasks
  • Sentence Rhythm (SD) — standard deviation of lengths; captures varied vs. uniform rhythm
  • Type-Token Ratio — vocabulary range; interpreted cautiously across different text lengths
  • Flesch-Kincaid Grade Level — reading complexity; stable across tasks for the same writer
  • Pairs naturally with Inspect's Process View report — together they answer how was this written and does it sound like them
  • All computation runs locally — no text ever transmitted
  • Minimum 500 words per sample required
$1.99/month
or $19.99/year — save 16%
See full pricing ↓
StyleMatch is a probabilistic tool. Genre-constrained writing (lab reports, TOK essays) may show inflated similarity on some metrics. Always apply professional judgment. All reports include a mandatory disclaimer.

"I want to be thorough before I have this conversation."

AI-assisted deep stylometric analysis — examine internal consistency within a single essay, or compare it against a controlled writing sample across ten research-grounded dimensions.

  • Consistency Report: chunks essay into segments, scores 6 dimensions each
  • Baseline Deviation Report: 10-dimension comparison vs. controlled sample
  • 🧪 Data & Methods Report: every reported statistic recomputed in code
  • AI-suggested follow-up questions — no text retained after analysis
🔬 Consistency Report — S. Okonkwo 0 of 29 chunks flagged
Stylometric Heatmap — 29 chunks · 0 flagged
1
2
3
4
5
6
7
8
Register5.6 / 10
Vocabulary5.8 / 10
Argument Depth6.0 / 10
Error Density7.5 / 10
🔬 Baseline Deviation — S. Okonkwo 2 Anomalous
2
Anomalous
2
Notable
6
Within Range
Function Word Distribution — 43/100 similarity
Error & Mechanics Fingerprint — characteristic errors absent
Syntactic Patterning — notable shift in sentence openings

Three reports. One conversation.

  • Consistency Report — analyses a single essay in 29 chunks across Register, Vocabulary, Sentence Complexity, Argument Depth, Error Density, and Cohesion. Produces a heatmap and writer baseline profile.
  • Baseline Deviation Report — compares submitted work against a controlled sample across 10 dimensions: Function Word Distribution, Orthographic Habits, Punctuation Micro-Patterns, Syntactic Complexity, Syntactic Patterning, Cohesive Device Profile, Error & Mechanics Fingerprint, Nominal vs. Verbal Style, Phrase Recurrence, and Lexical Diversity.
  • 🧪 Data & Methods Integrity Report — new for 2026–27, for lab reports and IAs: every reported statistic (t-tests, p-values, means, correlations, percentage changes) recomputed in code from the numbers printed in the document, plus protocol feasibility arithmetic and missing-data checks. An inconsistency is a question, not a finding.
  • Each dimension rated: Anomalous / Notable / Within Range
  • AI-generated recommended next steps and suggested follow-up questions for the student conversation
  • Research citations included — 10 peer-reviewed sources
  • Printable PDF report with Report ID for institutional records
  • No text retained after analysis is complete
  • Tokens never expire — use them when the situation calls for it
1 token/report
A token is $0.60 in the starter pack  ·  10 for $6  ·  35 for $18  ·  100 for $45
See all token packs ↓
Verify transmits submitted text to PaperTrail's secure backend and the Anthropic API for processing. Text is not retained after the report is generated. See our Privacy Policy for full details.

"Where did this quote come from?"

Every citation in the essay checked against the source itself — does the bibliography match the in-text citations, does the source exist, and does it actually say what the essay claims it says.

  • Bibliography ↔ in-text cross-match — both directions, before any tokens are spent
  • Live web search locates each source — never AI memory
  • Quoted passages searched in the reachable source text
  • "What the essay says" vs. "what the source says," side by side
  • A suggested question per finding — never a verdict
📚 PaperTrail Verify Citations 11 citations checked
6 located  ·  3 access limited  ·  2 not located  ·  2 works-cited entries never cited in the text
Whitfield, R. (2019). The Empirical Habit. Halloway Press.
✓ Located — full text ◐ Notable difference
What the essay says
Quotes Whitfield: "observation is the engine of all discovery."
What the source says
The quoted passage was not found in the source text.
Worth asking: where in the book did they find this quotation?
Tanaka, M. (2020). Measurement and inference in field biology. [Print]
◌ Access limited — print only Couldn't check
"Not located" is a question, not a finding. Educator judgment required.

Existence is table stakes. Claim-vs-source is the point.

The question Verify Citations answers: modern misuse rarely invents whole sources — it invents what they said. Real bibliography, invented specifics.

  • Cross-match — bibliography vs. in-text citations, both directions, plus format observations (free, before any tokens are spent)
  • Existence — live web search across journal databases, publisher pages, and Google Books: Located / Not located / Access limited
  • Claim vs. source — quoted passages searched in the reachable text; figures matched exactly; paraphrases compared with both sides quoted
  • Data plausibility — numbers sanity-checked, phrased only as questions
  • Evidence basis stated on every read — full text, abstract, snippet, or description
  • Suggested question per finding for the student conversation
  • Essay text not retained after the report is generated
2 tokens /essay
Covers 25 citations, +1 token per additional 25 — from $1.20 with the starter pack
See how a check runs →
Paywalled and print-only sources routinely fail lookup — which is why the report says "Not located," never "fabricated." A finding is an agenda for a conversation, not a determination of misconduct.

"Walk me through the choices you made."

The essay generates its own defense: AI drafts questions about the student's authorial choices, you curate them, and the student answers out loud at a private link — one take, no account.

  • AI-drafted question pool from the essay itself — you approve, edit, or replace every question
  • Students record short spoken answers — no account, questions hidden until read aloud
  • Answers auto-transcribed, with on-demand audio playback
  • Report describes how the student accounted for each choice — described, never judged
  • Audio auto-deleted after export or 30 days
🎙️ PaperTrail Oral Defense Question 2 of 5
Why did you open with the counterargument instead of your own thesis?
Question read aloud · 10s to prepare
0:47 Recording — one take, up to 2:00
🎙️ Ownership Report — E. Kowalski
✓ Accounted for Q1 — explained the Steinbeck example from memory, added context
✓ Accounted for Q2 — described the counterargument structure as deliberate
◐ Brief Q3 — thin on why this source over alternatives
⚑ Couldn't account Q4 — couldn't explain the conclusion's key term
Describes how the student accounted for their writing. Not a verdict — educator judgment required.

The essay generates its own defense

The question Oral Defense answers: can the student account for the choices in their own essay — out loud, in their own words, two weeks later? A genuine writer can talk about why they wrote what they wrote.

  • Generate — paste or auto-fill the essay; AI drafts ~10 questions about authorial choices: why this example, why this conclusion, why this structure (1 token)
  • Curate — you are the gatekeeper: keep 4–6, edit any, add your own
  • Send — the student gets a private link; no account, no app (1 token)
  • Defend — each question is hidden until it's read aloud, then 10 seconds to prepare, then recording starts — one take, up to 2 minutes
  • Report — per-question markers (✓ accounted for · ◐ brief · ⚑ couldn't account · — not answered), evidence from the transcript, and suggested follow-ups
  • Transcripts + on-demand audio playback in your dashboard
  • Audio permanently deleted on export (or 30-day backstop); report kept 90 days
2 tokens /defense
1 to generate questions + 1 to send — from $1.20 with the starter pack
See how a defense runs →
Oral Defense describes how a student accounts for their writing — it is not a lie detector and produces no verdict. Nerves, language background, and accommodations matter; the report exists to support your conversation, never to replace it.

"I want to see how they actually write."

A focused, timed writing environment for in-class assignments. Students write; the session records the process. Teachers see a full behavioral log on submission.

  • Timed sessions with teacher-controlled join code
  • Paste detection, focus-loss events & process log
  • Student transparency screen — they know what's captured
  • Session reports: submission text + behavioral timeline
PaperTrail Write Essay: The Ethics of AI
38:14
This session is being documented
Work is saved automatically
Save Now
Submit Essay →
Process Log — Teacher View
09:04:12 SESSION START Jordan M. — joined
09:06:38 PASTE 214 chars — external source
09:18:55 FOCUS LOST Window left — 42s
09:46:02 SUBMITTED 487 words — 41m 50s

What Write gives you that a shared Google Doc doesn't

  • Timed sessions — set a duration; students see a countdown; the window locks on expiry. No extensions without a deliberate teacher action.
  • Join-code access — students join via a short code, no account needed. Teachers control who can submit and when.
  • Paste detection — every paste event is logged with a timestamp, character count, and whether the content was subsequently deleted.
  • Focus-loss tracking — if a student leaves the writing tab, the duration and timestamp are recorded in the process log.
  • Student transparency — a mandatory, non-skippable screen tells students exactly what is and isn't captured before they begin writing.
  • Session reports — on submission, the teacher receives the full essay text alongside a structured process log. Reports can be downloaded as ZIP.
  • No student accounts — students never register or log in. Nothing is stored beyond the teacher's download window.
  • Plain text, no assistance — spell check, grammar check, and autocorrect are all off. Students write without AI-augmented support.
$49/yr
Includes unlimited sessions
Go to Write →
Write captures behavioral metadata — paste events, focus changes, and timestamps. Students are shown exactly what is recorded before the session begins. No clipboard contents, keystrokes, or off-tab activity are accessible.

An evidence base, not a verdict

Three independent signals. Together, an honest evidence base.

No single tool can determine whether work is authentic. PaperTrail is built around the idea that defensible educator judgment comes from multiple independent sources of evidence — each answering a different question.

📊 Process View FREE

How was this written?

Sessions, paste activity, struggle moments, and writing-process metrics — a printable observational report on how the document was assembled. Free, runs entirely in your browser.

🔍 StyleMatch

Does this sound like them?

Seven stylometric metrics compare the submitted writing to a known sample by the same student. The only PaperTrail tool that compares two pieces of writing against each other.

📑 Source matching External

Did the words come from somewhere?

Tools like Turnitin or your institution's source-matching service answer a question PaperTrail does not — surfacing matches against published sources, prior submissions, and AI-generated text.

A document that is flagged by source-matching, shows stylistic divergence in StyleMatch, AND shows surface editing only in Process View is a different conversation than any of those signals alone. The evidence stack is what makes the conversation defensible.

Verify (AI-assisted close reading), Verify Citations (every source checked against what it actually says), and Oral Defense (the student accounts for their writing, out loud) are further signals available when the situation calls for more — see below.

Pricing

Start free. Upgrade when you need to.

Inspect is free forever, no account required. StyleMatch is a small subscription; Verify, Verify Citations, and Oral Defense run on PaperTrail Tokens — pay only when the situation calls for more.

✏️ Inspect

Free

No account required

$0
  • Full revision timeline
  • Paste detection & paste log
  • Session history
  • ⚡ Struggle moment detection
  • Paste novelty analysis
  • Revision playback
  • 📊 Process View report (PDF)
  • StyleMatch authorship comparison
  • Verify deep analysis reports
⬇ Add to Chrome — Free
🔬 Verify  +  📚 Citations

Pay per report

AI-assisted close reading & citation checking

1token/report

Citations check: from 2 tokens per essay

  • Everything in Inspect
  • Consistency Report — chunks essay into segments, scores 6 dimensions each (1 token)
  • Baseline Deviation Report — 10-dimension comparison vs. controlled sample (1 token)
  • 🧪 Data & Methods Report — every reported statistic recomputed in code, for lab reports and IAs (1 token)
  • Citations Report — every source checked against what it actually says, via live web search (from 2 tokens)
  • AI-suggested follow-up questions for the student conversation
  • Printable PDF with Report ID for records
  • No text retained — tokens never expire
Buy PaperTrail Tokens
🎙️ Oral Defense

Pay per defense

The student accounts for their writing — out loud

2tokens/defense

1 to generate questions + 1 to send — from $1.20

  • AI-drafted questions from the essay itself — you curate every one
  • Students answer by voice at a private link — no account
  • Questions hidden until read aloud — like a real defense
  • Auto-transcribed answers + audio playback
  • Per-question ownership report — described, never judged
  • Audio auto-deleted after export or 30 days
Buy PaperTrail Tokens
🪙 PaperTrail Token Packs
One shared wallet for everything token-based. Tokens never expire. No subscription required.
What's a PaperTrail Token?

The pay-as-you-go unit for PaperTrail's AI-assisted tools. A token is $0.60 in the starter pack, down to $0.45 in the year pack. Buy once, spend from the same wallet:

🔬 Verify report · 1 token 📚 Citations check · 2 tokens* 🎙️ Oral Defense · 2 tokens

*Covers 25 citations, +1 token per additional 25.  Oral Defense: 1 token to generate questions + 1 to send.

Starter
10
tokens
$6.00
$0.60/token
Buy Now
Semester
35
tokens
$18.00
$0.51/token
Buy Now
Year Supply
100
tokens
$45.00
$0.45/token
Buy Now

Common questions

Questions teachers actually ask

Is PaperTrail an AI detector?

No. AI detectors return a probability — a percentage chance that text was AI-generated — and that number is often wrong, especially for ESL students and neurodivergent writers whose natural phrasing can read as "unnatural" to a model. PaperTrail doesn't guess at authorship from the text alone. Inspect shows you the document's actual revision history. StyleMatch compares writing style against a student's own past work. Verify adds AI-assisted analysis on top of that evidence — but the output is always a report you interpret, never a score that decides for you.

Is PaperTrail a plagiarism checker?

No — that's a different problem. Plagiarism checkers like Turnitin scan submitted text against a database of existing sources to find copied passages. PaperTrail doesn't do that. Verify Citations checks something more specific: whether the sources a student cites actually say what the essay claims they say. That catches fabricated or misrepresented evidence, which a plagiarism checker won't.

How can I tell if a student used ChatGPT to write an essay?

No single signal proves it either way, which is why PaperTrail is built around three independent ones instead of one verdict. Inspect shows the session history — how the document was actually typed, including any large paste events. StyleMatch flags whether the submitted essay's sentence structure, vocabulary, and punctuation diverge from that student's own controlled writing sample. Verify adds a deeper AI-assisted read of internal consistency, and Oral Defense lets the essay generate its own follow-up questions for the student to answer out loud. None of these alone is proof. Together, they're a documentable, defensible basis for a conversation.

Why do AI detectors give false positives?

Because they're built to output a probability, not a fact — and that probability is trained on patterns that don't hold evenly across writers. Formulaic, tightly-structured writing (the kind ESL students and many neurodivergent students are taught to produce) frequently scores as "more likely AI" even when it's entirely their own. That's a structural problem with score-based detection, not a tuning issue. It's why PaperTrail doesn't produce a single AI-probability number — we show you what happened in the document instead of asking you to trust a model's guess about the text.

How is Inspect different from Draftback or Revision History?

Inspect does the same core thing — replay a Google Doc's edit history — and it's free, with no account required, rather than a paid subscription. Where Inspect goes further is in what it surfaces automatically: paste-and-novelty scoring, struggle-moment detection, and a printable Process View report you can hand a student or parent, not just a scrubber timeline you watch by hand.

What's the difference between StyleMatch and an AI detector?

An AI detector asks "does this text look machine-generated?" StyleMatch asks a narrower, more answerable question: "does this sound like the same person who wrote their in-class sample?" It's authorship consistency, not AI-generation probability — seven stylometric metrics (function words, punctuation patterns, sentence rhythm, and more) computed client-side, nothing transmitted. A ghostwritten essay and a heavily AI-edited essay can both fail a style match without either one tripping an AI detector.

Can PaperTrail help a student prove they didn't cheat?

Yes — that's half the point. A revision history that shows steady, incremental work over several sessions is real evidence, and it goes both directions. Inspect's Process View report and Oral Defense's recorded answers give a student something concrete to point to, not just a denial.

For Schools & Departments

Built to pass the IT review.

When a teacher brings PaperTrail to their department head or IT coordinator, we want the answer to be easy.

  • Inspect (with the Process View report) and StyleMatch run entirely client-side — no student text ever leaves the browser
  • No student accounts, logins, or data collection of any kind
  • Inspect reads revision metadata only — not document content
  • Verify text processed securely and not retained after analysis
  • No advertising, no data resale, no third-party tracking
  • Teacher accounts only — students never interact with the product
Contact us for school & department inquiries
🔒
FERPA
Conscious
🇪🇺
GDPR
Aware
🚫
Zero Student
Data

Get in Touch

Every email gets a real reply.

Questions about how PaperTrail works, privacy concerns, school and department inquiries — PaperTrail Academic is built by a teacher.