How to Run a 30-Day Student Authorship Verification Pilot at Your University

A successful pilot does not ask, "Did the software catch cheaters?" It asks a more useful question: Can students and staff reliably produce, understand, review and verify writing-process evidence in ordinary coursework, without creating unacceptable privacy, accessibility or workload problems?

This guide provides a proposed 30-day evaluation framework for universities considering AuthMarks or another authorship-verification platform. It is a recommended pilot design, not a promise of a particular free trial, guaranteed implementation timeframe or guaranteed outcome. Confirm commercial scope and activation arrangements directly with the vendor.

Disclosure: AuthMarks provides writing-process evidence software and publishes this guide. The test framework is vendor-neutral by design. AuthMarks does not adjudicate student misconduct.

Still choosing vendors? The university authorship verification software buyer's guide frames the categories of evidence a pilot should assess.

The pilot's four decisions

At the end of the month the institution should be able to decide:

1. Instructional fit: Do the writing tasks and teaching workflows benefit from process evidence?
2. Evidence quality: Are normal drafting, pasting, revisions, gaps and attestation communicated clearly enough for meaningful review?
3. Operational fit: Can the university manage integration, access, accessibility, training, consent, retention and support?
4. Commercial fit: Is the workflow worth the purchase, support and review cost compared with alternatives?

A pilot that produces only positive screenshots but no evidence of failures cannot answer these questions.

Before day 1: define a deliberately narrow cohort

Choose one or two courses or departments, ideally with different writing workflows, and recruit faculty members who actually assess drafts or written assignments. A larger cohort is not automatically better; a controlled sample with useful observation often reveals more.

Agree the following in writing:

Item	Recommended pilot decision
Executive sponsor	One faculty or academic-integrity owner empowered to accept findings
Teaching lead	One instructor per course responsible for assignment design
Technical owner	LMS administrator plus a vendor integration contact
Privacy owner	Institution's privacy/data-governance representative
Accessibility owner	Disability/accessibility services representative
Student experience	Plain-language notice, consent procedure and non-punitive alternative
Scope	Which assignments, editors, devices and users are in the pilot
Exclusions	No covert monitoring; no algorithm-only misconduct conclusions
Data	Access, retention, exports and deletion/archival terms
Exit decision	Expand, retest with fixes, or stop

Important: Do not use student participation or refusal as evidence of dishonest behaviour. If recording is optional, measure voluntary uptake separately from evidence quality. If a course proposes mandatory participation, clear it through institutional policy and offer appropriate accommodation.

The 30-day schedule

Time	Main work	Output
Days 1–5	Policy, privacy, accessibility and technical setup	Approved scope and working test environment
Days 6–10	Instructor orientation and dry runs with synthetic writing	Baseline workflow, issue log and training notes
Days 11–20	Live student pilot in agreed assignments	Participation data and representative records
Days 21–25	Review evidence, support issues and student feedback	Scored evaluation and measured workload
Days 26–30	Procurement, security and educational review	Written go/no-go decision with conditions

The schedule is an illustrative project plan. Procurement and privacy approval can take longer than a month, so the institution may need to start those activities before day 1.

Days 1–5: setup and baseline

Document the current workflow before adding any software. Ask staff how they respond today to authorship concerns, what records they rely on, how long review takes, and what students are told.

For a proposed LTI deployment, the institution should test actual launches, permissions and submission links in a non-production course. AuthMarks publishes LTI 1.3 setup documentation, including Canvas and Blackboard paths, but registration and tenant activation are distinct steps. Do not treat a documented integration as proof that every permission, gradebook placement or launch has been validated in your environment.

Set up a consent explanation students can understand in under two minutes:

This assignment offers a way to preserve evidence of your writing process. Recording applies only when you explicitly enable it for the document. The record may show typed, pasted, suggested and unobserved writing. It is evidence for human review, not an automatic misconduct finding. You may ask about an alternative arrangement or raise a privacy/accessibility concern through the institution's normal channels.

This is an illustrative communication, not legal advice or an assertion that every course can offer identical consent rights.

Days 6–10: six controlled test scenarios

Run these with synthetic or explicitly volunteered test documents, not real disciplinary cases. Keep a dated record of each scenario and expected behaviour.

Test A — A normal essay draft

A participant drafts, revises and deletes passages over several short sessions. Check whether the instructor can understand the process without expert assistance. Record what is visible and what is not.

Test B — Legitimate external quotation

A participant pastes a properly cited quotation, then writes original commentary. The tool should allow the reviewer to distinguish evidence of pasting from a claim that pasting is prohibited. Pasted does not mean cheated.

Test C — Unrecorded beginning

Start recording after part of the document has already been written. Inspect whether the system identifies the gap or unobserved passage. Do not grade a vendor solely on whether every passage has a confident label.

Test D — Approved AI assistance

Create a writing task that permits certain AI uses, and require the author to declare them. Determine whether the system records useful context without misrepresenting the university's academic policy.

Test E — Interrupted connection or recording

Pause capture, switch networks or end the session partway through a synthetic draft. Record the actual behaviour and confirm that reviewers are not shown invented certainty for missing periods.

Test F — Export, access and independent check

Ask a second reviewer—someone other than the person who generated the document—to open the retained evidence or check an exported artifact. Distinguish seal integrity from proof that the human attestation is truthful.

These are functional and interpretability exercises, not penetration tests or tests of proprietary threat-detection mechanisms. Any separate adversarial study should use authorized test accounts and a scoped, agreed protocol.

Days 11–20: a real assignment, with safeguards

Give learners a meaningful writing assignment with the same assessment criteria as usual. Make clear which writing tools are accepted, when recording is expected, which staff can see the record, and how a learner can contest a factual error.

Keep three counts separate:

• Invited learners — those eligible to take part.
• Participating learners — those who explicitly enabled the tool as defined by the pilot.
• Usable evidence packages — submissions containing the expected evidence and accessible to a reviewer.

Also record exception categories: unsupported device, accessibility accommodation, declined consent, missing evidence, editor incompatibility, login problem, unclear instructions, offline work and technical failure.

Do not publish student writing, voice recordings, personal data, or individual suspected misconduct examples as marketing material without explicit authorization and appropriate protections.

Days 21–25: score the evaluation

Use a five-point scale, but report the raw measurements as well. Suggested dimensions and weights (illustrative, not industry standards):

Dimension	Weight	Evidence to inspect
Evidence interpretability	25%	Faculty can explain what each label and gap means
Workflow reliability	20%	Expected records arrive and can be opened
Consent, privacy and fairness	20%	Notices, permissions, alternatives and escalation
Instructor time	15%	Median minutes per routine review and escalated case
Technical integration	10%	LTI launch, roles, linking, error recovery
Student usability	10%	Feedback, activation and accessibility issues

For each criterion, score 1 = unacceptable, 3 = meets the agreed pilot threshold, 5 = exceeds it. Pre-register thresholds before reviewing results. A high weighted score should not override a serious privacy or accessibility failure.

Example decision gates

A department might require all of these:

• No unresolved critical privacy, security or accessibility issue.
• Reviewers can identify untracked text and explain why it is not evidence of misconduct by itself.
• Sample sealed records can be independently checked, where the vendor claims this capability.
• Technical issues have a documented support path.
• Staff time is measured and acceptable against an agreed budget.
• Evidence-sharing and retention permissions match the institution's approved policy.

Avoid promising a universal 95% success threshold or similar numbers without knowing the educational context. The institution should choose thresholds before testing, then publish actual outcomes honestly.

Days 26–30: deliver a decision memo

Use a one-page decision record that includes:

Recommendation: Expand / Retest / Stop.

Scope tested: courses, writers, platforms, submission count, date range.

Observed outcomes: participation, usable evidence, median review time, most common support failures, consent/accessibility feedback.

Outstanding risks: anything not tested, unsupported platforms, export/retention uncertainties, legal or procurement conditions.

Next condition: which technical, policy or contract changes must be completed before expansion.

A credible pilot report includes limitations, not just screenshots that favour a vendor.

How to arrange an AuthMarks pilot

The AuthMarks institution page describes consent-based capture, named attestation, sealed records, an instructor Action Center, and optional LMS-linked submissions. The LTI guide explains administrator registration and vendor-side activation for supported LMSs.

For an evaluation request, email sales@authmarks.com with:

• Institution, department and sponsor.
• LMS (Canvas, Blackboard, Moodle, Brightspace or other).
• Intended cohort and number of assignments.
• Preferred writing environments and device types.
• Whether the goal is instructor review, procurement evaluation or technical integration.
• Required privacy, accessibility and evidence-retention conditions.

Ask for the exact pilot scope, availability, cost, support and start conditions in writing. Do not assume a free 30-day institutional licence merely because individual student accounts are free.

Frequently asked questions

Can a pilot prove that cheating has stopped?

No. A month-long implementation study can assess workflows and evidence usability; it generally cannot establish the true incidence of cheating or prove a causal reduction in misconduct.

Can students write outside AuthMarks' native editor?

The current institution page describes browser workflows for Google Docs and Word plus a separate AuthMarks editor. Confirm which surfaces are generally available versus pilot/early-access before scoping a course.

Can a sealed record determine guilt?

No. A sealed record can support review and help establish artifact integrity, but institutions must apply their policies and consider context, alternative explanations and student responses.

Is LMS integration required for the pilot?

Not necessarily. A non-LMS workflow may be useful for initial evidence tests. If the eventual purchase decision depends on Canvas or Blackboard integration, however, test that integration during the pilot rather than assuming it will work based solely on a vendor demo.

Related: Procurement checklist · Pricing and total-cost guide · Authorship verification software buyer's guide · Canvas integration guide · Blackboard integration guide