Academic Integrity After COVID: What the Data Shows

TL;DR

COVID-19 forced rapid online-assessment adoption. Misconduct rates spiked 25–80% in 2020–21 across various institutions. Detection improved through 2023. ChatGPT (late 2022) then added a new misconduct category before COVID-era issues fully resolved. Post-pandemic infrastructure investment is durable; the misconduct landscape has changed permanently.

The COVID shock (March 2020 – mid 2022)

What happened

When universities globally moved teaching and assessment online from March 2020:

• Most institutions had no robust online proctoring infrastructure
• Examination question banks were not designed for online use
• Faculty had limited training in online assessment design
• Students faced novel assessment situations without prior experience
• Stress and instability created conditions favourable to misconduct

Measured impact

Multiple peer-reviewed studies documented misconduct spikes:

• Lancaster & Cotarlan (2021) — Chegg query data showed substantial increases [verify specific numbers]
• Various institutional FOI data — confirmed misconduct cases increased 25–80% in 2020–21 across UK institutions
• Newton & Essex (2024) — meta-analysis documenting cross-country patterns [verify]

The increases reflected both higher actual misconduct rates and improved (or newly possible) detection in online environments.

What happened next (mid 2022 – late 2022)

By the time most institutions had returned to in-person teaching, online assessment infrastructure had matured:

• Proctoring software (Proctorio, ProctorU, Examplify, etc.) widely deployed
• Question banks redesigned for online security
• Assessment-design training extended across faculty
• Institutional integrity offices significantly expanded

The expected trajectory was a return to pre-pandemic misconduct levels with stronger institutional infrastructure. This trajectory did not complete.

The ChatGPT disruption (November 2022 onwards)

ChatGPT's launch in late November 2022 disrupted the recovery before it completed:

• An entirely new misconduct category (AI-generated submissions) emerged
• Detection capability for the new category was minimal
• Student awareness and uptake was rapid (100 million users within two months of launch)
• Institutional policy frameworks had not anticipated the technology

The AMI's D2 dimension tracks the new category. The maxed D2 scores in multiple countries (Colombia, Argentina, Greece, Egypt, Iran, Saudi Arabia, Italy, France, Spain, Poland — 11 countries at 100) reflect demand signals for AI submission tools.

What the AMI data shows about the post-COVID period

Detection investment paid off

Q1 countries (Australia, UK, Ireland, Canada, NZ, Netherlands, US) have R_det (Detection sub-score) values from 65 to 90 — substantially higher than pre-pandemic baseline. The investment in detection infrastructure during the pandemic transferred to AI detection capability.

Disclosure infrastructure expanded

Mandatory institutional misconduct reporting requirements emerged in several jurisdictions during 2021–2024. The UK 2022 Skills and Post-16 Education Act partly reflected the post-COVID context.

The legislative wave

Three specific contract cheating bans came after the start of COVID:

• Ireland 2019 (pre-COVID)
• Australia 2020 (COVID context contributed to passage timing)
• UK 2022 (post-COVID, with COVID-era data feeding policy case)

The COVID context accelerated policy attention to academic integrity.

What the research literature shows

Newton & Essex (2024) meta-analysis

Newton & Essex [verify specifics] analysed cross-country misconduct trends through the pandemic period. Key findings:

• COVID-era detected misconduct rates increased substantially in jurisdictions with mature detection
• Underlying rate increases were smaller than detected-rate increases (detection improvement effect)
• Post-pandemic rates partially returned toward baseline but remained elevated

Discipline-specific patterns

Misconduct increases were not uniform across disciplines:

• STEM courses with problem sets — significant collusion increases
• Essay-based humanities and social science — significant contract cheating increases
• Open-book final examinations — substantial new misconduct opportunities
• Project-based and demonstrable-understanding assessment — smallest increases

The pattern informed post-pandemic assessment redesign approaches.

The institutional infrastructure that persisted

Detection deployment

Institutions that licensed detection tools during the pandemic typically retained them post-pandemic. The deployment ratchet is largely one-way.

Integrity office expansion

Most major universities expanded their integrity offices during 2020–2022. The expansion has been substantially retained even after returning to in-person teaching.

Online assessment options

Many institutions retained online assessment as an option for some course types even after returning to in-person. The pandemic accelerated assessment format diversification.

Faculty training

Faculty academic integrity training programmes expanded during the pandemic and have largely persisted.

What was lost or compromised

Assessment authenticity gaps

Many online assessment adaptations from the pandemic period remained in use post-pandemic — sometimes without adequate authenticity safeguards. Open-book essay assessments, for example, became more common; AI tools subsequently undermined the integrity of such formats.

Detection-incidence confound

Strong detection environments report more misconduct. The post-COVID infrastructure investment makes cross-time comparisons harder — we cannot directly compare 2019 detected-rates with 2024 detected-rates because detection capability differs.

Student culture changes

Some research suggests post-pandemic student attitudes toward online assessment integrity remain more permissive than pre-pandemic. Cultural change is slow to reverse.

Implications for the AMI

D1 (Contract cheating)

Demand signals remained elevated through 2022 but showed partial decline in 2023–2024 — likely a substitution effect toward AI tools rather than a true reduction.

D2 (AI submissions)

Created as a meaningful category by ChatGPT. The dimension did not exist pre-2022. Current AMI scores reflect 2022–2026 demand.

D5 (Collusion)

Pandemic-era online assessment created collusion opportunities. Detection of online collusion remains harder than in-person; the post-pandemic data shows elevated D5 signals in countries with mature detection.

What comes next

Detection-incidence confound stabilisation

As 2023–2024 detection infrastructure becomes the new baseline, cross-time comparisons will become possible again. The AMI's time series capability will improve.

Assessment redesign maturation

The most promising long-term response is assessment redesign that makes AI use largely irrelevant. Adoption is ongoing.

Post-COVID norm-setting

Three years after returning from pandemic-era online assessment, institutional norms around AI use, online proctoring, and integrity infrastructure are stabilising. The AMI will track which norms prove durable.

Sources

• Newton, P. M., & Essex, K. (2024) [verify exact citation]
• Lancaster, T., & Cotarlan, C. (2021), International Journal for Educational Integrity
• Institutional FOI data (UK, US, Australia)
• AMI v1.5 methodology document
• Guardian FOI investigation (June 2025)

Full methodology | Download dataset