Research, technical documentation, product announcements and commentary support different kinds of claims. Inspect what was actually read and what remains unverified.
Golding, J. M., McGavran, M. B., Waikel, L. R., Phillips, N. L., Lynam, D. R., Lippert, A., & Miller, J. D. (2026). The use of generative artificial intelligence by psychology graduate students in the United States. Teaching of Psychology. Advance online publication. https://doi.org/10.1177/00986283261496432Consulted source
Psychology survey · Published 2026-10-10 · Checked 2026-10-24
US clinical and nonclinical psychology graduate students.
Access: Publisher abstract and online publication date inspected; full article not accessed.
Interpretation limits: Self-reported use is not evidence of teaching effectiveness. Restricted access prevented review of sampling and analyses.
Castellana, M. A., Imundo, M., Christhilf, K., Hansen, W., Szeli, É., McNamara, D. S., & Arner, T. (2026). Training undergraduate psychology students in ai-supported scientific writing in a research methods course. Psychology Learning &, Article 14757257261471430. https://doi.org/10.1177/14757257261471430Consulted source
Psychology course study · Published 2026-08-19 · Checked 2026-10-24
Two undergraduate research-methods course sections.
Access: Publisher abstract and first-online date inspected; restricted full article not accessed.
Interpretation limits: Section allocation and detailed methods were not inspected. This is not evidence that every form of AI writing instruction fails.
Bailey, P. E., Parylo, A., Tremayne, K., O’Gorman, C., & Watson, P. (2026). Predictors of psychology student GenAI use and grades on GenAI secure versus unsecure assessments. Psychology Learning &, 25(2), 158–174. https://doi.org/10.1177/14757257261430600Consulted source
Psychology observational study · Published 2026-03-12 · Checked 2026-10-24
University of Technology Sydney; data concern early 2024.
Access: Publisher abstract, introductory context and limitations inspected; tables and complete methods not audited.
Interpretation limits: Associations do not establish causation, authorship or the absence of misconduct. General coursework use was self-reported.
Delvenne, J.-F. (2026). Navigating GenAI in psychology education: Assessment validity, academic integrity, and the realities of teaching in an ai-rich era. Psychology Learning &, 25(2), 113–126. https://doi.org/10.1177/14757257261427151Consulted source
Psychology review · Published 2026-03-04 · Checked 2026-10-24
Discipline-focused review informed by UK academic-integrity practice.
Access: Publisher abstract, introduction and challenge table inspected; cited studies not independently re-reviewed.
Interpretation limits: An interpretive review and practice perspective, not a trial validating a particular assessment redesign.
Wu, Xinning, Zhu, Pei, Zhang, Jinliang, Yin, Mengwei, Wang, & Yingxi. (2026). ChatGPT’s impact on student learning outcomes: A meta-analysis of 35 experimental studies. Humanities and Social Sciences Communications, 13(1), 684. https://www.nature.com/articles/s41599-026-07019-z
Meta-analysis · Published 2026-03-26 · Checked 2026-10-24
Studies published in 2022–2024 across educational contexts.
Access: Publisher abstract and date inspected. Crossmark distinguishes March 26 first online from May 19 version of record.
Interpretation limits: A pooled effect does not predict benefits in Psy101 or establish durable independent learning. Included studies were not individually audited here.
Professional guidance · Updated 2026-03-19 · Checked 2026-10-24
University of Queensland guidance hosted in TEQSA’s toolkit.
Access: Official case-study text inspected; date is the page’s last-updated date, not an established original-publication date.
Interpretation limits: This is design advice, explicitly not a mandatory policy position. It does not replace local institutional rules or demonstrate intervention effectiveness.
Working paper · Published 2026-06 · Checked 2026-10-24
Elementary literacy; two randomized trials.
Access: Stanford institutional summary and working-paper cover abstract inspected; full methods and analyses not reviewed. June precision only.
Interpretation limits: A school-literacy implementation study, not an introductory-psychology trial. Increased engagement and improved achievement are separate outcomes.
Access: Stanford repository summary inspected. Linked NBER paper returned 403; full paper not accessed. August precision only.
Interpretation limits: The intervention combines a practice platform with tutor access; it does not isolate an AI-only benefit. Working-paper findings may change.
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), Article e2422633122. https://doi.org/10.1073/pnas.2422633122
Randomized field experiment · Published 2025-06-25 · Checked 2026-10-24
High-school mathematics at one Turkish school, fall 2023.
Access: Publisher-indexed abstract, design and results excerpts inspected; supplements not audited. August 2025 correction concerns an author affiliation.
Interpretation limits: Short-term mathematics outcomes do not establish effects in psychology. Removing harm is not the same as showing a learning benefit.
Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). AI tutoring outperforms in-class active learning: An RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15(1), Article 17458. https://doi.org/10.1038/s41598-025-97652-6Consulted source
Randomized course experiment · Published 2025-06-03 · Checked 2026-10-24
College physics; custom tutor informed by pedagogical principles.
Access: Publisher-indexed abstract and introductory context inspected; direct article fetch failed. Complete methods, materials and supplements not audited.
Interpretation limits: This concerns particular lessons and a purpose-built tutor. It does not establish superiority of unrestricted chatbots or replacement of instructors.
Scarfe, P., Watcham, K., Clarke, A., & Roesch, E. (2024). A real-world test of artificial intelligence infiltration of a university examinations system: A “turing test” case study. PLOS ONE, 19(6), Article e0305354. https://doi.org/10.1371/journal.pone.0305354Consulted source
Psychology assessment case study · Published 2024-06-26 · Checked 2026-10-24
Five undergraduate psychology modules at one UK university.
Access: Publisher abstract, study setup, ethics, submitted-exam description and detection results inspected; not a complete methodological audit.
Interpretation limits: Marker recognition is not a test of commercial detection software. This does not estimate misconduct prevalence or learning by actual students.
Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. https://doi.org/10.1177/1529100612453266
Learning-science review · Published 2013-01-08 · Checked 2026-10-24
Conceptual background for the original teaching illustrations.
Access: Publisher abstract and PubMed abstract inspected; full review not read.
Interpretation limits: A broader review from 2013, not a study of current AI tools. Implementation and the criterion task matter; this does not establish effects for the exact examples on this site.
Vendor release notes · Undated living page · Checked 2026-10-24
Dated API changes; selected September and October 2026 entries inspected.
Access: Official full page opened; relevant dated entries read.
Interpretation limits: Living vendor record. A release is not evidence of account access or instructional effectiveness. Individual development dates are given separately.
Vendor announcement · Published 2026-10-09 · Checked 2026-10-24
Adapted keynote describing a broad work-agent direction.
Access: Official article opened; architecture, Workspace, governance, and cost-control sections inspected.
Interpretation limits: The opened article dates itself October 9; search indexing showed October 8. Use the visible article date. Broad announcement does not establish general availability for every described feature.
Vendor announcement · Published 2026-09-09 · Checked 2026-10-24
Cross-application document work and rollout.
Access: Official article opened in full; examples and rollout paragraph read.
Interpretation limits: Rollout and administrator permissions matter. Some surfaces were described as coming later. Productivity examples are vendor demonstrations.
Vendor announcement · Published 2026-09-01 · Checked 2026-10-24
Goal-directed inspection of uploaded and YouTube video through the API.
Access: Official article opened; mechanism, capabilities, availability, and benchmark descriptions read.
Interpretation limits: Performance figures are provider benchmarks. API availability and later consumer rollout are distinct; instructional interpretation remains untested here.
Vendor announcement · Published 2026-10-17 · Checked 2026-10-24
DeepSeek: dated API change log.
Access: Official October17 and September10 entries inspected. Benchmarks were not independently reproduced.
Interpretation limits: Benchmarks and service descriptions are provider reports, not evidence about learning. Earlier release plans on the site can be superseded by later change-log entries.
Vendor documentation · Undated living page · Checked 2026-10-24
DeepSeek: official search documentation.
Access: Official guide description and request/response examples inspected; no API trial performed.
Interpretation limits: Retrieval does not guarantee a complete search, full-text access or that a generated claim is supported. No classroom evaluation is reported.
Vendor documentation · Undated living page · Checked 2026-10-24
DeepSeek: official tool-calling guide.
Access: Official guide and example loop inspected; no code executed against a service.
Interpretation limits: A tool request is not execution or permission. The application controls what is actually available. No teaching effectiveness evidence.
Vendor announcement · Published 2026-01-15 · Checked 2026-10-24
Microsoft: education product announcement by Matt Jubelirer.
Access: Official announcement text and visible January15 date inspected; year comes from the canonical URL.
Interpretation limits: A product announcement and sponsored material do not establish independent teaching or learning gains. Availability statements belong to the dated announcement.
Vendor announcement · Published 2026-05-19 · Checked 2026-10-24
Microsoft: product-team post by Anoo Padte.
Access: Official indexed article text inspected; direct fetch returned no readable page text. No audio, product interaction or linked training was reviewed.
Interpretation limits: The pedagogical rationale is the vendor’s account. It is not a randomized evaluation of this product in psychology. Language, license and tenant conditions must be checked locally.
Vendor documentation · Undated living page · Checked 2026-10-24
Microsoft: current education product page.
Access: Official overview and student-access sections inspected October24; no release date inferred from the check date.
Interpretation limits: The page mixes product description with marketing claims. This review does not validate its effectiveness, contractual scope or an institution’s configuration.
Vendor announcement · Published 2026-09-08 · Checked 2026-10-24
Meta: personal-agent announcement.
Access: Official announcement sections inspected. Page also displays September30 as an update; launch and update dates are distinct. No product trial.
Interpretation limits: Vendor characterization is not a safety guarantee or an education trial. Availability and approvals require product-specific checking.
Vendor announcement · Published 2026-09-02 · Checked 2026-10-24
Meta: model release and demonstration report.
Access: Official release, agentic-workflow and demonstration-label sections inspected; benchmark methodology not audited.
Interpretation limits: Demonstrations and provider benchmarks are not independent evidence of reliable classroom operation. An example labeled prototype is not a released product.
Vendor announcement · Published 2026-10-15 · Checked 2026-10-24
Meta: open-weight release announcement.
Access: Official release, hardware example and license sections inspected; no weight download, license analysis or deployment performed.
Interpretation limits: Open weights do not mean every license permits every use, that ordinary laptops can run the model, or that a whole application is private.
Technical documentation · Undated living page · Checked 2026-10-24
Conceptual introduction to a connection protocol.
Access: Official versioned introductory page inspected; version path is not represented as the article publication date.
Interpretation limits: A protocol connection is not an agent or automatic authorization. The page does not establish educational effectiveness or the configuration of any installed tool.
Professional guidance · Published 2023-09-07 · Checked 2026-10-24
UNESCO education and research guidance; official overview updated January 16, 2026.
Access: Official overview, author metadata, original date, update date and scope paragraphs read. Earlier transferred book-ledger scope remains separately preserved.
Interpretation limits: Guidance, not a classroom experiment or a determination of local legal requirements. The book was not reviewed in full for this entry.
Autio, C., Schwartz, R., Dunietz, J., Jain, S., Stanley, M., Tabassi, E., Hall, P., & Roberts, K. (2024, July 26). Artificial intelligence risk management framework: Generative artificial intelligence profile. https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf
Governance framework · Published 2024-07-26 · Checked 2026-10-24
NIST AI 600-1, a voluntary cross-sector companion to AI RMF 1.0.
Access: Official NIST publication metadata and PDF introduction; selected definitions and privacy/bias passages on printed pages 1–9; governance passages on printed pages 21/26/27 and deactivation recommendations on page 41 inspected. The full 64-page PDF was not reviewed.
Interpretation limits: A governance framework, not evidence about a particular course, product or compliance outcome. Referenced empirical studies were not independently audited here.
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), Article e2422633122. https://doi.org/10.1073/pnas.2422633122
AI learning trial. Checked 2026-10-23.
Original inspection scope: Publisher-indexed full article sections and PMC record
Randomized high-school mathematics field trial, nearly 1000 students in Turkey. Unguarded assistance harmed later unaided performance; specialized tutor mitigated problem. Not psychology evidence.
Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). AI tutoring outperforms in-class active learning: An RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15(1), Article 17458. https://doi.org/10.1038/s41598-025-97652-6
AI learning trial. Checked 2026-10-23.
Original inspection scope: Publisher-indexed abstract/discussion and first PDF-page excerpt; direct open failed
Custom college physics tutor trial, not generic chatbot use or an entire Intro course. Avoid unverified exact gains or claims of teacher replacement.
Multiple inclusive assessments; process and judgment; clear AI boundaries; secure points for independent learning. Expert guidance, not Intro intervention trial.
Scarfe, P., Watcham, K., Clarke, A., & Roesch, E. (2024). A real-world test of artificial intelligence infiltration of a university examinations system: A “turing test” case study. PLOS ONE, 19(6), Article e0305354. https://doi.org/10.1371/journal.pone.0305354
Psychology assessment study. Checked 2026-10-23.
Original inspection scope: Full publisher article
GPT-4 submissions across five undergraduate psychology modules at one UK institution; 94% missed by markers. Does not estimate actual student misconduct or AI-detector accuracy.
Original inspection scope: Official page indexed text
A detector score alone is insufficient for allegations; false positives and mixed/edited text remain problems. Detector score is not probability the paper was AI-generated.