Original research · Public research-and-methods edition

When a school questions authorship

AI disclosure, evidence and a student’s opportunity to respond in 30 California districts

The Research University · Released November 4, 2026
Publication authorized by Dr. Freedom Cheteni, Editor-in-Chief

Abstract

A student can know that AI assistance must be acknowledged and still have little guidance about what happens if a teacher questions their authorship. We examined publicly accessible documents for 15 unified and 15 high-school districts selected at random within each district type. Of 30 districts, 11 expressly required disclosure in at least one circumstance. Among these, 2 documented both a way to evaluate evidence and an opportunity for the student to respond. The documents included district guidance for teachers as well as board policies. These counts describe what the search located by November 4, 2026.

Research question

Among sampled districts with an explicit duty to disclose AI assistance, how often can a reader also find both an evidentiary standard for disputed authorship and a way for the student to present their account? California’s model policy discusses these three elements. Earlier studies and a public policy tracker have examined AI governance and disclosure. We studied whether the three provisions could be traced to documents in the same district.

Methods

We froze the California Department of Education district directory on November 4, 2026. The eligible frame contained 345 active unified districts and 76 active high-school districts. Within each group, district codes were sorted before a seeded random draw without replacement (Python seed 20261104, unified first). We selected 15 from each group before searching for policies and retained every selected district.

Searches paired each district’s name with terms for AI disclosure and academic integrity, then looked for handbooks, evidence, detectors and appeals. The review considered districtwide policies, official guidance and cross-referenced academic-integrity procedures. A rule at one school counted only if its districtwide applicability could be established. State guidance supplied context. District coding required local documentary support.

Each row represents one district at the cutoff. Disclosure required an explicit direction to acknowledge AI assistance. The evidence code required a concrete evidentiary method or standard, such as evaluating a student’s explanation or ruling out a detector score as sufficient proof. A student response required an express opportunity to explain or rebut the allegation. The analysis distinguishes explicit provisions, explicit disallowance, silence in reviewed text, unavailable material and ambiguity. Only explicit provisions enter the joint count.

Results

Documented provisions by district type
District type	Sampled	Disclosure required	All three provisions
All sampled districts	30	11	2
Unified districts	15	5	1
High-school districts	15	6	1

Within the 11 districts with a disclosure requirement, 3 had an ambiguous or unavailable code for evidence or student response. Two disclosure districts documented an evidentiary standard and three documented a student response opportunity. Searchers located applicable material in 25 of the 30 sampled districts, encountered inaccessible material in two and could not establish scope in three. Each row links to the documents reviewed and their relevant sections. Open the district evidence index.

Two ways a safeguard can appear

Oakland Unified’s family guidance calls for citing AI assistance. Its separate teacher guide directs a teacher who suspects misuse to meet privately with the student and assess their command of the assignment. That passage identifies both an evidentiary method and an opportunity to explain. It is professional guidance for teachers. Oakland sources · D02

Merced Union High’s toolkit sets a limit on what a detector result can establish. Its academic-honesty policy gives a student the opportunity to demonstrate that their use of technology conformed to policy. Reading the toolkit alongside the board policies brought the evidentiary limit and response procedure into view. Merced sources · D19

Interpretation and limitations

Where a board policy refers users to other materials, a family may have to read a staff guide and a student-facing policy together to understand the procedure. A count of policy titles alone would miss that distinction. The companion worksheet asks a district to trace one disputed assignment through its own documents.

Equal sampling across two district types with different population sizes limits interpretation of the pooled count. The study did not observe classrooms, interview students, measure sanctions, or assess legal compliance. Archived documents, inaccessible policy manuals and search terminology can change what a reviewer finds. Silence describes the text inspected. Human disagreement and interrater agreement have yet to be measured.

Review history and contributions

The private first pass reported 0 of 11. Targeted review of district sources before release identified provisions the first pass had missed. The public count above uses the corrected dataset. The original private snapshot is retained in the editorial record. Read the version notes.

Dr. Freedom Cheteni commissioned the pilot and authorized public release. AI assistance covered retrieval, initial coding, targeted source checks, computation and drafting. Reproducing the counts checks the arithmetic. Independent human review and formal journal acceptance are separate, outstanding steps. The pilot used public institutional documents and incurred no new paid-service costs.

Sources and reproducibility

Dataset CA-AI-2026-11-04-v2.7afb1874. The district table contains original annotations. The source table links to documents held by their respective publishers. Districts CSV · Sources CSV · Analysis script

Editorial status and contributor criteria are documented in the journal’s publishing policy.