R0029/2026-03-27/Q003/H3¶
Statement¶
AI authorship is universally prohibited, but the specifics of disclosure, accountability, and enforcement differ significantly across venues.
Status¶
Current: Supported
The evidence shows a two-layer reality: absolute consensus on AI authorship prohibition, but meaningful variation in disclosure requirements. Key differences include: where to disclose (dedicated statement vs. Methods vs. Acknowledgments), what triggers disclosure (all AI use vs. only methodology-relevant use), grammar-check exemptions (some exempt, some do not), and peer review restrictions (some prohibit AI, others allow limited use).
Supporting Evidence¶
| Evidence | Summary |
|---|---|
| SRC01-E01 | Nature: no AI authorship; AI-generated images prohibited; copy editing AI does not require disclosure |
| SRC02-E01 | Science: reversed initial total ban; now allows AI assistance with acknowledgment; requires full prompts in methods |
| SRC03-E01 | ACM: Acknowledgments disclosure; grammar/spelling exempt |
| SRC04-E01 | IEEE: Acknowledgments disclosure; editing/grammar recommended but not required |
| SRC05-E01 | NeurIPS: only disclose if part of methodology; editing does not require disclosure |
| SRC06-E01 | Big 5 vary: Elsevier uses dedicated AI declaration; Springer Nature requires Methods section; Wiley allows Methods or Acknowledgments |
Contradicting Evidence¶
No evidence contradicts this hypothesis.
Reasoning¶
The two-layer finding is robust across all examined venues. Layer 1 (prohibition) is absolute: not a single venue allows AI as author. Layer 2 (disclosure) is fragmented: venues differ on location, granularity, exemptions, and enforcement. This fragmentation suggests the field has reached consensus on the principle but not the implementation.
Relationship to Other Hypotheses¶
H3 subsumes H1's correct observation (prohibition is universal) while adding the critical nuance that disclosure requirements are not standardized. H2 is eliminated. H3 is the most complete characterization.