R0027/2026-03-26/Q003/H3¶
Statement¶
Coverage is partial and inconsistent. Some guides address multilingual topics and/or offer translations, but coverage is inconsistent and no formal standard exists. The gap is primarily in vendor documentation and formal standards.
Status¶
Current: Supported
H3 is the best-supported hypothesis. The evidence shows a clear pattern: (1) All three major vendors provide English-only prompt engineering guides with no multilingual content. (2) The most comprehensive community guide is available in 14 languages. (3) ISO/IEC has AI management standards but nothing specific to prompt engineering. (4) Academic research on multilingual prompting is substantial but has not been incorporated into practitioner guides.
Supporting Evidence¶
| Evidence | Summary |
|---|---|
| SRC01-E01 | OpenAI: English-only, no multilingual section |
| SRC02-E01 | Anthropic: English-only, no multilingual section |
| SRC03-E01 | promptingguide.ai: 14 languages — community fills the vendor gap |
| SRC04-E01 | Google: English primary, minimal Spanish/Portuguese |
Contradicting Evidence¶
No evidence directly contradicts H3.
Reasoning¶
The evidence paints a consistent picture of a significant gap between: (a) the multilingual nature of the AI user base, (b) the multilingual capability of the models themselves, and (c) the language and content of the official guidance for using those models effectively. Vendors build multilingual models but provide English-only prompting guides. This inconsistency is the core finding.
Relationship to Other Hypotheses¶
H3 subsumes the strongest elements of both H1 (some coverage exists via community) and H2 (vendor/standards coverage is missing). It provides the most accurate and nuanced characterization of the landscape.