# Publication checklist for an AI-assisted HTML guide

Canonical: https://githtml.com/guides/ai-documentation-publication-checklist
Published: 2026-09-26
Publisher: WaveTech LLC / gitHtml

Before publishing an AI-assisted guide, perform separate factual, operational, privacy, and presentation checks. Verify important claims against sources, label illustrative examples, remove sensitive material, and inspect the HTML with optional resources disabled. Record the reviewed revision and responsible owner so readers know the scope of the approved document.

## Review substance before visual polish

Check the purpose, audience, and central answer first. A beautifully formatted page can still describe the wrong version or invent a feature. Use a short claim list for limits, permissions, data handling, and required actions, then compare each claim with its source. NIST's generative AI profile provides background on generated-content risk; your approval needs concrete evidence from this document. Ensure that uncertainty survives in the title and summary when it matters. Remove empty sections rather than filling them with unsupported generalities to make the guide look comprehensive.

## Inspect examples and disclosure boundaries

Every example should be clearly synthetic, illustrative, or actually verified, with the distinction stated honestly. Review commands for side effects and make sure a reader cannot mistake a dangerous operation for an ordinary setup step. Check source links, screenshots, comments, and metadata for secrets or inappropriate internal details. Do not infer permission to publish from permission to read a repository. An illustrative checklist result might say factual review complete, command execution not performed, and public disclosure review pending. That is a useful status, not a reason to mark the whole guide approved.

1. Evidence: verify high-consequence claims and the document's version scope.
2. Safety: inspect commands, identifiers, links, and disclosure suitability.
3. Presentation: confirm headings, readable code, alternative text, and narrow-screen layout.

## Verify the exact output readers receive

Render the actual HTML file, not just the source draft. Follow important links, check heading order, and ensure the main explanation works with JavaScript and remote resources disabled. W3C's page-structure guidance helps assess semantic organization, but a quick visual check is not a complete accessibility audit. Fix missing content before committing through the user's normal authorized workflow. gitHtml indexes and reads committed HTML from GitHub; it does not supply an automatic AI review or publication pipeline. Keep the approval record outside decorative page metadata that readers may never see.

## Sources and further reading

- [NIST AI 600-1: Generative AI Profile](https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf)
- [W3C WAI: Page Structure Tutorial](https://www.w3.org/WAI/tutorials/page-structure/)

## Related guides

- [Write a read-only AI documentation prompt with explicit boundaries](https://githtml.com/guides/read-only-ai-documentation-prompt)
- [Build an evidence ledger for AI-generated documentation](https://githtml.com/guides/ai-documentation-evidence-ledger)
- [Verify AI-generated API documentation against the actual contract](https://githtml.com/guides/verify-ai-generated-api-documentation)
- [Review AI-generated architecture diagrams for invented relationships](https://githtml.com/guides/review-ai-generated-architecture-diagrams)

Editorial approach: https://githtml.com/guides/about