AI Content Disclosure Template

Why You Need an AI Content Disclosure

Readers, regulators, and AI assistants increasingly want to know which parts of a page were written, drafted, or assisted by AI. A clear disclosure builds reader trust, complies with regional rules (the EU AI Act, several US state laws), and gives AI crawlers a signal about provenance that improves citation quality. The template below covers three disclosure tiers: site-wide, page-level, and per-section.

Tier 1: Site-Wide Disclosure

Place this in your editorial policy, footer, or /about page.

## Use of AI on This Site

We use AI tools in our content workflow. Specifically:

1. We draft outlines and first drafts with [model names, e.g., GPT-5, Claude Sonnet 4.6].
2. Every published article is reviewed, edited, and fact-checked by a named human editor.
3. AI-generated images are labelled in their caption and tagged with C2PA provenance metadata where possible.
4. We do not publish AI-generated content on the topics of [list sensitive topics, e.g., medical advice, legal advice, financial recommendations].
5. We disclose AI use on a per-article basis at the top of each article.

Last updated: [date]
Editor responsible: [name]

Tier 2: Page-Level Disclosure

Place this at the top of any article that used AI in its production. Adjust the language to reflect the actual workflow:

• AI-drafted, human-edited: "This article was drafted with the help of [model] and reviewed and edited by [author name] before publication."
• AI-summarised research, human-written: "Background research for this article was summarised using [model]. The article itself was written by [author name]."
• AI-translated: "This article was translated from [source language] by [model] and reviewed by a human translator."
• Fully AI-generated: "This article was generated by [model] from a curated set of inputs and reviewed by [editor name] before publication."

Tier 3: Per-Section Disclosure

For data tables, code samples, or charts that were generated or computed by AI, add an inline note:

<figure>
  <table>...</table>
  <figcaption>Source: [data source]. Table organised with assistance from [model]; reviewed [date].</figcaption>
</figure>

Machine-Readable Disclosure (for AI Crawlers)

Add structured metadata so AI assistants treating provenance as a citation signal see the disclosure even if the prose is paraphrased:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Article title",
  "author": {
    "@type": "Person",
    "name": "Editor name"
  },
  "creativeWorkStatus": "Published",
  "aiAssisted": true,
  "aiAssistedDetails": "Drafted with GPT-5, edited by human author. No AI-generated images.",
  "datePublished": "2026-05-15",
  "dateModified": "2026-05-15"
}
</script>

Tone Guidance

Disclosures should be matter-of-fact, not defensive. Avoid language that apologises for AI use; AI is now a standard editorial tool. Phrases like "AI-drafted, human-edited" describe the workflow without prescribing a value judgement. Keep disclosures to one or two sentences at most; longer disclosures dilute reader attention and rarely improve compliance posture.

Common Mistakes

1. Buried disclosures. A disclosure at the bottom of a 3,000-word article is technically present but functionally invisible. Put page-level disclosures above the fold.
2. Over-disclosing. Marking grammar checks and spell-checks as "AI-assisted" floods the signal and reduces its value. Disclose meaningful AI contribution, not autocomplete.
3. Generic site-wide language only. A blanket "we sometimes use AI" line is weaker than per-article disclosure for trust and compliance.
4. No editor named. A named human editor is the single strongest trust signal in a disclosure.
5. No machine-readable form. Schema.org markup is becoming the canonical machine-readable disclosure surface for AI assistants.