AI Content Metatag Standard v1.1

1. Introduction

The AI Content Metatag Standard provides a simple, universal way to indicate AI involvement in web content creation. As AI becomes increasingly prevalent in content generation and enhancement, this standard aims to promote transparency and trust by allowing content creators to clearly disclose the level and nature of AI involvement in their work.

The existence of this standard enables the development of tools for detecting, analyzing, and filtering AI-generated content. Search engines and other automated systems can use these metatags to flag content appropriately in their indexes, which is particularly important for training datasets used by large language models (LLMs).

2. Purpose

The purpose of this standard is to:

• Provide a clear method for indicating AI involvement in content creation
• Promote transparency in AI-assisted content creation
• Enable users to make informed decisions about the content they consume
• Facilitate the development of tools for identifying and analyzing AI-generated content
• Allow search engines and other systems to appropriately categorize and handle AI-generated content

3. Scope

This standard applies to all types of web content, including but not limited to:

• Text (articles, blog posts, comments)
• Images
• Videos
• Audio content
• Interactive elements (e.g., chatbots)

4. Core Attributes

The standard defines five primary data attributes:

4.1 data-ai-generated

• Type: Boolean
• Description: Indicates whether the content was primarily generated by AI.
• Usage: Set to "true" if AI was the primary creator of the content, "false" otherwise.

4.2 data-ai-enhanced

• Type: Boolean
• Description: Indicates whether AI was used to enhance or modify existing content.
• Usage: Set to "true" if AI tools were used to significantly modify or enhance human-created content, "false" otherwise.

4.3 data-ai-free

• Type: Boolean
• Description: Indicates that no AI was involved in creating or modifying the content.
• Usage: Set to "true" if the content was created entirely without AI assistance, "false" if any AI involvement occurred.

4.4 data-ai-tools

• Type: String
• Description: Lists the AI tools or models used in content creation or enhancement.
• Usage: Comma-separated list of AI tools or models used. Optional, but recommended when known.

4.5 data-ai-tagged-date

• Type: Date
• Description: Indicates when the content was tagged with AI metatags.
• Usage: ISO 8601 format (YYYY-MM-DD). Optional, but recommended for version control and auditing purposes.

5. Implementation

5.1 General Usage

Apply these attributes to the HTML element containing or representing the AI-involved content.

<div data-ai-generated="true" 
     data-ai-enhanced="false" 
     data-ai-free="false" 
     data-ai-tools="GPT-4" 
     data-ai-tagged-date="2024-08-26">
    <!-- AI-generated content here -->
</div>

5.2 Text Content

For text content, apply the attributes to the containing element:

<article data-ai-generated="false"
         data-ai-enhanced="true" 
         data-ai-free="false"
         data-ai-tools="Grammarly,GPT-3"
         data-ai-tagged-date="2024-08-26">
    <h1>Article Title</h1>
    <p>This article was enhanced using AI tools...</p>
</article>

5.3 Images

For images, apply the attributes to the <img> tag:

<img src="ai-artwork.jpg" alt="AI-generated artwork" 
     data-ai-generated="true"
     data-ai-enhanced="false"
     data-ai-free="false"
     data-ai-tools="DALL-E"
     data-ai-tagged-date="2024-08-26">

5.4 Videos and Audio

For videos and audio content, apply the attributes directly to the <video> or <audio> tag:

<video src="enhanced-video.mp4"
       data-ai-generated="false"
       data-ai-enhanced="true"
       data-ai-free="false"
       data-ai-tools="Adobe Sensei"
       data-ai-tagged-date="2024-08-26">
</video>

5.5 Nested Content

The AI Content Metatag Standard supports nesting, allowing for more granular tagging of content. This is particularly useful for content that contains a mix of AI-generated, AI-enhanced, and AI-free sections.

Rules for Nesting:

1. Child elements inherit the AI status of their parent unless explicitly overridden.
2. A child element can override any of the parent's AI attributes.
3. The most specific (deepest) tag takes precedence for any given piece of content.

Example of Nested Content:

<article data-ai-enhanced="true" 
         data-ai-generated="false"
         data-ai-free="false"
         data-ai-tools="Grammarly,GPT-3"
         data-ai-tagged-date="2024-08-26">
    <h1>Article Title</h1>
    <p>This paragraph was enhanced by AI.</p>
    
    <section data-ai-generated="true" 
             data-ai-tools="GPT-4">
        <p>This entire section was generated by AI.</p>
    </section>

    <section data-ai-free="true">
        <p>This section was written entirely by a human, overriding the article-level enhancement tag.</p>
    </section>

    <section>
        <p>This paragraph inherits the article-level AI-enhanced status.</p>
        <p data-ai-generated="true" 
           data-ai-tools="GPT-3">
            But this specific paragraph was generated by AI.
        </p>
    </section>
</article>

In this example:

• The entire article is marked as AI-enhanced.
• The first section overrides this and is marked as fully AI-generated.
• The second section overrides the article-level tag and is marked as AI-free.
• The third section contains a mix:
  • The first paragraph inherits the article-level AI-enhanced status.
  • The second paragraph overrides this and is marked as AI-generated.

Best Practices for Nesting:

1. Use nesting judiciously. Only override parent tags when there's a significant difference in AI involvement.
2. Be as specific as possible. Tag at the most granular level that makes sense for your content structure.
3. Remember that all attributes can be overridden, including data-ai-tools and data-ai-tagged-date.
4. When overriding a parent tag, explicitly set all relevant attributes to ensure clarity.

6. Best Practices

1. Always use all four primary attributes (generated, enhanced, free, and tools) for any content using this standard. This provides a clear and complete picture of AI involvement and serves as a sanity check.
2. If data-ai-free is true, both data-ai-generated and data-ai-enhanced should be false.
3. Be as specific as possible when listing AI tools in data-ai-tools.
4. Apply attributes to the highest-level container for a piece of content. For example, tag an entire article, video, or image, rather than individual paragraphs or segments.
5. When in doubt about the level of AI involvement, err on the side of disclosure.
6. Include the data-ai-tagged-date attribute whenever possible to provide context about when the AI involvement was assessed.

7. Standardized Names for LLM Models

To ensure consistency, we recommend using the following standardized names for common LLM models in the data-ai-tools attribute:

• GPT-3
• GPT-4
• BERT
• T5
• DALL-E
• Midjourney
• Stable Diffusion

For models not listed here, use the most commonly recognized name for the model or tool. If you would like your AI model or tool to be included in this list, please contact us at .

8. Edge Cases and Complex Scenarios

8.1 Minimal AI Involvement

For content with minimal AI involvement (e.g., spell-checking), use discretion. If the AI's impact on the content is negligible, you may set data-ai-enhanced to "false", but should still list the tool in data-ai-tools.

8.2 Multiple AI Tools

When multiple AI tools are used, list all known tools in the data-ai-tools attribute:

<article data-ai-generated="false"
         data-ai-enhanced="true"
         data-ai-free="false"
         data-ai-tools="Grammarly,Hemingway App,GPT-3"
         data-ai-tagged-date="2024-08-26">
    This article was edited using multiple AI tools.
</article>

8.3 Dynamically Generated Content

For content that is dynamically generated or enhanced by AI in real-time, use JavaScript to add or update the attributes as needed, including the data-ai-tagged-date.

9. Future Considerations

1. This standard may evolve to include more granular indicators of AI involvement as AI technologies advance.
2. Future versions may introduce attributes for indicating the percentage of AI contribution.
3. Integration with other metadata standards and schemas may be considered in future iterations.

10. Versioning

This document represents version 1.1 of the AI Content Metatag Standard. Future updates will be numbered incrementally (e.g., 1.2, 1.3) for minor changes, and with new major version numbers (e.g., 2.0) for significant changes.

11. Adoption and Implementation

We encourage web developers, content creators, and platform providers to adopt this standard. By implementing these metatags, you contribute to a more transparent web ecosystem and help users, search engines, and AI systems make informed decisions about the content they consume or process.

For questions, suggestions, or to report issues, please contact us at .