Kan noai-meta-taggen faktiskt skydda mitt innehåll från AI-träning? Eller är det bara önsketänkande?

Let me explain what the noai tag actually does and its real limitations.

How it works:

You add this to your HTML head:

<meta name="robots" content="noai">

Or for images specifically:

<meta name="robots" content="noimageai">

The honest reality:

What it CAN do:

• Signal your intent to ethical AI crawlers
• Establish documented preferences
• Work with platforms that have adopted the standard
• Provide some protection from responsible companies

What it CANNOT do:

• Stop malicious scraping
• Protect content already collected
• Force compliance from all AI companies
• Guarantee your content isn’t used

The key limitation:

This is a voluntary standard. It’s like a “No Trespassing” sign - polite visitors will respect it, but it won’t stop determined trespassers.

Who respects it:

DeviantArt, ArtStation, Sketchfab, Fab - they’ve committed in their terms of service. But OpenAI, Google, Anthropic? No formal commitments to honor this specific tag.

Artist perspective here.

My experience:

Implemented noai and noimageai on my portfolio site 6 months ago. Here’s what happened:

The good:

• DeviantArt now shows my “no AI” preference
• ArtStation respects the setting
• Peace of mind (maybe false, but still)

The bad:

• Found my art in AI training discussions on Reddit
• No way to know if major models already have my work
• Still see AI-generated “inspired by” versions of my style

My conclusion:

The noai tag is like locking your front door - it won’t stop a determined thief, but it’s still worth doing.

What actually helps more:

• Watermarking strategically
• Posting lower resolution versions
• Keeping highest quality behind paywalls
• Documenting and timestamping original work

The uncomfortable truth:

If you’ve had public content online for years, it’s probably already in training datasets. The tag only helps going forward.

Legal context on the noai tag.

Why it matters legally:

Even though it’s voluntary, implementing it creates a record of your intent. In potential future litigation, being able to say “I explicitly opted out using industry-standard methods” strengthens your position.

Current legal landscape:

• No specific laws require respecting noai
• Copyright law still applies to AI training (ongoing cases)
• Some jurisdictions recognize opt-out mechanisms

The evolving situation:

Several countries are developing AI-specific regulations. Having documented opt-out preferences now may matter when these regulations take effect.

My recommendation:

Implement it, but don’t rely on it. Combine with:

• Clear terms of service
• Copyright statements
• Monitoring where your content appears
• Documentation of original creation dates

On retroactive protection:

You cannot remove content from existing training datasets. The tag only affects future collection. This is a significant limitation that many don’t understand.

Technical implementation details.

The variations:

Platform-specific implementation:

WordPress:

• Add to functions.php using wp_head hook
• Use plugins like “Simple NoAI and NoImageAI”
• Add to theme’s header.php

Squarespace:

• Settings > Advanced > Code Injection
• Requires Business/Commerce plan

Wix:

• Settings > Advanced > Custom Code
• Place in head section

The technical limitation:

This is a meta tag, not an authentication system. Any crawler can simply ignore it. It’s asking nicely, not enforcing.

What I wish existed:

A technical standard that actually prevents access, not just requests it. But that’s essentially impossible with public web content.

Research perspective on AI training data practices.

The reality of AI training:

Most major AI models were trained on datasets collected before noai tags existed. Common Crawl, WebText, and similar datasets contain billions of web pages without any opt-out filtering.

Current state of compliance:

Some AI companies check:

• GPTBot (OpenAI) respects robots.txt
• GoogleBot-Extended can be blocked
• Some smaller AI companies honor noai

Many don’t:

• Countless scrapers ignore all signals
• Historical data already collected
• Some companies explicitly disregard opt-outs

The platform perspective:

Art platforms like DeviantArt and ArtStation adopted noai because:

• User backlash over AI training concerns
• Legal pressure
• Competitive differentiation

They honor it within their ecosystems but can’t control external scrapers.

My take:

The noai tag is necessary but insufficient. It represents an important step toward giving creators control, but technical and legal enforcement mechanisms need to catch up.

Photographer’s practical approach.

What I do:

1. noai meta tag - Yes, implemented site-wide
2. robots.txt blocking - All known AI crawlers
3. Watermarks - On all portfolio images
4. Resolution limits - Full res only for clients
5. Monitoring - I check AI image generators for my style

The monitoring piece is key:

I use Am I Cited to track if my brand/site gets mentioned in AI responses. If AI systems are citing me as a source, I want to know about it - both for opportunity and protection reasons.

My philosophy:

I can’t prevent all AI training, but I can:

• Document my preferences clearly
• Track where my content appears
• Build evidence if needed later

The business angle:

Some photographers are pivoting to licensing specifically for AI training. If you can’t beat them, monetize them. Different approach, but worth considering.

Small business owner perspective.

My different take:

I actually WANT AI systems to cite my content. Visibility in AI responses is increasingly valuable.

The distinction:

• Training on content - AI learns from your content to generate new content (what noai blocks)
• Citing content - AI references your content in responses (potentially valuable)

Why I haven’t implemented noai:

1. I want AI systems to know about my business
2. Citations drive traffic and awareness
3. Being excluded might hurt visibility

The nuance:

noai is about training data, not citations. But being blocked from training might affect whether AI “knows” about you enough to cite you.

What I do instead:

• Monitor AI citations (Am I Cited)
• Optimize content for AI visibility
• Focus on being cited as an authority

Different goals require different strategies.

This discussion clarified a lot for me.

My takeaways:

1. noai is limited but worth implementing - Low effort, documents intent
2. It’s voluntary - Won’t stop determined scrapers
3. Retroactive protection doesn’t exist - Past content is past
4. Layer protections - Tag + robots.txt + ToS + monitoring
5. Different goals matter - Prevent training vs. ensure citation

What I’m doing:

For my art:

• Implementing noimageai
• Watermarking portfolio pieces
• Limiting public resolution
• Documenting creation dates

For my blog:

• Keeping noai OFF for articles
• I want text content to be cited
• Monitoring citations with Am I Cited

The key insight:

The noai tag is about training data, not visibility. For my art, I want protection. For my articles, I want visibility. Different content, different strategies.

Thanks everyone for the nuanced perspectives!