The December 2025 core update was brutal for AI-generated content. Mass-produced pages without human oversight saw traffic losses of 60-95%. Generic "SEO content" optimized for keywords rather than users dropped 63%.
But some sites using AI didn't get hit at all. The difference wasn't whether they used AI. It was how.
What Google actually targeted
Google didn't penalize AI content for being AI content. They targeted specific quality patterns that happen to be common in AI output:
Scaled production without review
Sites publishing hundreds of AI-generated pages with minimal human oversight got hit hardest. The pattern: generate, lightly edit for obvious errors, publish, repeat.
Google got better at detecting this at scale. The telltale signs:
- Similar structure across many pages
- Generic information available everywhere
- No unique insights or original data
- Missing author expertise signals
Content that answers the prompt, not the user
AI is trained to respond to prompts. When the prompt is "write a 1500-word article about X," the AI writes 1500 words about X, whether or not that serves the user.
This creates content that:
- Pads word count with unnecessary filler
- Covers obvious points without adding value
- Lacks the specific details users actually need
- Reads like it was written to hit SEO metrics
Missing experience markers
AI can describe things. It can't experience them. The December update got better at detecting content that lacks genuine experience:
- No first-hand usage or testing
- No original photos or screenshots
- No personal opinions or lessons learned
- No specific examples from real situations
The 87% that got hit
Let's break down what happened to different types of AI content:
Fully automated content farms
These sites used AI to generate thousands of pages across broad topics with no human involvement beyond publishing.
Result: 90-95% traffic loss. Many were effectively deindexed.
"Edited" AI content at scale
These sites generated AI drafts, had humans do light editing (fix grammar, check facts), and published.
Result: 70-85% traffic loss. The editing wasn't adding enough value.
AI-assisted affiliate content
Product review sites using AI to write reviews without actually testing products.
Result: 71% saw significant drops. Google increasingly requires evidence of actual product experience.
Generic how-to content
AI-generated tutorials and guides covering topics already well-covered elsewhere.
Result: 63% ranking losses. No unique value over existing content.
The 13% that survived
Some AI-assisted content came through fine. What did they do differently?
AI for research and drafting only
Use AI to:
- Gather and organize information
- Create initial outlines
- Draft sections for human expansion
Then have humans add:
- Original insights and opinions
- Specific examples from experience
- Expert analysis and recommendations
- Unique data or research
Human expertise as the foundation
Start with human expertise, use AI to help express it:
- Subject matter expert provides core insights
- AI helps structure and articulate
- Expert reviews and adds specificity
- Final product reflects genuine expertise
Quality over quantity
Instead of 100 AI-generated articles:
- 10 thoroughly researched pieces
- Each with original data or testing
- Deep coverage rather than surface-level
- Regular updates to stay current
How to use AI responsibly going forward
AI isn't going away, and Google isn't against AI content. They're against low-quality content that happens to be efficiently produced by AI.
The test that matters
Before publishing AI-assisted content, ask:
"Would an expert in this field be proud to put their name on this?"
If no, if it's generic, obvious, or lacks real insight, it's not ready.
Practical guidelines
Do use AI to:
- Speed up research and information gathering
- Create outlines and first drafts
- Help articulate complex ideas
- Check grammar and improve clarity
Don't use AI to:
- Replace subject matter expertise
- Generate content at scale without review
- Cover topics you don't actually understand
- Create "SEO content" for ranking purposes
Adding the human layer
For every AI-assisted piece, add:
- Original examples: Real situations you've encountered
- Specific data: Numbers, dates, results from your experience
- Expert opinion: Your actual take, not just "experts say"
- Visual evidence: Screenshots, photos, videos you created
- Unique insights: Things the AI couldn't know
The opportunity
Sites that relied on AI shortcuts are now struggling. That's an opportunity for sites willing to invest in genuine quality.
The bar has risen. Meet it by using AI as a tool to enhance human expertise, not replace it.