Does AI penalize thin content differently than Google? We tested it with some surprising results

Great test. Let me add context on why AI treats thin content differently.

Google vs AI evaluation:

Why AI “ignores” thin content:

AI systems ask: “Does this content add something I can cite?”

Thin content typically:

• Restates common knowledge
• Lacks specific data
• Doesn’t answer directly
• Has no quotable insights

AI’s implicit quality test:

“If I cite this, does it add credibility to my response?”

Generic content fails this test. Original insights pass.

The implication:

Thin content isn’t “penalized” - it’s simply not useful for AI’s purpose. Different from Google, where thin content might rank with enough links.

We’ve audited hundreds of sites for AI visibility. Here’s what thin content looks like in practice.

AI-invisible content patterns:

1. “What is X” without differentiation

   • Generic definitions
   • No unique perspective
   • Could be on any site

2. Listicles without depth

   • “10 tips for X”
   • Each tip is one sentence
   • No explanation or evidence

3. Aggregated information

   • Compiles existing sources
   • No original analysis
   • Nothing AI couldn’t find elsewhere

4. Keyword-stuffed content

   • Written for Google, not users
   • Repetitive phrases
   • No genuine value

AI-visible content patterns:

1. Original research with data
2. Expert perspectives with credentials
3. Case studies with specific outcomes
4. Unique frameworks or methodologies
5. First-hand experience reports

The audit question:

“If this content disappeared, would anything unique be lost?”

If no, it’s thin for AI purposes.

Defending shorter content:

Short ≠ Thin

Some of the most AI-cited content is SHORT:

• FAQ answers (50-100 words)
• Data points with source
• Definition with unique insight
• Quick how-to with specific steps

What makes short content valuable:

1. Answers directly - No fluff, just the answer
2. Citable format - Easy to extract and quote
3. Unique information - Something not found elsewhere
4. Clear expertise - Author credibility visible

Example structure:

[Direct answer - 50 words]
[Supporting data - 100 words]
[Expert context - 100 words]
[Practical application - 100 words]

Total: 350 words AI citation rate: Often higher than 2,000-word guides

The insight:

AI rewards density over length. Be concise but substantive.

How to upgrade thin content for AI:

The thin content audit:

For each piece, ask:

1. Does it contain original information?
2. Does it directly answer questions?
3. Would an expert find it valuable?
4. Is there something quotable?

If 0-1 yes: Thin (needs major work) If 2 yes: Medium (needs enhancement) If 3-4 yes: Substantial (optimize format)

Upgrading thin to substantial:

Our process:

1. Audit all content for “thickness”
2. Prioritize pages with ranking potential
3. Add original value (not just words)
4. Monitor AI citation changes

Typical results:

Upgraded thin content: 200-400% increase in AI citations. Just adding length: Minimal change.

E-commerce thin content reality:

The problem:

Most product pages are “thin” by AI standards:

• Generic product descriptions
• Copied manufacturer specs
• No unique value

Our solution:

Transform product pages into information resources:

Before (thin): “High-quality running shoe. Comfortable fit. Great for runners.”

After (substantial): “Weighs 9.2oz with 32mm stack height. Tested by 47 marathon runners in our panel, rated 4.7/5 for cushioning on runs over 10 miles. Best for: neutral gait, road running, distances 10K+.”

The additions:

• Specific measurements
• Original testing data
• Use case specificity
• Expert recommendations

Results:

Product pages with unique data: 28% AI citation rate for product queries Generic product pages: 3% AI citation rate

The investment:

Yes, it’s more work. But generic product content is invisible to AI.

Operationalizing content thickness:

Our content scoring system:

Each piece gets scored on:

Scoring guide:

• 0-2 points: Thin (don’t publish or upgrade)
• 3-5 points: Medium (acceptable for supporting content)
• 6-10 points: Substantial (pillar content potential)

Our rule:

No content published below 3 points. All pillar content must score 7+.

The result:

Average AI citation rate went from 8% to 31% after implementing scoring.

Tool:

We use a simple checklist in our CMS. Writers self-score, editors verify.

Great discussion. Here’s my framework from these insights:

AI Content Thickness Framework:

The core principle:

AI doesn’t penalize thin content - it ignores content without citation value. The question isn’t “is this long enough?” but “is this worth citing?”

Thickness factors (prioritized):

1. Original research/data (highest value)
2. Expert perspectives
3. Specific, verifiable facts
4. Unique insights/frameworks
5. Direct answer format
6. Depth of coverage (not length)

The thickness test:

Ask these questions:

• Does this contain information unique to us?
• Would an expert find this valuable?
• Is there something specifically quotable?
• Does this directly answer user questions?

Score 3-4: Substantial content Score 1-2: Medium (enhance before publishing) Score 0: Thin (don’t publish for AI visibility)

Practical upgrades:

The bottom line:

Word count ≠ thickness Unique value = thickness

Focus on what makes your content irreplaceable.

Thanks everyone for the excellent insights!