The Study
Spoiler: it's not your meta descriptions.
Our data team spent three months tracking which content sources AI engines actually reference when making product recommendations. We analyzed over 10,000 AI-generated product recommendations across 50 categories.
The results challenge everything we thought we knew about digital marketing.
Methodology
We queried ChatGPT, Claude, Gemini, and Perplexity with standardized product recommendation questions. Each query was structured as a real consumer would ask:
- "What's the best [product] for [use case]?"
- "Should I buy [Brand A] or [Brand B]?"
- "[Product category] recommendations for [persona]"
We then analyzed the responses, tracking which sources were cited (explicitly or implicitly) and how recommendations were structured.
Key Findings
1. Reviews Trump Everything
Third-party reviews account for 47% of all recommendation influence. Not brand websites. Not paid content. Independent reviews.
The most-cited sources:
- Wirecutter (12% of citations)
- Reddit (11%)
- Specialized review blogs (9%)
- Consumer Reports (8%)
- YouTube reviews (7%)
2. Your Website Matters Less Than You Think
Brand websites were cited in only 18% of recommendations. And when they were cited, it was almost always for:
- Technical specifications
- Pricing information
- Availability
Marketing copy? Almost never referenced.
3. Community Sentiment Is Gold
Reddit threads and forum discussions appeared in 23% of recommendations. AI engines seem to weight "real user experiences" heavily.
Particularly influential:
- Detailed comparison posts
- Long-term ownership reviews
- Problem/solution discussions
4. Recency Varies Wildly
Training data cutoffs mean some AI engines reference content from 2+ years ago. Others (like Perplexity) access real-time data.
This creates inconsistency—your brand might be recommended by one AI and completely ignored by another.
What Gets Cited vs. What Gets Ignored
High citation rate:
- Comparison articles ("X vs Y")
- Best-of lists from trusted publications
- Detailed Reddit reviews
- Technical specification pages
Low citation rate:
- Press releases
- Sponsored content
- Social media posts
- Most blog content without substantive information
Implications
If you're optimizing for AI visibility, traditional SEO tactics fall short. The content that influences AI recommendations is:
- Third-party validated
- Comparison-focused
- Community-generated
- Technically detailed
Your marketing strategy needs to extend beyond your owned channels into the sources AI actually trusts.
What's Next
We're continuing this research with a focus on how recommendation patterns change over time. Subscribe to get updates as we publish new findings.