How to Reverse-Engineer ChatGPT's Citation Algorithm
A technical analysis of the signals ChatGPT uses to decide which brands to recommend, based on our analysis of 10,000+ commercial queries.
Citerecon Labs
July 28, 2026
Methodology
We analyzed 10,847 commercial queries across 42 B2B software categories over a 90-day period. For each query, we recorded which brands ChatGPT cited, their rank order, the language used, and cross-referenced this with the cited brands' web presence characteristics.
Key Findings
Finding 1: Structured Data Correlation
Brands with comprehensive JSON-LD schema markup were 2.7x more likely to be cited than brands without it. The effect was strongest for Organization and SoftwareApplication schemas.
Finding 2: Third-Party Authority
The single strongest predictor of citation was the number of mentions in authoritative third-party sources:
- G2 reviews (strongest signal)
- Wikipedia presence
- Crunchbase profiles
- TechCrunch/major media mentions
- Stack Overflow discussions
Finding 3: Content Depth Threshold
There appears to be a content depth threshold around 2,500 words per topic below which ChatGPT is less likely to treat a source as authoritative. Brands with multiple deep-dive articles on their core topics were cited more consistently.
Finding 4: Recency Bias
ChatGPT shows a measurable recency bias. Content published within the last 90 days was 1.8x more likely to influence citations than older content, controlling for other factors.
Finding 5: The Comparison Page Effect
Brands that published their own comparison pages (e.g., "YourBrand vs Competitor") were cited 40% more often in comparative queries. This suggests that self-published comparison content directly influences the model's comparative reasoning.
Implications
These findings suggest a clear action plan for brands seeking to improve their ChatGPT citation rate:
- 1Invest in structured data - this is the highest-ROI technical intervention
- 2Secure third-party mentions - especially G2, Wikipedia, and major tech media
- 3Publish deep content - target 3,000+ words for core topic pages
- 4Maintain freshness - publish or update key content at least monthly
- 5Create comparison content - own the narrative before your competitor does
Ready to Audit Your AI Visibility?
Use Citerecon's Brand Scanner to instantly see how AI engines perceive your brand.
Scan Your Brand Free