Wills: ChatGPT Ranking Factor Study

Key Takeaways:

• Wills analyzed citations in ChatGPT of 100k+ prompts including 145 industries and 1.595 personas to draw conclusions on Citation factors:

  • 1. SEO is the foundation. OpenAI is using search data today and building their own index. As that matures, the connection between search authority and LLM visibility deepens. Traditional SEO principles are not obsolete, they're the starting point for LLM visibility too.

  • 2. Persona is the measurement unit. The #1 airline for a frequent flyer is a different site from the #1 for a student flying abroad. Same model, same industry, different person, different result. You don't have one LLM rank, you have a rank per buyer segment. Monitor by persona or the number is meaningless.

• I scraped the data and took a look at comparisons...

  • across personas for each industry: there are clear patterns for each industry which domains dominate and for which personas the general visibility order does not apply

  • across industries for each factor: the spearman correlation of a factor in comparison to other factors was quite similar across industries

• Key Measures:

  • monitor how well LLMs and retrieval systems cite or mention you in each customer segment e.g. by using personas in prompt tracking

  • monitor changes over time and optimize content and features for specific personas or rather customer segments/user groups

  • analyze citation patterns and focus on information gain

Net diagram showing the top N domains of an industry across all tested personas: only few domains are similarly visible and mentioned across personas

Spearman correlation of selected factors across industries in the insurance sectors - facors correlation should be similar across similar industries, but it is not

Sources:

• https://x.com/benwills/status/2054227566303846574

• https://oppalerts.com/LLM-Ranking-Factors/