| Research overview | The conclusions and the key data, shortened |
|---|---|
| Blog post with the full details | LLM Ranking Factors Report Data (May 2026) |
| Live report | LLM Ranking Factors report |
| Also posted | LinkedIn post, May 11, 2026 |
What the research found
In May 2026, I compared 13 signals with the domains ChatGPT recommended across 145 industries. Search result signals had the strongest correlations, followed by backlink measures. But no single signal explained more than 5.8% of the variation. The strongest signal changed by industry, and the recommendations changed with the buyer persona. In July, I repeated the work across 13 models from five providers, using about ten times the data.
Findings you can quote
- Search engine appearances (0.241), best search engine rank (0.238) and search engine outbound links (0.230) were the three strongest signals.
- Backlink count (0.204), backlink authority (0.200) and PageRank (0.194) came next.
- Reddit comments correlated at 0.111, Reddit posts at 0.096, Wikipedia citations at 0.077 and homepage keywords at 0.072.
- Even taken together, the external signals I measured explain well under 20% of the variation. Roughly 80 to 85% happens inside the model.
- Wikidata was the strongest signal in hotels (0.650), ERP software (0.655) and furniture stores (0.706). Reddit posts were strongest in enterprise AI platforms (0.528) and live entertainment and ticketing (0.503).
- In airlines, backlink authority explains about three times as much of the variation as search appearances: 24.5% against 8.3%.
- Five airline personas produced five different top recommendations.
Quotes from Ben Wills
These quotes are from the report and the LinkedIn post.
“Same model, same industry, different person, different result.”
“If you are measuring LLM visibility without specifying who is asking, you are measuring noise.”
“Search is not dead. It is becoming one layer of AI recommendations.”
Story angles by audience
AI search visibility
Every signal had a small effect on its own. What the report gives you is a way to see which signals move with recommendations in your industry. None of the signals measured here explains the recommendations on its own.
SEO
The three strongest signals came from search results, with backlink measures next. Homepage keywords had the weakest correlation in May. When I re-scored homepage relevance for the July report, it rose to the second strongest.
Reputation monitoring
Reddit had the strongest correlation in categories where people use it to evaluate their options. For Reddit posts, that included enterprise AI platforms (0.528) and live entertainment and ticketing (0.503). Reddit comments also had a correlation of 0.321 for cosmetic-surgery clinics. My reading is that when a brand dominates real community discussion, that conversation can become part of how the model recommends it.
Competitive analysis
You can look up your own industry, compare your brand with the domains ChatGPT recommends, and see the scores and signals for each domain, broken out by buyer persona. The leaders can differ depending on what you measure. In hotels, tripadvisor.com led search appearances (223), while marriott.com led ChatGPT recommendations.
What to keep in mind
- A correlation tells you what moves together. It does not establish why the model makes a recommendation.
- I used one model, ChatGPT 5.4, for this research. The July 2026 report used 13 models from five providers and re-scored two signals. Keep those differences in mind when comparing the reports.
- I released the report on May 11, 2026. It scored 29,562 unique domains that ChatGPT recommended.
- The analysis used 1.15 billion Common Crawl pages, 4 billion+ backlinks, 5 billion+ Reddit submissions and comments, 15,697 Google searches and 300 million+ Wikimedia records.
- You can read the full report for free at oppalerts.com/LLM-Ranking-Factors.
Interviews
For an interview with Ben Wills, get in touch. You can find his background and the other research in the media kit.