Data from Novi Show What Sources ChatGPT Uses to Recommend Beauty Products

New analysis from Novi, the platform that helps brands and retailers ensure AI assistants like ChatGPT and Gemini can find and recommend their products in queries, shares the top sources ChatGPT references. 

Reddit, Who What Wear, Wikipedia, and Sephora are the top sources ChatGPT references most frequently when responding to beauty product recommendation prompts.

The Novi analysis examined 10.7 million citations across over 98,000 source websites that ChatGPT considered when generating responses to user queries to identify the sources large language models (LLMs) most frequently pull from when generating beauty product recommendations. 

Top Overall Beauty Sources

1. Reddit
2. Who What Wear
3. Wikipedia
4. Sephora.com
5. Allure

Kimberly Shenk, CEO of Novi, said,

“AI models weigh both SKU-level data and trust signals when recommending what products consumers should buy. Our research shows products with verified trust signals, like certifications and badges, are presented to shoppers significantly more often than products without them. For beauty brands, that means the most effective path to AI visibility is pairing well-structured data with verified trust signals, not chasing visibility via any single source.”

These findings come as shoppers increasingly make AI platforms their first stop for product discovery, with nearly three-quarters (73%) using AI search to learn about categories, products, brands or services. 

Over the most recent holiday shopping season, shoppers who were referred to retail sites from AI platforms converted 31% more often than others. However, nearly half of brands (47%) don’t know if or how they show up in the answers AI platforms generate in response to shoppers’ questions.

Skincare & Fragrance, AI Prioritizes Different Sources

The Novi analysis shows that AI platforms prioritize different combinations of sources depending on the beauty category covered by the consumer’s particular query. 

For skincare questions, the AI algorithms lean more heavily on editorial outlets and major retailers, while for fragrance queries, retailer websites and brand-owned content are cited more frequently.

Top 5 Skincare Sources

1. Reddit
2. Who What Wear
3. Sephora.com
4. Allure
5. Ulta.com

In the fragrance category, brands such as Neutrogena and La Roche-Posay rank in the top 10, indicating that AI engines prioritize brand-owned content alongside editorial and retailer sources when generating product recommendations in this category.

Top 5 Fragrance Sources

1. Reddit
2. Wikipedia
3. Who What Wear
4. Fragrantica
5. Sephora.com

Novi’s AI citation analysis shows that brands benefit from building consistent, structured product data across their digital footprint. 

Data also shows that certain social, editorial, and retail sites where that information is validated may matter far more than others. 

In AI product recommendations, those citations can influence how LLMs understand product relevance, evaluate credibility and determine which products to recommend.