The Top Sites ChatGPT is Using for Beauty Recommendations

As AI-powered shopping continues to reshape product discovery, a new analysis from Novi reveals which online sources most influence the beauty recommendations generated by ChatGPT. The results highlight the growing importance of community discussions, editorial authority and retailer validation.

• Related: AI Search Is Rewriting Influencer Marketing. Beauty Brands Are Racing to Feed the Machines.

The study analyzed 10.7 million citations referenced by ChatGPT between January 22 and May 20, 2026, spanning more than 98,000 source websites. The goal was to identify which sources large language models (LLMs) rely on most frequently when generating beauty product recommendations.

Across beauty queries overall, Reddit emerged as the most-cited source by a substantial margin, followed by Who What Wear, Wikipedia, Sephora and Allure. The findings suggest that AI recommendation engines draw heavily from a combination of consumer-generated discussions, editorial content, encyclopedic reference material and retailer product information when determining which products to surface.

"The takeaway should not be to optimize for one specific website or even a small set of websites," says Novi CEO Kimberly Shenk. "AI systems pull from a broad mix of brand-owned, retailer, editorial, social and community sources, but authority does not weigh equally across all of them."

She adds, "Brands need to understand which sources matter most for their category, how those sources shape AI’s understanding of the competitive set, and what signals help certain products get selected over others."

Shenk concludes, "The shift is away from chasing keyword optimization, which will not meaningfully change how AI systems evaluate products, and toward building a digital footprint that helps AI understand where a product fits, when it is relevant, and why it should be recommended."

The research arrives as AI becomes an increasingly important discovery channel. According to Novi, 73% of shoppers now use AI search tools to learn about products, brands and categories, while consumers arriving at retail websites from AI platforms converted 31% more frequently than other shoppers during the most recent holiday season. Despite this shift, nearly half of brands (47%) remain uncertain about whether or how their products appear in AI-generated recommendations.

"Rather than simply auditing visibility, brands need to understand how platforms like ChatGPT and Gemini interpret their products across the categories that matter most to their business," says Shenk. "This means looking beyond whether a brand appears in a handful of prompts and understanding how LLMs actually think through product recommendations, from how they group products into categories to the signals they use to compare options and decide what to recommend. The immediate step is to evaluate whether their products are being placed in the right categories, associated with the right use cases, and understood as relevant options within the competitive set."

How the Analysis Was Conducted

Novi examined 10.7 million citations generated in response to beauty-related product recommendation prompts within ChatGPT. The dataset covered more than 98,000 individual websites, though the overwhelming majority were cited fewer than 10 times.

By analyzing citation frequency, Novi sought to determine which sources AI systems most often reference when constructing recommendations across beauty categories and within specific segments such as skin care and fragrance.

The methodology offers a window into how LLMs evaluate product relevance and credibility. Rather than relying on a single source, AI systems synthesize information from multiple channels, including consumer discussions, editorial reviews, retailer product pages and brand-owned content.

In this context, category specifics matter significantly.

"Brands need to invest in analytics that show how LLMs are actually making product selections within their competitive category," says Shenk. "That does not mean tracking a limited set of prompts where a brand wants to show up, because consumers do not research in predictable ways."

She continues, "Instead, brands need to understand which products are being selected, what attributes are influencing those selections, and which sources carry authority for that specific category. From there, they can adapt their strategy to strengthen the right product attributes, claims and trust signals across the sources AI systems are already using."

Finally, she says, "Once brands understand where and how LLMs are making these decisions, they can more intentionally target the publications, retailer pages"

Skin Care Queries Favor Editorial and Retail Validation

For skin care-related questions, ChatGPT's citations leaned more heavily toward editorial publications and major beauty retailers.

The top five most-cited sources for skin care recommendations were:

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

Notably, brand-owned websites also appeared among the top-ranked sources. Companies including Neutrogena and La Roche-Posay placed within the top 10, indicating that well-developed brand content remains influential when AI systems evaluate skin care products.

Fragrance Recommendations Draw More Heavily from Category-Specific Sources

Fragrance-related queries produced a different citation profile.

The top five sources cited for fragrance recommendations were:

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

The appearance of Fragrantica among the leading sources highlights the value of category-specific databases and enthusiast communities in fragrance discovery. The findings also suggest that AI models may place greater emphasis on specialized fragrance knowledge and brand information when generating scent recommendations.

Trust Signals Matter More Than Individual Platforms

While Reddit's dominance may attract attention, Novi argues that the broader takeaway is not that brands should focus exclusively on any single website.

AI systems evaluate both structured product information and signals that indicate credibility and trustworthiness. The company's research found that products carrying verified certifications, badges and other trust indicators appear more frequently in AI-generated recommendations than products lacking those signals.

Shenk explains, "AI systems give meaningful weight to claims and proof points that help them understand why a product is relevant for a specific category or consumer need. The priority for brands is to make sure those claims are clearly surfaced, readable and substantiated across the places AI systems are likely to reference. Certifications are one way to do this, but so are clear ingredient explanations, testing details, clinical support, retailer badges, standards, reviews and other forms of validation."

She cautions, however, "The takeaway is not that every brand needs to chase more certifications. It is that brands need to make the evidence behind their product claims easy for AI systems to find, understand and trust."

For beauty marketers, the findings reinforce the importance of maintaining accurate and consistent product information across retailer sites, editorial coverage, brand-owned channels and third-party platforms. As AI assistants increasingly influence purchase decisions, visibility may depend less on traditional SEO tactics and more on ensuring that trustworthy, structured product data exists across the digital ecosystem.

"They [brands] should make sure product information is clear, structured, easy to verify and machine-readable," says Shenk. "Product pages should include detailed ingredient information, use cases, benefits, testing data, certifications and other attributes that are important to that specific item and category."

The executive adds, "Marketing copy alone is not enough. The information needs to be presented in a way that consumers can easily understand, but also in a format that search engines and AI systems can easily interpret, connect back to the product and use when evaluating recommendations."

The study also highlights a broader shift in how beauty brands should think about discoverability. Rather than optimizing solely for search engines, marketers may need to optimize for AI systems that synthesize information from thousands of sources simultaneously and weigh credibility signals when deciding which products to recommend.