When someone asks ChatGPT which foundation to buy, they probably assume the answer is coming from somewhere authoritative. Maybe a beauty expert, a clinical study, or at the very least a reputable publication they recognise. The reality is a little more complicated, and for beauty brands, it matters enormously.
AI tools are now a primary discovery channel for product recommendations. The customer journey that once ran through Google, then to Vogue, then to a product page is increasingly shortcutting straight to the AI answer. Which means if you want to show up in that answer, you need to understand what AI is using to find its sources.
We ran our own research to find out, and what we found had some surprises.
What our own research showed
We ran 20 beauty prompts through four major AI platforms: ChatGPT, Gemini, Perplexity, and Claude. Prompts ranged from “What’s the best foundation for dry skin?” to “Best vegan makeup brands available in the UK”. The kinds of questions real shoppers are genuinely asking AI platforms.
Across three rounds of testing, we recorded 683 citations, we categorised each source by type, and looked at which domains came up most often.
The headline finding: editorial coverage matters the most. Across both rounds of testing, 42% of all citations came from what we classified as editorial or glossy publications like Vogue, Allure, Cosmopolitan, Who What Wear, Marie Claire, Byrdie, and Glamour, and these were the publications which all appeared consistently. A further 31% came from SEO blogs and general web content.
Reddit ranked as the single most-cited individual source across both rounds, appearing 39 times in total. That’s more than Allure (24) and Vogue (23) combined. YouTube, Cosmopolitan, and Who What Wear also featured heavily.
Brand-owned content, like your own product pages, brand blogs, brand directories, accounted for just 6% of citations across the dataset.
How each AI platform sources differently
Not all AI tools behave the same way, and the differences are meaningful.
Gemini was the most citation-heavy platform in our research, averaging around 10.8 sources per prompt response. It leaned heavily toward editorial content (52.8% of its citations) with Reddit and TikTok both featuring prominently. It appears to weight prestige publications highly when forming beauty recommendations.
ChatGPT showed a similar editorial preference (42.7%), with Reddit appearing 16 times as its most-cited individual source. It also surfaced YouTube frequently and drew on a broad range of glossy titles.
Perplexity was notably different. It showed the most balanced spread across source types, with editorial (30.5%) and SEO/blog content (31.1%) roughly equal. Retailers like Boots appeared far more often here than with any other platform, 29 of the 80 retailer citations in our dataset came from Perplexity alone. It also cited brand-owned content more than any other platform (9.6%). This suggests Perplexity is more willing to treat commercial sources as credible.
Claude behaved very differently from the other three. It cited fewer sources overall, averaging 5.6 per prompt compared to Gemini’s 10.8, but showed a distinct preference for SEO and blog content (58.4% of its citations). Crucially, it returned zero social or forum citations across the entire dataset. It also cited content it described as coming from its own training data on five occasions, which no other platform did. Claude appears to be the most conservative and least editorially driven of the four.
What the wider research says
Our findings sit neatly alongside what other studies have been showing.
Muck Rack’s “What Is AI Reading?” report, which has now analysed over 25 million AI citations across ChatGPT, Claude, and Gemini, found that earned media accounts for 84% of all AI citations — and that this figure has remained consistent across three editions of the study going back to July 2025, ranging from 82% to 89%. Paid and advertorial content accounts for just 0.3%.
In beauty specifically, an analysis by Goodie covering 135,419 beauty and personal care citations across ChatGPT, Gemini, Claude, and Perplexity found that editorial publications like Byrdie, Allure, Vogue, and Cosmopolitan provide the expert-vetted recommendations and award validation that AI leans on most heavily, and that their “Best of” lists effectively become the scaffolding for AI recommendations. Reddit ranked first overall, present across all four LLMs and pulling from both posts and threads.
The brands that appear in AI answers are not getting there through their own content. They are getting there because the sources AI trusts have written about them.
What this means for beauty brands
If a beauty brand is not appearing in Vogue, Allure, Byrdie, or Cosmopolitan or being discussed on Reddit, it is largely invisible to the AI tools that an increasing number of shoppers are turning to for recommendations.
Brand-owned content like your product pages, your blog, your press releases are not what AI is reading. Across our research, they accounted for just 6% of citations. That is not zero, but it is not significant, it should be a gauge to show where to recalibrate where marketing effort is directed.
Research from 2025 suggests that content structure matters too, with AI models extracting 44% of citations from the first 30% of a page and rewarding content that leads with a direct, clear answer rather than building to it. You can read more about this here.
The Muck Rack data adds a PR-specific angle worth noting: the first seven days after publication generate the highest citation rates in both ChatGPT and Claude, and half of all citations across models come from content published in the last 11 months. Recency is a live variable. Past coverage fades. A placement from 2022 is not doing the same work it once did.
For digital PR services, this is a genuine opportunity. The brands consistently appearing in AI beauty answers are not the ones spending the most on paid media. They are the ones with sustained editorial presence in the publications that AI actually reads.
Where to focus your marketing efforts
Our research and the wider evidence point toward a consistent set of priorities for beauty brands that want to appear in AI-generated recommendations.
Editorial coverage in tier-one beauty titles is a non-negotiable.
Vogue, Allure, Byrdie, Cosmopolitan, and Marie Claire dominate citation lists across every platform we tested and every study we reviewed. Getting into those publications, consistently and repeatedly, is the single highest-leverage activity for AI visibility.
Community presence on Reddit
This matters more than most brands realise. It was the top-cited source in our dataset and appears in the top tier of every major platform. Organic brand presence in relevant subreddits, through honest reviews, real conversations, and community engagement, feeds directly into AI answers.
Retailer presence is platform specific
Boots appeared prominently in our Perplexity data specifically, suggesting that brands with strong retail visibility on Boots.com may have an advantage on that platform. For makeup brands with a broad retail footprint, this is worth monitoring.
Be on YouTube
Video content is being cited more than most beauty marketers expect. It appeared 15 times in our dataset, more than most individual editorial titles, suggesting that long-form video reviews are being indexed and referenced by AI in ways that shorter social content is not.
Next steps
The brands that understand this shift are already adapting and winning. They are investing in editorial relationships, building community presence, and thinking about digital PR not just as a brand-building tool but as an AI visibility strategy. The ones still optimising primarily for traditional search rankings are playing a game the rules of which are changing rapidly.
What we’ve found in our research is that AI beauty recommendations are not neutral. They are a reflection of what the media has said, what communities have been discussing, and what editorial gatekeepers have endorsed.
The question for beauty marketers is not whether AI will shape their customers’ decisions, we know and have proof that it already is. The question should be whether your brand is in the room when those recommendations are being made. We can help you make that happen, contact us to today to discuss your AI visibility strategy, we’d love to hear from you.
