By The Darl Team • Last Updated: June 2026 • 8 min read
| ⚡ChatGPT recommends products it has encountered in structured, authoritative sources: trusted editorial coverage, brand websites with clear factual descriptions, and review platforms. Brands that appear across multiple credible sources with consistent claims are far more likely to be cited than those relying on SEO alone. |
Sofia ran the same query fifty times. The pattern that emerged surprised everyone on her team.
As part of a research project at The Darl, Sofia and our team set out to answer a specific question for our beauty and skincare clients: when someone asks ChatGPT to recommend a product, what determines which brands actually get named? Not in theory; in fifty real queries, across real product categories, with real outputs we could analyse line by line.
The brands that showed up again and again were not always the biggest. They weren’t always the ones with the strongest SEO. Some genuinely excellent, well-funded brands never appeared once across fifty queries in their own category.
This wasn’t a small sample by accident. Across the 50 product recommendation queries we ran and analysed, a clear pattern emerged in what separated the brands that got cited from the brands that didn’t. This article walks through exactly what we found, why beauty and skincare brands are unusually exposed to this dynamic, and the four-step process we now run with clients to fix it.
How ChatGPT Builds Its Product Knowledge
ChatGPT doesn’t maintain a live product catalogue. Its knowledge of specific brands and products comes from two places: patterns absorbed during training from the broader web, and — when browsing is active — content it can access and verify in real time. Neither of these works the way a retailer’s search engine does.
Think of it less like a store directory and more like a well-read friend who’s seen a lot of articles, reviews, and forum threads about your category, but has never personally used any of the products.
That friend will recommend whatever left the clearest, most repeated impression; not necessarily whatever is objectively best. Clarity and repetition beat quality of product alone, every time, in how the recommendation actually gets generated.
This is the foundational reason most brands never appear: a great product with no clear, repeated, structured presence across credible sources leaves no impression for the model to draw on. The product could be the best in its category. If it never appeared anywhere ChatGPT can point to with confidence, it functionally doesn’t exist for recommendation purposes.
Read more: What is GEO?
The Sources ChatGPT Trusts Most
Across our 50-query analysis, we traced every product mention back to the type of source it most plausibly originated from. The pattern was consistent enough to rank.
| Source Type | % of Citations Traced To It | What It Signals to ChatGPT |
|---|---|---|
| Editorial / press coverage | 34% | Independent validation; a third party vouching for the product, not the brand vouching for itself |
| Brand’s own website | 27% | Clear, factual descriptions ChatGPT can extract; but weighted lower than third-party sources alone |
| Review & retail platforms | 23% | Aggregated consumer signal and structured comparison data |
| Forums & community discussion | 11% | Repeated, organic mentions across real conversations |
| Other / unclear origin | 5% | Mixed or untraceable signal |
The headline finding: editorial and press coverage outweighed every other source category — including the brand’s own website. This single data point reframes where most beauty and skincare marketing budgets should be going if AI visibility is a goal.
PR and credible third-party coverage isn’t a brand-awareness nice-to-have anymore. It’s now directly upstream of whether ChatGPT can recommend you at all.
Why Beauty Brands Are Uniquely Vulnerable
Beauty and skincare came up disproportionately often as a category where recommendation gaps were widest in our research; for three structural reasons specific to the industry.
Category Saturation
Beauty has more brands competing for the same descriptive language — “hydrating,” “brightening,” “clean” — than almost any other consumer category. When every brand uses near-identical language to describe near-identical benefit claims, ChatGPT has fewer distinguishing signals to work with, and tends to default toward whichever brand has the clearest third-party validation breaking the tie.
Ingredient-Led Queries
A large share of beauty product queries are ingredient-first — “best retinol serum,” “vitamin C for dark spots” — rather than brand-first. This means ChatGPT is frequently selecting from an open field of any brand using that ingredient, rather than a query that already names you.
Winning ingredient-first queries requires being the clearest, most credible voice on that specific ingredient; not just having a good product that contains it.
Influencer Noise Without Structure
Beauty has enormous influencer and UGC volume, but influencer content is rarely structured in a way ChatGPT can extract cleanly. A hundred TikTok mentions with no consistent factual claim behind them carry less weight in AI recommendation than a single clear, structured editorial review.
Volume of mentions and structure of mentions are not the same thing, and beauty brands often have abundant volume with very little structure.
What Our Research Revealed About AI Product Picks
The figures below come from The Darl’s internal research: 50 ChatGPT product recommendation queries run across skincare, wellness, and beauty categories, analysed for citation patterns.
| 58%of queries returned the same 3–4 brands per category, regardless of phrasing | 71%of cited brands had at least one piece of third-party editorial coverage | 22%of cited brands had inconsistent product claims across the sources we checked |
The most actionable finding: 58% of category queries returned the same handful of brands no matter how the question was phrased.
This tells us ChatGPT’s product recommendations are not highly sensitive to query wording; they’re sensitive to which brands have built a strong enough cross-source presence to dominate the category regardless of phrasing. That’s a structural position to compete for, not a keyword to chase.
The most surprising finding: 22% of cited brands had inconsistent claims across the sources feeding their citation; different ingredient lists, different positioning, different claims about results. ChatGPT cited them anyway, but the inconsistency is a vulnerability.
As AI systems get better at cross-referencing for accuracy, the brands with airtight consistency across sources are the ones positioned to hold their citation share.
| The pattern across all 50 queries: brands didn’t get recommended because they had the best product. They got recommended because they had the clearest, most repeated, most independently validated story about their product across the open web. |
The 4-Step Process to Get Recommended
This is the process we run with skincare and beauty clients directly off the back of this research.
1. Run your own category citation test: Ask ChatGPT to recommend products in your category, phrased five different ways. Document which brands appear consistently. If you’re not one of them, you’re not yet competing in the format that matters. 2. Audit your third-party coverage gap: List every piece of independent editorial, review, or press coverage your brand currently has. Compare it against the brands that did appear in your citation test. The gap here is usually the single biggest lever. 2. Fix claim consistency across every source” Pull your product descriptions from your website, retail listings, and any press coverage. Align ingredient claims, benefit claims, and positioning so every source tells the same factual story. 4. Pitch ingredient-first and comparison editorial: Target coverage built around the specific ingredients and use-cases your buyers are actually asking ChatGPT about; not just brand-name press. This is what wins the open-field, ingredient-first queries that dominate beauty search. |
A Skincare Brand We Made ChatGPT-Visible
The example below reflects a specific client engagement at The Darl and is not a guarantee of identical results for every brand.
A mid-sized skincare brand with a genuinely strong vitamin C serum came to us after running their own version of the citation test in step one above and finding themselves absent from every variation of “best vitamin C serum” they tried. Their product reviews were strong. Their SEO was solid. Their AI visibility was zero.
We ran the audit from step two and found the gap immediately: they had almost no independent editorial coverage. Every mention of their product online traced back to their own website or their own paid placements; both lower-weighted source types in our research.
We rewrote their product claims for consistency across every channel, then spent eight weeks securing four pieces of ingredient-focused editorial coverage specifically about vitamin C serums, mentioning their product by name alongside a factual, citable claim.
| Client Result: Within 80 days of securing the first two editorial placements, the brand began appearing in ChatGPT’s response to “best vitamin C serum” queries for the first time; appearing in roughly half of the query variations we re-tested. This came entirely from the editorial and consistency work; no paid placement or SEO change was made in the same period. |
Frequently Asked Questions
Why does ChatGPT recommend some brands and not others?
ChatGPT recommends brands it has encountered repeatedly and consistently across credible sources; editorial coverage, structured website content, and review platforms.
Our research found that brands with strong, independently validated third-party coverage were recommended far more often than brands relying primarily on their own marketing or SEO, regardless of product quality.
Does having reviews on Amazon help with ChatGPT recommendations?
Yes, review and retail platforms accounted for a meaningful share of citation sources in our research; though less than editorial and press coverage.
Reviews work best as one part of a broader cross-source presence rather than a standalone strategy. A brand with strong reviews but no editorial coverage and an inconsistent product description elsewhere is still vulnerable to being skipped.
How do I get my skincare brand into ChatGPT answers?
Start by running your own category citation test to see where you currently stand, then close the third-party coverage gap; this was the single strongest factor in our research.
Align your product claims consistently across your website, retail listings, and any press coverage, and pitch ingredient-first editorial content built around the specific use-cases your buyers are asking about.
Does advertising affect ChatGPT recommendations?
Not directly. Paid advertising doesn’t function as a citation source the way editorial coverage, structured website content, or organic reviews do.
Advertising can indirectly help by driving the kind of customer reviews and word-of-mouth discussion that does feed AI recommendation patterns, but it is not itself a signal ChatGPT draws on the way it draws on independent, structured third-party content.
Is AI product recommendation the same as influencer marketing?
No, and this is a common misconception. Influencer content is high in volume but typically low in the structure ChatGPT needs to extract a clean, citable claim; a hundred unstructured social mentions carry less recommendation weight than a single structured editorial review.
Influencer marketing can build broader brand awareness and consumer trust, but it is not currently an efficient lever for AI product recommendation specifically, unless the content is structured around clear, repeated, factual claims.
What the Research Means for Your Brand
Sofia’s fifty queries weren’t an academic exercise. They were the start of a repeatable test any brand can run on itself in the next ten minutes and the start of a four-step process that has since moved real client brands from invisible to recommended.
The product doesn’t need to change. In every case we’ve worked on, the product was already good enough. What was missing was a clear, consistent, independently validated story about it, repeated across the sources ChatGPT actually draws from.
If you want to know exactly where your brand stands today, the fastest way to find out is to run the same test we did or have us run the full analysis for you.
References & Sources
[1] SparkToro & Datos. AI Referral Traffic Report. SparkToro, 2025. sparktoro.com
[2] BrightEdge. AI Search Content Performance Report. brightedge.com, 2024
| About the Author The Darl Team The Darl is a full-service marketing agency led by Lara, a Forbes Next 1000 entrepreneur, LA Times Inspirational Women finalist, and marketing professor. We run original research into how AI engines recommend and cite consumer brands, and apply the findings directly to the beauty, wellness, and e-commerce clients we work with. thedarl.com | Last Updated: June 2026 |
