A woman with fine, low porosity waves has tried four curl creams this year. Every one of them left her hair limp by the afternoon. In 2020 she would have searched Google, opened six tabs, and read her way through a pile of listicles. In 2026 she does something quicker. She opens ChatGPT and types: "I have fine 2b waves, low porosity, my hair goes flat and greasy by lunchtime. Which curl cream should I try, and which ingredients should I avoid?"
Back comes a considered answer. It explains why heavy butters weigh down low porosity hair, it warns her about a few specific ingredients, and it names three products she should consider.
Here is the question every textured hair brand must now ask itself, and it is uncomfortable: were you one of the three?
If you were not, the reason is almost certainly not the quality of your formula. The truth is that AI assistants do not rank brands the way a search engine ranks pages. They assemble an answer out of sources they can retrieve, read, and quote. A brand that publishes beautiful campaign copy and very little else gives the model nothing it can use. You are not being rejected. You are simply not present in the evidence.
What the citation data actually shows
We should begin with facts rather than opinion, because this subject attracts a great deal of confident guessing.
In June 2026, the beauty platform Novi published an analysis of the sources ChatGPT draws on when answering beauty questions. The study covered 10.7 million citations gathered between 22 January and 20 May 2026, across more than 98,000 source websites. It is, as far as public data on this subject goes, unusually large.
The findings deserve to be read slowly:
- Reddit was the single most cited source for beauty queries, and by a significant margin.
- The next most cited were Who What Wear, Wikipedia, Sephora.com, and Allure.
- Those five sources together accounted for roughly a third of all citations in the dataset.
- For skincare specifically, brand-owned websites did appear in the top ten, but never in the top five.
Read that last point again, because it is the one that matters most to a brand founder. Your own website can reach the conversation. It rarely leads it.
Meanwhile the traffic itself is no longer a rounding error. Adobe Analytics reported that traffic to retail sites from generative AI tools rose 693.4% year over year across the 2025 holiday season, from 1 November to 31 December, and named personal care among the categories where shoppers leaned on AI assistants most heavily. The base is still modest compared to classic search. The direction is not ambiguous.
Why the curl category is unusually exposed to this shift
Every category is affected, but textured hair care is affected more than most, for three reasons.
The questions are highly specific. Nobody with curls searches for "shampoo". They search, or now they ask, for a sulphate-free cleanser for high porosity 3c hair that does not cause build-up in hard water. This kind of question is precisely what an assistant is good at handling and precisely what a category page is bad at answering.
The knowledge is genuinely complicated. Porosity, curl pattern, protein and moisture balance, humectants in humid weather, drying alcohols, silicone build-up. Shoppers know they need guidance, and they know a product page will not give it to them honestly. So they ask something that has no commercial interest in the answer.
The community is exceptionally strong. Textured hair communities have been trading detailed routine advice online for well over a decade. That accumulated discussion is exactly what the citation data shows the models leaning on. This is why Reddit sits at the top of the list, and it is why a brand with no genuine community presence is invisible in a way that paid social cannot fix.
The mistake almost every curl brand is making
Let me describe the typical product page in this category, because I suspect it will be familiar.
There is a beautiful photograph. There is a name, perhaps something poetic. There is a paragraph about weightless definition and bouncy, frizz-free curls. There is a price, an add-to-cart button, and a review widget that loads a second after the rest of the page.
Now consider what an assistant can actually extract from that page in order to answer our shopper with fine, low porosity 2b waves. Which curl patterns is this formulated for? Unstated. Which porosity? Unstated. What are the key ingredients and what do they do? Unstated. What has been deliberately left out? Unstated. What do real customers with her hair type say? Loaded by a script, therefore invisible.
The page is a lovely piece of brand expression and a completely empty document. When the model looks for something to cite, it finds nothing, so it goes to Reddit, where a stranger has written four paragraphs about exactly this problem.
This is the whole difficulty in one image. You are not losing to a better competitor. You are losing to a forum post.
What to change, in order
1. Make every product page state what it is for
Not what it feels like. What it is for. Curl patterns, porosity levels, hair density, the concerns it addresses, and the conditions where it performs poorly. That last one earns more trust than any marketing sentence you will ever write, and it makes the page quotable.
If a shopper can ask "is this suitable for low porosity fine waves" and your page answers plainly, you have become citable. If the answer requires interpretation, you have not.
2. Publish the real ingredient story, including the exclusions
Name the key actives and explain what each one does for textured hair, in plain language. Then be equally explicit about what you have excluded, whether that is sulphates, drying alcohols, silicones, or protein.
Exclusions are among the most searched attributes in this entire category. A shopper who has decided that a certain ingredient ruins her hair is searching for its absence, and an assistant asked to avoid that ingredient needs a page that says so in words.
3. Get your customer reviews into the actual HTML
This one is quietly expensive to ignore. Most brands collect reviews through a third-party widget that injects the text after the page renders. The reviews exist for a human visitor and do not exist for a crawler or a retrieval system.
In a category where the citation data shows models leaning heavily on real user experience, this is the strongest evidence you own, and it is sitting behind a script. Server-render it.
4. Add and validate structured data across the catalogue
Product, Offer, Review, FAQPage, Organization. Publish price, availability, variants and ratings as machine-readable data, then validate it rather than assuming it works.
There is an unpleasant irony here. When your own product page lacks clean structured data, the retailer listing that does have it becomes the authoritative source for your own product. You created the demand and handed the citation to a marketplace.
5. Build the education layer that comes before the purchase
The porosity test, the curly girl method, refreshing day three curls, why a product stopped working, protein overload, hard water. These questions arrive long before anybody is ready to buy, and at present publishers and forums answer all of them.
Answer the question fully and honestly before you mention a single product of yours. A guide that solves the reader's problem and then suggests a relevant product is useful. A guide that is an advertisement wearing a guide's clothing is transparent to readers and to models alike.
6. Participate honestly where the discussion already happens
We have to be direct about this, since the data leaves little room for comfortable alternatives. Reddit was the most cited source for beauty answers by a significant margin. A brand entirely absent from genuine community discussion is absent from a large share of the material the assistant is reading.
This does not mean astroturfing, which communities detect quickly and punish permanently, and which would damage you far more than silence. It means real participation: a founder or formulator who answers technical questions properly, discloses affiliation openly, and follows each community's rules without trying to bend them.
And do the on-site work first. Someone who finds you in a discussion will open your website within seconds, and that website has to confirm what was said about you.
What this is not
It is worth naming a few things that will not solve this, because the market is currently full of people selling them.
Adding a file of instructions for AI crawlers to your server does not make an assistant recommend you. Stuffing a page with the phrase "best curl cream" does not make it true or citable. Publishing forty thin blog posts a month produces exactly the kind of low-value content that both search engines and models are increasingly able to recognise and discount.
There is no shortcut here, which is, in the end, encouraging news for brands that actually know their category. The thing that gets you cited is publishing genuinely useful, specific, verifiable information about products you understand deeply. That is a real advantage for a focused curl brand and a real disadvantage for a large house writing broad copy for everybody.
The small brand advantage
One might assume the large beauty houses will simply dominate this the way they dominate shelf space. I do not believe they will, at least not quickly.
Big brands write cautiously and generally, for the widest possible audience, with legal review at every step. That produces copy which says very little in a great many words, which is the worst possible input for a system trying to extract a specific answer.
A focused textured hair brand can do the opposite. It can say, with authority, that a particular formula suits high porosity 4a coils in humid climates and performs poorly on fine 2a waves. That sentence is specific, honest, and useful. It is exactly the kind of statement a model reaches for when a real person describes a real head of hair.
Precision, in this arena, beats budget. It is not often one can say that in beauty.
Where to begin this week
If you do one thing, take your three best-selling products and rewrite their pages so that each one states the curl patterns, porosity levels, and concerns it is built for, the key ingredients and what they do, and what the formula excludes. Then get your reviews server-rendered into those same pages.
That is not a rebrand and it is not a six-month project. It is the difference between a page a machine can quote and a page it cannot.
Licheo builds and improves websites for curly and textured hair brands and beauty brands, covering product and ingredient content, e-commerce structured data, education content, and the technical repairs that image-heavy beauty sites usually need. Get in touch to talk about your brand.
Sources
- Novi analysis of 10.7 million ChatGPT beauty citations, 22 January to 20 May 2026, reported by PPC Land
- Adobe Analytics holiday shopping season data, 1 November to 31 December 2025