Opinion Briefing (Crabstone)

Renfrew Is Selling 'Clean.' The Machine Reads Only Ingredients.

Gregg Renfrew's Gen Z brand Beecee launches on a 'clean' pitch built on what a formula leaves out. But machine-mediated discovery rewards ingredients and evidence — The Ordinary alone takes 7% of AI beauty answers — and 'clean' is a category no regulator defines and no algorithm can weigh.

Sir John Crabstone

Gregg Renfrew is betting Gen Z still buys provenance. Her brand Beecee launches on September 1 with three products. The pitch is “clean,” a claim you can only check by reading a label. The reading is now done by machines.

She has standing to make it. Renfrew founded Beautycounter, watched Carlyle buy it for $1 billion in 2021, then bought it back out of foreclosure three years later. Beecee is her third attempt to sell safety as a category. She is making, a second time, the argument she won thirteen years ago.

Her distribution follows the same instinct. Beecee is recruiting young women as commissioned advocates and touring meet-ups from Nashville to Charleston. Renfrew calls the model “collective commerce” and is careful to distinguish it from the MLM label critics attached to Beautycounter — a charge she has never accepted. It is a bet on human word of mouth, placed just as the recommending is handed to software.

Asked whether the young still care about clean, Renfrew answers “unequivocally yes”, citing young women alert to “chemicals that are disrupting their hormones.” She may be right about the caring. That is not a demand problem — it is a discovery problem.

Ask a chatbot for a good lip gloss and it does not return a value. It returns ingredients. The brands winning that contest name their actives. Glossy reported an analysis by the firm 5W AI: The Ordinary turns up in 7 percent of AI beauty answers. The best-placed legacy brand, Estée Lauder, ranks eighteenth. AI rewards “ingredient transparency” and “more facts,” 5W AI’s founder said.

Here is the trap in Renfrew’s defense. The generation that reads ingredient lists is the one the machine serves best; a curiosity about what a formula contains leads to the brand that names its actives, not the brand that advertises what it leaves out.

What the regulator will not define, the algorithm cannot weigh.

The machine will not even quote her. In an audit of 1,500 AI beauty answers, a brand’s own website supplied 2.4 percent of the sources the models cited; the rest came from pages the brand does not control. A label may say “clean” as loudly as it likes. The recommendation is built from everyone but the seller.

The other machine, TikTok, rewards a hook, not a heritage claim; a gloss trends for how it catches the light, not for what it omits. Renfrew spent thirteen years teaching people to read the label. To the machine that reads for them now, “clean” is a password it was never issued.