Poetry vs. Product Data: Why Boy Smells Is Invisible to AI Shopping Agents

Executive Summary

• Brand: Boy Smells. Founded in 2016 in Los Angeles by Matthew Herman and David Kien. Gender-neutral ("post-gender") candle and fragrance brand. Proprietary coconut and beeswax blend. Paraben-free, phthalate-free, paraffin-free. Made in the USA. $44 candles. Available at Nordstrom, Saks Fifth Avenue, and Revolve. Estimated $15-30M annual revenue.
• AI visibility score: 18/150 tests (12%). ChatGPT 22%, Copilot 10%, Gemini 4%.
• The pattern: Editorial presence does not equal AI visibility. One of the strongest editorial footprints in the indie candle category — Vogue, GQ, lifestyle publications — yet surfaces in only 1 in 8 AI tests.
• Key finding: The brand's core differentiator (coconut and beeswax blend) appears nowhere in product descriptions, tags, or structured data. Descriptions are 54-76 words of literary fragrance copy with zero functional data. Automated description score: 5.0/10.
• Root cause: Poetry-first descriptions with no product attributes. Hidden Shopify taxonomy that crawlers cannot access. Zero on-site reviews. Unclaimed Trustpilot profile (254 reviews, 3.7/5). The term "gender-neutral" — the brand's defining position — does not appear in product descriptions or tags.
• Fix complexity: Low. The scent taxonomy already exists in Shopify tags. The sustainability claims are real. The fix is surfacing existing data.

The brand

Boy Smells was founded in 2016 in Los Angeles by Matthew Herman and David Kien. They built a gender-neutral — they call it "post-gender" — candle and fragrance brand around a proprietary coconut and beeswax blend, known for its distinctive pink wax. All products are paraben-free, phthalate-free, and paraffin-free. Made in the USA. $44 candles. Available at Nordstrom, Saks Fifth Avenue, Revolve, and specialty retailers. Editorial coverage in Vogue, GQ, Variety, Marie Claire, and dozens of gift guides.

By any traditional measure, this is a brand that has made it.

The test

I ran five queries across ChatGPT, Gemini, and Copilot, each repeated 10 times per platform — 150 total tests. All tests were run in incognito mode via Playwright with anti-detection measures. No authentication, no history. The 10x repeat methodology measures frequency of brand surfacing rather than relying on single-run results.

The queries:

1. "What's a good gender-neutral scented candle?"
2. "Can you recommend a luxury candle with a unique scent?"
3. "What's the best coconut wax candle brand?"
4. "I need a candle that smells like a spa. Suggestions?"
5. "What are some cool indie candle brands?"

The results

Overall: 18 out of 150 tests (12%)

The platform disparity is stark. ChatGPT surfaces Boy Smells 22% of the time — and 60% for indie brand queries, averaging position #3.4. Gemini surfaces them just 4%. This suggests ChatGPT's training data includes more lifestyle and editorial content where Boy Smells appears, while Gemini may weight structured product data and shopping feeds more heavily.

The "cool indie brands" query (27%) is Boy Smells' strongest showing. But the query that should define them — "gender-neutral candle" — surfaces the brand in only 1 of 5 tests. The brand should own this query. It does not.

The coconut wax query is the one that should keep them up at night: "What's the best coconut wax candle brand?" Boy Smells makes a coconut wax candle. It is arguably their strongest differentiator. They surfaced 7% of the time. Wax type is not mentioned in their descriptions.

Spa candle queries: zero across all 30 tests. Boy Smells' bold, unconventional scents do not align with "spa" query intent — a structural mismatch rather than a data gap.

The description problem

Here is what AI agents actually read when they look at a Boy Smells product page. Automated description score: 5.0/10.

KUSH (their signature candle, the one most cited in editorial coverage):

"A special delivery from our local dealer, Kush is back as a Limited Icon — exclusively at BoySmells.com. Herbal. Floral. Balanced — Kush is a potent rush of earthy indulgence: getting lost in a botanical garden, a secret whispered by the breeze, a touch of amber, the pulse of the earth, a perennial favorite. Light it up when you need to tap into a pure, grounded escape."

63 words. No wax type. No burn time. No scent note pyramid (top, heart, base). No room guidance. No clean ingredient claims. No mention of the coconut-beeswax blend.

ASH:

"We read your sternly-worded emails — and knew we had to bring back this best seller. Another Limited Icon, exclusively at BoySmells.com. Ash is a reminder of the cyclical nature of renewal — that even in destruction, there's a beautiful beginning. It's a smoky baptism — a ritualistic rise to meet your shadow self released through notes of dry hay, black charcoal, and rare woods. Ready to accompany your phoenix era."

65 words. Same pattern. Pure literary brand storytelling. An AI agent cannot extract a single functional product attribute from this text.

VIOLET ENDS is the shortest at 54 words. LES runs to 76 words. All four products follow the same pattern: editorial fragrance copy with zero functional data.

The hidden taxonomy

Here is what makes it worse. Boy Smells actually has scent profile data. Their Shopify tags include scent-botanical, scent-floral, scent-earthy, scent-smoky, scent-woody, and scent-fruity. Each product carries 8-10 tags with reasonably consistent scent family classifications. Tag quality score: 4.0/10 — better consistency than Apotheke's extreme fragmentation, but still not comprehensive.

But tags live in the Shopify JSON API. They do not appear on the product page. Crawlers do not read them. AI agents do not see them. The data exists. It is locked away.

The gender-neutral gap

Boy Smells' entire positioning is gender-neutral fragrance. The phrase "gender-neutral" appears nowhere in their product descriptions or tags. The positioning exists in PR and brand marketing but not in the product data AI agents consume. The brand's core differentiator is invisible to the systems that generate product recommendations.

The review gap

None of the four products I audited had visible on-site reviews. No review system. No aggregateRating in their JSON-LD structured data. Nothing.

Their Trustpilot profile has 254 reviews at 3.7 out of 5. The profile is unclaimed — nobody at Boy Smells is managing it or responding to reviews.

In a category where Diptyque and Jo Malone carry thousands of reviews on Nordstrom and Sephora, the complete absence of on-site review infrastructure is not a minor gap. It is a missing trust signal that affects every AI recommendation engine.

Why this is happening

This is the clearest example in the audit programme of the editorial-to-AI pipeline being broken. Boy Smells has one of the strongest editorial footprints in the indie candle category — frequently cited in Vogue, GQ, and lifestyle publications. Yet they surface in only 12% of AI tests.

Being "known" is not the same as being "recommended."

Boy Smells can surface when an AI agent is asked "name some cool candle brands." They cannot surface when that same agent is asked "recommend a specific candle for my living room that's made from natural wax."

The first query is brand recall. The second is a purchase decision. AI commerce lives in the second query.

P.F. Candle Co. — Boy Smells' closest indie competitor — has the opposite data profile: thin descriptions (23 words) but massive review volume (909 reviews on one product). P.F. Candle Co. wins on social proof where Boy Smells wins on brand voice. Apotheke has the opposite strength — excellent structured descriptions (9/10) but no reviews. Boy Smells sits between both: moderate descriptions, strong brand recognition, but neither the data depth of Apotheke nor the review volume of P.F. Candle Co.

The brands that dominate AI candle recommendations share one trait: functional product data alongside editorial brand presence. Boy Smells has the editorial. They are missing the data.

What Boy Smells could do, in priority order

Phase 1 (quick wins):

• Add structured product data to every description. Keep the brand voice but add a "Details" section: burn time, weight, wax type (coconut-beeswax blend), dimensions, scent note pyramid. This is the single most impactful change.
• Add "gender-neutral" to product descriptions and tags. The brand's core differentiator is invisible to AI agents. Add it explicitly.
• Install a review collection system and add aggregateRating to JSON-LD structured data.
• Claim the Trustpilot profile. 254 reviews at 3.7/5 is a foundation. Claim it and begin responding.

Phase 2 (medium effort):

• Enrich JSON-LD with additionalProperty fields. Add burn time, wax type, scent notes, fragrance family as structured properties.
• Surface the scent tags as visible text on product pages. The taxonomy already exists — it just needs to be rendered.

Phase 3 (longer term):

• Target "gender-neutral candle" and "indie candle brand" roundups. These are the queries where Boy Smells has the best chance of consistent surfacing.
• Build product-attribute editorial coverage. Target roundups for "best coconut wax candles," "best natural wax candles," and "best clean candles" — the queries where Boy Smells should dominate but currently does not appear.
• Submit product feed to Bing Merchant Center. Copilot draws from Bing Shopping data.

Close

Boy Smells could fix this in a week. The literary descriptions can stay. They are good brand copy. But they need a foundation of facts underneath them.

The most editorially visible candle brand in this audit programme is invisible to AI shopping agents. Being culturally relevant and being AI-recommended are two different things. One gets you written about. The other gets you bought.

Your brand story gets you noticed. Your product data gets you recommended.

Which one is your AI visibility actually built on?