I Audited Apotheke Against AI Shopping Agents. The Best Description Template in the Set, Zero Social Proof.

Executive Summary

• Brand: Apotheke (apothekeco.com). Brooklyn-based home fragrance brand founded in 2011 by Chrissy Fichtl. $46 candles. Premium soy wax blend with essential and perfume-grade fragrance oils. Made in USA. Shopify store. Estimated $10-20M annual revenue. Stocked at Bloomingdale's, Nordstrom, and Anthropologie.
• AI visibility score: 21/150 tests (14%). ChatGPT 8%, Gemini 24%, Copilot 10%.
• The pattern: Best-in-class descriptions undermined by zero social proof and wildly inconsistent metadata
• Key finding: The structured description template (Fragrance / Product / Vessel / Details) is the most AI-friendly format in the audit set — automated score 9.0/10. But the brand has zero on-site reviews, no aggregateRating markup, and a dormant Trustpilot profile. Tag quality scores 3.2/10. The gap between description quality and tag quality is the widest in the entire audit programme.
• Root cause: The 9/10 descriptions exist in a vacuum. No reviews to validate. No consistent tags to classify. The signal is half-built.
• Fix complexity: Medium. The description template is already done — most brands need to build that first. Apotheke needs to build everything else.

The brand

Apotheke was founded in 2011 in Brooklyn by Chrissy Fichtl, with her husband Sebastian Picasso joining production from 2012. The range centres on $46 candles made from a premium soy wax blend with essential oils and perfume-grade fragrance oils. All hand-poured. All made in the USA. The aesthetic is modern and minimalist — matte black and white glass vessels — anchored by the bestselling Charcoal scent.

The retail distribution is strong. Bloomingdale's, Nordstrom, Anthropologie, plus 1800Flowers, Verishop, and Orchard Mile. They also run candle-making classes at their Brooklyn Studio — an experiential angle that most home fragrance brands do not have.

This is a brand with genuine premium positioning, real retail credibility, and a product that competes directly with Diptyque and Jo Malone on quality if not on name recognition. It sits between the premium accessible tier (Nest, Voluspa) and ultra-luxury European houses (Diptyque, Byredo) — premium pricing and Brooklyn artisan positioning, but without the editorial frequency of Nest or the brand recognition of Diptyque.

The test

I ran five queries across ChatGPT, Gemini, and Copilot, each repeated 10 times per platform — 150 total tests. The queries ranged from niche scent queries ("What's a good charcoal-scented candle?") to generic category queries ("What's the best candle for a clean, modern scent?"), lifestyle queries ("Can you recommend a minimalist luxury candle brand?"), attribute-specific queries ("I need a 3-wick candle for a large room"), and provenance queries ("What are the best Brooklyn-made candle brands?").

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 results

Overall: 21 out of 150 tests (14%)

The pattern is stark. The Charcoal niche query accounts for 12 of Apotheke's 21 surfacings — more than half of all AI visibility from a single product on a single query type. When Charcoal surfaces, it consistently ranks #1. Gemini called it "the definitive charcoal candle" and "the gold standard." Copilot surfaced it via Liberty London at £23.

Strip out the Charcoal query and Apotheke's visibility drops to 9/120 (8%). For generic category queries and attribute-specific queries — the kind of questions where any premium candle brand should appear — zero surfacings across 60 tests.

Brooklyn provenance is a secondary asset (6/30, 20%), driven primarily by Gemini (4/10). ChatGPT and Copilot largely ignore the Brooklyn angle. Brooklyn Candle Studio and Keap consistently outrank Apotheke on their own home turf.

The descriptions: a model for the category

This is where Apotheke stands apart from every other brand in this audit programme. Their product description template is genuinely excellent. Automated score: 9.0/10.

I reviewed four products. Take the Charcoal Classic Candle (174 words). The description is broken into structured sections:

"The Fragrance" — the scent story plus a full note pyramid: "Top Notes: Smoky Embers, Pink Pepper / Middle Notes: Black Tea, Cedarwood / Bottom Notes: Labdanum, Vetiver."

"The Product" — soy wax blend, clean-burning, essential and perfume-grade fragrance oils.

"The Vessel" — matte-black glass, dimensions.

"The Details" — net weight 10.5oz (296g), burn time 60-70 hours, fragrance family: Woody, Made in USA.

Earl Grey Bitters follows the same template at 169 words. Elderflower Sugar at 152 words. Kindling at 146 words. Every product delivers structured, extractable product data across four clearly labelled sections. Every section answers a different type of AI query. The scent pyramid answers "what does it smell like." The product section answers "what is it made from." The details section answers "how long does it burn" and "what size is it."

This is what I mean when I tell brands to write descriptions for AI agents. Apotheke is the template. Literally.

So why only 14%?

Three problems, all outside the descriptions themselves.

1. Zero reviews. Anywhere.

Apotheke has no on-site review system. None. No aggregateRating in their JSON-LD structured data. Their Trustpilot profile sits at 3.5 out of 5 with zero reviews — it appears unclaimed and unmanaged.

For a brand generating an estimated $10-20M annually, stocked at Bloomingdale's and Nordstrom, this is an extraordinary omission. Reviews are one of the strongest signals AI agents use to validate product quality. Without them, even the richest product description is an unverified claim.

Compare this to P.F. Candle Co., which has 909 reviews on a single product. Or Boy Smells with 254 Trustpilot reviews. Apotheke has none.

2. Tags are wildly inconsistent.

The Charcoal candle has 9 tags including scent family (Woody) and room size ("Medium space"). Earl Grey Bitters has 13 tags including seasonal tags ("Summer-scents") and room size. The Kindling candle has 1 tag.

Tag quality score: 3.2/10. The gap between description quality (9.0) and tag quality (3.2) is the widest in the entire audit set. No other brand has such a large delta between their best and worst product data attributes.

This is not a tag strategy problem. It is a tag consistency problem. The taxonomy exists for some products and is completely absent for others. A fragrance family tag on one product out of five is worse than no tags at all, because it means the classification system exists but was never completed.

3. Limited editorial footprint.

The Charcoal candle has editorial traction — it appears in gift guides and roundups, and when surfaced it ranks #1. But the brand does not consistently appear in the mainstream "best candles" listicles that AI agents draw from when generating recommendations.

Diptyque, Jo Malone, and Byredo dominate the "luxury candle" query space across all platforms. Nest and Voluspa dominate the "premium but accessible" tier. For generic queries like "best candle for a clean, modern scent," zero surfacings across 30 tests.

Why this is happening

In the candle category, the brands that score highest on AI visibility tend to have one of two advantages: either massive brand recognition (Diptyque, Jo Malone) or strong review ecosystems (P.F. Candle Co.).

What no brand in this audit set has is the combination of excellent descriptions AND strong social proof. Apotheke has the best descriptions. P.F. Candle Co. has the most reviews. Neither has both.

The brand that figures out this combination first — Apotheke-quality descriptions with P.F. Candle Co.-level review volume — will own the AI visibility category for candles.

Department store presence does not equal editorial presence. Availability at Bloomingdale's, Nordstrom, and Anthropologie provides retail credibility but does not generate the editorial roundup citations that feed AI recommendations.

What Apotheke could do, in priority order

Phase 1 (quick wins):

• Implement a review collection system. Klaviyo Reviews, Judge.me, Yotpo, or Stamped.io — any app that collects reviews and feeds them into JSON-LD aggregateRating markup. Trigger post-purchase review request emails. At the volume Apotheke does across DTC and department stores, even a 5% response rate should generate hundreds of reviews within months. This is the single highest-impact fix.
• Claim and build the Trustpilot profile. The page exists but is empty and unclaimed. Even 50-100 reviews at 4+ stars would create a social proof signal that currently does not exist.
• Standardise tags across all products. Apply a consistent taxonomy to every SKU: fragrance family, room size, season, scent intensity. Every product should have the same tag categories filled. Kindling having 1 tag while Earl Grey Bitters has 13 is unacceptable.

Phase 2 (medium effort):

• Enrich JSON-LD with additionalProperty fields. The description text already contains burn time, weight, dimensions, fragrance family, and scent notes. Add these as structured fields in the schema to make the data machine-readable, not just human-readable.
• Add room size and occasion context to descriptions. The templated structure has room for one more section covering recommended room size, occasion (everyday, entertaining, gifting), and season. These are high-frequency query modifiers.
• Add cross-product comparison context. When two products share a fragrance family (Charcoal and Kindling are both Woody), the descriptions should differentiate them. AI agents need to recommend ONE product, not a category.

Phase 3 (longer term):

• Target editorial roundups. Pitch the Charcoal Classic Candle to editors at Architectural Digest, NYT Wirecutter, Bon Appetit, Domino, and Real Simple. One citation in each publication's shopping guide would dramatically change AI recommendation probability for generic queries.
• Create fragrance family landing pages for each category (Woody, Citrus, Floral, Fresh) targeting long-tail queries like "best woody candle for winter."
• Submit product feed to Bing Merchant Center. Copilot draws from Bing Shopping data. Without a Bing feed, Apotheke is structurally invisible on one of the three major AI shopping platforms.
• Leverage the candle-making classes as a content angle. "Brooklyn candle maker" is a differentiated story that could earn editorial coverage and build the off-site signals that drive AI recommendations.

Close

Apotheke is the most instructive brand in this audit programme. Not because they are the best at AI visibility. Because they show exactly where the ceiling is when you get one thing perfectly right and neglect everything else.

The description template is a 9 out of 10. Genuinely the best I have seen across 17 brand audits. If you are a DTC brand looking for a model of how to write product descriptions for AI agents, start with Apotheke's Charcoal candle. The Fragrance, The Product, The Vessel, The Details. Scent pyramid, burn time, dimensions, fragrance family.

But descriptions alone do not break through. Zero reviews. Zero aggregateRating. An unclaimed Trustpilot. Tags that range from excellent to non-existent within the same catalogue. One product carrying more than half the brand's entire AI visibility.

The template is the easy part to copy. The hard part is building everything around it — the reviews, the editorial presence, the consistent metadata, the social proof that tells an AI agent "this product is not just well-described, it is well-validated."

Apotheke has built the best room in the house. They just forgot to put a door on it.