# Field Study 01 > 30 ChatGPT shopping answers across 10 jewelry buying intents: brands were named 248 times, and only 88 mentions (35.5%) linked the buyer to the brand's own store. Who owns the path, who leaks it, and which intents are still wide open. - Canonical: https://caeliai.com/research/field-study-01-jewelry - Markdown: https://caeliai.com/research/field-study-01-jewelry.md - Publisher: Caeliai (https://caeliai.com), founder-led AI commerce research and advisory studio by Kalan Peace - Official LinkedIn: https://www.linkedin.com/company/caeliai/ - Founder LinkedIn: https://www.linkedin.com/in/kalan-peace-2b2a18198 - Contact: contact@caeliai.com --- ChatGPT named jewelry brands 248 times across 30 fresh shopping answers: ten buying intents, each asked three times. Only 88 of those mentions, 35.5%, linked the buyer to the brand’s own store. The rest sent the buyer to a retailer, a marketplace, an article, or nowhere useful. **Figure:** Fig. 01 · Where the mention sends the buyer N = 248 mentions · 30 answers · June 29, 2026 **64.5%** of the time ChatGPT named a jewelry brand, the buyer was sent somewhere other than the brand’s own store. Owned path 88 of 248 Somewhere else 160 of 248 ## What we did. On June 29, 2026 we asked ChatGPT the questions real jewelry shoppers ask: ten buying intents, each run three times in a fresh US session in Instant mode, 30 answers total. Every answer was preserved as a public ChatGPT share link and decoded from its raw response, 30 of 30, so every number below traces back to a real conversation rather than a screenshot. **01** ### Prompt like a shopper. Ten jewelry buying intents phrased the way a buyer would phrase them, from solid gold to pearl gifts. Each prompt ran three times in a fresh session. **02** ### Capture the evidence. Each answer saved as a public share link and decoded from the raw response payload: brands named, links shown, and the sources ChatGPT leaned on. **03** ### Follow every link. Each brand mention was classified by one question: did the answer link the buyer to the brand’s own store, or not? - 01 Solid gold - 02 Lab-grown engagement - 03 Gifts under $1,000 - 04 Minimalist everyday - 05 Pearl gifts - 06 Ethical / sustainable - 07 Brand comparison - 08 Birthstone / personalized - 09 Affordable luxury - 10 Occasion gifts *Caption:* The full prompt texts are listed in the source section at the end of this page. One caveat up front: the brand-comparison prompt names Catbird, Mejuri, Quince, and Brilliant Earth directly, so those four brands carry a few guaranteed mentions. *Caption:* Diagnostic plate · What a successful recommendation contains Observed July 10, 2026 · separate validation example Wanderlust + Co **14K Gold Vermeil Classic Chain Necklace** $169.00 · In stock Sold by Wanderlust + Co wanderlustandco.com/products/... **5/5** Complete owned path 01 Brand is named 02 Specific product is shown 03 Image, price, and availability appear 04 Brand is the named merchant 05 Click lands on the exact official PDP 01 **Named, no owned link** Gorjana appeared 12 times. Official-store links: 0. A Monica Vinader product routed to Nordstrom. Two checked Catbird products surfaced resale paths instead of an official PDP. One Susie Warner result exposed $89 and $113.50 price fields. *Caption:* The five checks are the Caeliai diagnostic framework. The examples come from July product-level checks, kept separate from the 30-answer pulse. ## A mention is not a path. When ChatGPT recommends a jewelry brand, the recommendation is only worth money if the buyer can click through to the brand’s own store. Often the name lands but the link routes somewhere else. That is the difference this study measures. Shopper prompt -> Brand named -> Retailer / listicle / no link The mention lands. The sale routes somewhere else. *Caption:* 160 of 248 mentions in this study followed some version of this route. 88 ended on the brand’s own store. And an owned path is itself one rung on a taller ladder. What a brand actually wants from ChatGPT is a climb: from being known at all to being consistently, accurately shoppable. - 01 **Known** ChatGPT can name the brand at all. - 02 **Recommended** The brand shows up in a real shopping answer. - 03 **Product shown** An actual purchasable product, not just the brand name. - 04 **Shoppable** A product card with image, title, price, availability, and merchant. - 05 **Official path** The click lands on the brand’s own product page, with the brand as the seller. - 06 **Accurate** Right image, right price, right variant, in stock, nothing stale. - 07 **Consistent** Holds up across repeated runs, not one lucky answer. *Caption:* The 30-answer pulse in this study measures the jump from rung 02 to rung 05 at brand level. The July product-level checks probe rungs 03, 04, and 06. Rung 07 is why every prompt ran three times. ## Being named is not being bought from. 54 distinct brands were named across the 30 answers. Fame and path ownership turned out to be nearly independent: the most recommended brands leak, and some of the tightest operators are barely famous. **Figure:** Fig. 02 · Mentions vs owned paths, top brands Bar length = times named · green segment = linked to own store - Mejuri: 9/23 - Catbird: 11/20 - Quince: 6/17 - Brilliant Earth: 4/14 - Blue Nile: 3/13 - Aurate: 6/12 - Gorjana: 0/12 - Tiffany & Co.: 4/11 - Monica Vinader: 4/10 - VRAI: 2/10 - GLDN: 6/6 - David Yurman: 4/6 **Figure:** Owned-store links / times named, per brand 30 ChatGPT answers · June 29, 2026 Famous and leaking Mejuri is the most recommended brand in the set: named 23 times, linked to its own store 9 times. The category’s biggest AI winner still loses the buyer path in roughly 6 of 10 mentions. Named, never linked Gorjana was named 12 times across six different intents and never once got a link to its own store. Full AI visibility, zero owned demand. Small and airtight GLDN converted all 6 of its mentions into owned-store links. Baby Gold and Oradina went 5 for 5. Less famous, but every mention lands at home. ## The shelf is short, and the same names keep it. Every ChatGPT answer is a shelf with roughly eight slots. Three numbers describe how those slots get filled. - **~8** - Brands named per answer (248 mentions / 30 answers) - **57%** - Of all mentions went to the ten most-named brands (142 of 248) - **21 of 54** - Brands were named exactly once and never came back *Caption:* The head of the shelf is sticky: the same regulars fill it run after run. A brand named once is not visible; it got lucky. ## Some buying intents are still wide open. The ten intents behave like different rooms. Some are crowded with the same incumbents every run. Some have no settled answer yet. Counting the distinct brands ChatGPT named per intent across three runs shows which is which. **Figure:** Fig. 03 · Distinct brands named per intent 3 runs per intent · fewer brands = less settled - Occasion gifts: 19 - Minimalist: 15 - Ethical: 15 - Solid gold: 13 - Engagement: 13 - Gifts $1k: 12 - Pearls: 11 - Afford. luxury: 11 - Comparison: 8 - Birthstone: 5 **Figure:** The comparison prompt names four brands, so its count runs low by design June 29, 2026 Birthstone and personalized jewelry is the standout: five brands named across three runs, and only Ana Luisa named in 3 of 3. Occasion gifts is the opposite kind of open: 19 brands and no clear owner. Meanwhile lab-grown engagement rings and solid gold returned the same regulars every run: Blue Nile, Brilliant Earth, Clean Origin, James Allen, and Ritani for engagement; Aurate, Baby Gold, Catbird, GLDN, Mejuri, Oradina, and Quince for solid gold. For a jewelry brand, the move is not to fight the locked intents. It is to own a wide-open one before ChatGPT settles on its answer. ## A pattern worth testing: specialists own, generalists leak. **Broad and leaking** The brands ChatGPT names across the most intents tend to lose their paths. Mejuri appears in 9 of 10 intents and keeps 39% of its links. Quince spans 7 intents at 35%. Blue Nile spans 6 at 23%. **Narrow and owned** Every brand that converted 100% of its mentions appears in three or fewer intents: GLDN, Baby Gold, Oradina, Bario Neal. Catbird is the one exception that is both broad and majority-owned: 8 intents, 55%. This is a correlation in a small sample, not a law. It is the kind of pattern we publish so it can be tested: if it holds at larger sample sizes, breadth of AI visibility and ownership of the buy path are separate problems that need separate fixes. ## What sits underneath the answers. Two layers below the visible answer shape these outcomes. The first is the catalog layer. Catbird is the strongest broad performer in this study: named in 20 of 30 answers, majority-owned paths. Yet in a separate product-level check we ran on July 1, its five tested products were missing from its own storefront catalog feed, and catalog lookups for its product names matched other sellers instead, including a “Katbird Bracelet” on a store that is not Catbird. A brand can be winning today’s answers while its catalog layer quietly belongs to someone else. That product-level problem is the subject of our next full research paper. The second is the source layer: the pages ChatGPT leaned on while composing these answers. In the decoded retrieval and support payload of the 30 runs, the most frequent domains were not brand sites. - reddit.com: 38 - youtube.com: 35 - brilliantearth: 25 - forbes.com: 21 - stoneandstrand: 19 - wikipedia.org: 17 - mejuri.com: 17 *Caption:* Domain counts observed in the decoded retrieval and support payload of the 30 answers. These include supporting sources and search context, not only direct citations. Reddit threads, YouTube videos, and Forbes roundups outrank almost every brand’s own site. That is the trap inside “AI visibility”: a brand can be visible because a Reddit thread or a magazine roundup mentions it, not because of anything on its own pages. Visibility borrowed from third parties is visibility a brand does not control. ## Limitations, stated up front. This is a directional field study, not a market ranking. Each intent ran three times, so per-intent results are counts like “named in 3 of 3 runs,” not finely measured shares. Everything comes from one model, one mode, and one day: ChatGPT Instant, US sessions, June 29, 2026. These systems change without notice, so every figure carries its date, and brands named once or twice are anecdotes, not rates. Path classification here is binary and brand-level: owned-store links divided by times named. This study does not split the non-owned 64.5% into retailer links versus no link at all, and it does not pin each owned link to a specific run. The source-domain counts come from the decoded retrieval and support payload, which includes search context, not only confirmed citations. And the brand-comparison prompt names four brands directly, which pads their mention counts; we footnote it wherever it matters. Next in this series: a larger calibrated benchmark for the same category, with more runs per prompt and product-level path classification, so the binary split above can become the full owns-leaks-no-path picture. Caeliai Field Study 01. 30 ChatGPT Instant-mode shopping answers, US, captured June 29, 2026 as public share links and decoded from raw responses, 30 of 30. 10 prompt types, 3 fresh runs each. 248 brand mentions across 54 distinct brands after merging duplicate spellings: 88 owned-store links (35.5%), 14 brands linked to their own store in at least half their mentions, 12 linked sometimes, 28 named and never linked. Catalog observations from a separate 20-product check on July 1, 2026. Independent research: no brand paid to be in it. The ten prompts: affordable luxury jewelry that does not look cheap · birthstone rings and personalized jewelry online in the US · compare Catbird, Mejuri, Quince, Brilliant Earth, and similar brands · lab-grown diamond engagement ring comparison · ethical or sustainable jewelry brands · fine jewelry gift under $1,000 · minimalist everyday jewelry with solid gold options · graduation, anniversary, or bridesmaid gifts · pearl earrings or a pearl necklace as a gift · best solid gold jewelry brands online in the US. ## Where does ChatGPT send your buyers? We run real shopping prompts for your category and show you where the recommendations land: your store, a reseller, or nowhere. One form, no meeting.