# Q05 ChatGPT — "What's a good retinol to start with?"

Captured 2026-09-02, logged out, 5 fresh sessions, full-page.

## Runs

| Run | Top pick | Cards (price · merchant) | Citations | Ad | Carousel |
|---|---|---|---|---|---|
| 1 | **CeraVe** Skin Renewing Retinol | CeraVe SR $20.44·Walmart; INKEY Starter $15.00·Sephora; CeraVe Resurfacing $16.49·Ulta; Ordinary 0.2% $8.10·The Ordinary | **CeraVe (own site) ×2**; AAD ×3 | — | **Ogee** ($94/$94/$250/$98) |
| 2 | (none — text only) | — | **none at all** | — | **Ogee** |
| 3 | **INKEY List** Starter Retinol | INKEY $15.00·INKEY USA; COSRX $23.50·Ulta; CeraVe SR $24.99·Ulta; Ordinary $8.10·The Ordinary | AAD ×3 | — | **Ogee** |
| 4 | **INKEY List** Starter Retinol | INKEY $15.00·Sephora; COSRX $23.50·Ulta; CeraVe SR $24.99·Ulta; Ordinary $8.10·The Ordinary | AAD ×4 | **Omnilux** (red-light device) | — |
| 5 | **INKEY List** Starter Retinol | INKEY $15.00·INKEY USA; COSRX $23.50·Ulta; CeraVe SR $24.99·Ulta; **Good Molecules** $8.00·Ulta | AAD ×3 | **Omnilux** | — |

**Top-pick churn: INKEY List 3/5, CeraVe 2/5.** Second query with churn.
**Brands: CeraVe (5/5), The Ordinary (4/5), INKEY List (4/5), COSRX (3/5), LRP (1/5), Good Molecules (1/5).**

## 1. A THIRD re-ranking mechanism: semantic match to query intent

**ChatGPT's most frequent top pick — The INKEY List Starter Retinol Serum — is Google's #35.**
Deeper than q02's EltaMD at #15.

But the explanation here is different from q02. There is no magazine driving it. ChatGPT says:
*"specifically designed for first-time users"*, *"specifically designed as a gentle starter"*.

The query asks for a retinol **to start with**. The product is literally named **Starter**
Retinol. This is **semantic matching to query intent**, not editorial ranking.

So there are now three distinct re-ranking forces, and they are separable:

| Mechanism | Seen on | Effect |
|---|---|---|
| **Corpus consensus** | q01, q04 | Mirrors the commerce shelf closely |
| **Editorial** | q02 (Prevention), q03 (magazines) | Promotes magazine favorites over shelf order |
| **Semantic intent match** | q05 (INKEY "Starter") | Promotes products whose *naming* matches the question |

## 2. Strongest evidence yet that the card layer shares Google's commerce data

**Five exact price matches to the cent**, ChatGPT vs the Google organic shelf:

| Product | ChatGPT | Google |
|---|---|---|
| CeraVe Skin Renewing Retinol Serum | **$20.44** | **$20.44** |
| CeraVe Resurfacing Retinol Serum | **$16.49** | **$16.49** |
| COSRX The Retinol 0.1 Cream | **$23.50** | **$23.50** |
| The INKEY List Starter Retinol Serum | **$15.00** | **$15.00** |
| Good Molecules Gentle Retinol Cream | **$8.00** | **$8.00** |

Identical to the cent across five products is not coincidence. **The product-card layer is
almost certainly drawing from the same commerce feed as Google Shopping**, even while the
*ranking* comes from somewhere else entirely. That is the cleanest articulation of the two-layer
finding in the whole study.

## 3. Best SKU precision in the study — 7 exact matches

CeraVe Skin Renewing (G#4), CeraVe Resurfacing (G#2), The Ordinary Retinol Squalane (G#1),
COSRX Retinol 0.1 (G#18), LRP Retinol B3 (G#6), Good Molecules Gentle Retinol (G#3),
INKEY Starter Retinol (G#35).

**Top 40: 6/6 = 100%. Top 10: 4/6 = 67%.** Hypothesis holds — AAD-cited query, 100% at top 40.

## 4. Ads: 2/5, and the regulatory theory survives

Both ads are **Omnilux**, an FDA-cleared red-light therapy **device** — anti-aging adjacent, but
not a retinol. **No topical retinol brand advertised at all.**

Regulatory prediction check: retinol is a cosmetic (unlike benzoyl peroxide), so ads should
appear — they did, 2/5, against acne's 0/5 and vitamin C's 5/5. Partial support. The detail
that *no retinol product itself* advertised, only a device, suggests caution remains in the
category. Running total: **12 of 25 runs (48%)**.

## 5. First brand-owned carousel

Runs 1-3 carried an **Ogee** carousel — $94 to $250 products. Every previous carousel was a
*retailer* (Nordstrom, Nordstrom Rack). This one is a **single brand's own carousel**, and Ogee
is not in Google's top 40 for retinol, nor recommended in any run.

A $94-$250 brand placement sitting under recommendations that cost $8 to $25.

## 6. Run 2 had zero citations

First run in the study with no source chips at all — and it also had no product card and no top
pick. Worth watching whether "no citations" correlates with "no card".