Skip to main content

Field note 01 / Five live Shopify audits

An A is notan answer.

Product-page readiness tells you whether an agent can resolve the evidence. It does not tell you whether ChatGPT, Gemini, or another engine recommended the product for a buyer's actual question.

Cohort result

1:4

one deterministic Fix candidate
four answer-observation candidates

Permanent boundary

Readiness evidence is not proof of placement. Colter claims a live recommendation only when a real imported answer shows it.

MEASUREDINFERREDOBSERVED

The useful split

The audit found two different jobs hiding behind one score.

MEASURED43/F PRODUCT

Misen / 5.5 QT Braiser

A fix worth testing

The page visibly answers the buyer's specification question: induction-compatible, oven-safe to 500°F, wide and shallow, with a 5.5-quart capacity. But the exact product page exposed no measured Product entity, Offer, stable identifier, or fulfillment evidence.

INFERRED

Publish truthful Product and Offer evidence, then rerun the unchanged induction and oven-safe prompts.

NOT YET OBSERVED · no engine answer imported

MEASURED100/A PRODUCT

Graza / Drizzle

A perfect score with an open question

Every deterministic check passed. The commercial question did not: will an agent correctly choose Drizzle for finishing and Sizzle for roasting, searing, frying, or baking? Schema presence cannot resolve that comparison by itself.

NOT YET OBSERVED

Empty until a real answer records the product mentioned, the comparison made, the sources cited, and the facts stated.

NOT YET OBSERVED · readiness alone is not the outcome

The five-store ledger

One fix. Four proof questions.

These are public-page audits, not customer results or endorsements. No merchant authorized or endorsed this analysis, and no answer-engine placement was claimed.

Merchant / productExact pageOverallNext jobBoundary
Misen5.5 QT Braiser43/F72/CFIXProduct entity and offer evidence missing
GrazaDrizzle100/A100/APROVEFinishing-versus-cooking answer unobserved
Diaspora Co.Aranya Black Pepper100/A100/APROVEOrigin and use-case answer unobserved
FishwifeSpicy Albacore Tuna100/A100/APROVEHeat and meal-occasion answer unobserved
P.F. Candle Co.Amber & Moss Candle100/A100/APROVEScent-strength answer unobserved

The recommendation job

Measure the page. Change one thing. Prove the same answer moved.

The working AEO thesis is that prompts and comparisons—not raw crawler compliance—are the unit of work. Engine answers need separate observation, and citations can appear, disappear, or decay. These are practitioner hypotheses, not causal findings from this five-store audit.

Thesis source: Gamma growth lead Ravish Agrawal on the Distribution podcast, reviewed August 13, 2026. Colter has not independently established Gamma's reported traffic or attribution claims.

  1. 01

    Baseline

    Measure the store and exact product evidence against an unchanged buyer prompt.

  2. 02

    Fix

    Choose one correction tied to a measured gap—not a generic list of optimizations.

  3. 03

    Prove

    Import a real named-engine answer and retain mention, citation, comparison, and fact-accuracy evidence.

  4. 04

    Monitor

    Rerun the same prompt and targets; report the delta without turning readiness into placement.

Your next evidence point

Find out whether your first job is Fix—or Prove.

Run one read-only audit with exact product pages and the buyer questions that matter. No account required.

Run the recommendation audit