Project
    eCommerceKilled

    Shopify AI Product Optimizer

    Shopify AI Product Optimizer was a killed Shopify app experiment that tried to make product-catalog optimization measurable, reviewable, and reversible. The source app combined deterministic SEO checks with merchant-contextual AI suggestions, Shopify product updates, background jobs, audit logs, and one-click revert behavior.

    Table of contents

    Narrative

    01

    Could a product catalog become a review queue?

    Shopify AI Product Optimizer was a February 2026 experiment in making catalog SEO work inspectable. A deterministic analyzer scored the existing listing and identified field-level gaps before any model was asked to write. The model then produced bounded alternatives with rationales, leaving the merchant with a queue of evidence-backed choices rather than an unexplained rewrite.

    02

    Generation did not equal permission

    The app separated suggestions by operational risk. SEO titles and meta descriptions could enter a constrained safe-auto mode; product titles and descriptions remained broader display changes; handles were high risk because they alter URLs. Suggest-only was the default, and the product model preserved current values, proposed values, rationales, statuses, snapshots, audit records, and a revert path. The design treated AI output as a candidate mutation that still needed policy, scope, or human approval.

    03

    The machinery reached staging before the idea reached proof

    The experiment accumulated an embedded Remix/Polaris app, Shopify GraphQL and webhooks, Postgres/Prisma state, BullMQ jobs, a separate worker, billing controls, and Render/Neon/Upstash deployment plans. But the last source handoff still listed the real install-to-revert loop as the primary work ahead. Development stopped on February 6, the former deployment no longer responds, and Maggie classifies the Shopify app experiment as killed. The archive keeps the useful system boundaries without pretending the product reached validation.

    System surfaces

    Catalog health dashboard

    Shopify merchant or catalog operator

    A merchant-facing summary of product count, analysis coverage, pending suggestions, and quick paths into review or automation.

    Product analysis queue

    Catalog operator prioritizing product work

    Product-level SEO scores, issue state, suggestion counts, and timestamps make optimization work sortable and inspectable.

    Suggestion review

    Merchant responsible for brand and catalog quality

    Current and proposed field values, rationales, and risk levels support selective human approval instead of wholesale rewriting.

    Automation controls

    Operator defining automation scope

    Suggest-only, safe-auto, and full-auto modes combine with triggers and product criteria to limit which listings and fields automation may touch.

    Audit and revert

    Merchant reviewing or reversing catalog changes

    Snapshots, change logs, statuses, conflict checks, and revert jobs preserve an attributable path back from applied model output.

    Implementation stack

    Embedded Shopify app

    • Remix
    • React 18
    • TypeScript
    • Shopify Polaris
    • App Bridge
    • Shopify GraphQL Admin API 2026-01

    Analysis and generation

    • deterministic weighted SEO checks
    • OpenAI chat completions
    • field-specific prompts
    • merchant brand voice
    • target audience
    • JSON suggestion/rationale responses
    • field risk tiers

    State and jobs

    • Postgres
    • Prisma
    • BullMQ
    • Redis
    • product snapshots
    • suggestion lifecycle
    • change logs
    • analysis runs
    • automation settings

    Commerce and operations

    • Shopify OAuth sessions
    • product and bulk-operation webhooks
    • privacy webhooks
    • usage billing
    • Render web and worker
    • Neon Postgres
    • Upstash Redis

    Safety boundary

    • suggest-only default
    • safe-field allowlist
    • human review
    • billing caps
    • audit trail
    • conflict checks
    • revert

    Evidence trail

    data model

    Automation and permission boundary

    A product can enter the pipeline through a webhook, schedule, or manual run, but criteria, billing, risk mode, and explicit review determine whether generation becomes action.

    10entities1diagram
    data model

    Product evidence and suggestion lifecycle

    The experiment preserved a source product snapshot, deterministic analysis evidence, generated alternatives, and a reversible action record instead of treating model output as the catalog's new truth.

    8entities1diagram
    workflow

    Shopify AI Product Optimizer catalog review loop (killed experiment)

    Historical process model for the killed Shopify app experiment: deterministic catalog evidence identified field-level gaps, merchant context bounded generated alternatives, risk and human review controlled permission to act, and snapshots plus change logs kept applied work reversible.

    11steps1record
    Created: Jul 13, 2026
    note

    Shopify AI Product Optimizer Intake Dossier

    Shopify AI Product Optimizer was a killed Shopify app experiment that tried to make product-catalog optimization measurable, reviewable, and reversible. The source app combined deterministic SEO checks with merchant-contextual AI suggestions, Shopify product updates, background jobs, audit logs, and one-click revert behavior.

    1topic1scope
    note

    README

    Source Summary AI Product SEO Optimizer A Shopify app that uses AI to automatically analyze and optimize product listings for SEO and agentic commerce readin...

    1topic1scope
    note

    ARCHITECTURE

    Source Summary Architecture High-level overview of how the system fits together. Components - Remix app (app/) - Embedded Shopify admin UI (Polaris/App Bridg...

    1topic1scope
    note

    2026-02-04

    Source Summary Handoff — 2026-02-04 TL;DR - What changed today: Staging deployment on Render is now aligned with the “min fixed monthly cost” approach: Rende...

    1topic1scope
    note

    staging-e2e

    Source Summary Staging end-to-end checklist (Render + Neon + Upstash) Goal: validate the full loop works in a live environment: embedded UI → OAuth/session →...

    1topic1scope
    note

    seo-analyzer

    Source Summary / Deterministic SEO Analyzer Runs a set of deterministic checks on product data to produce a normalized SEO score and list of issues. / export...

    1topic1scope
    note

    ai-suggestions

    Source Summary import OpenAI from "openai"; import type { SEOAnalysis, SEOIssue } from ". /seo-analyzer"; import { DEFAULTPRODUCTOPTIMIZATIONMODEL } from "....

    1topic1scope
    note

    03-product-review

    Source Summary Al Product SEO Optimizer Dashboard Products Automation Activitylog Settings Read-only Intake fixture synthetic store data < Hand-Cut Paper Col...

    1topic1scope

    Visual evidence

    Key dates

    2026-02-02
    Experiment started

    Initial Shopify app, catalog, worker, and data-model implementation landed.

    2026-02-03
    Operational surface expanded

    Privacy webhooks, automation execution, billing hardening, and Render deployment configuration were added.

    2026-02-04
    Staging architecture stabilized

    The deployment settled on Render web/worker plus Neon and Upstash, while the handoff still identified core end-to-end product validation as the main unresolved risk.

    2026-02-06
    Last source commit

    The implementation window ended without a source-recorded completion or kill decision.

    2026-07-13
    Fresh killed-experiment intake

    Maggie classified the project as a killed Shopify app experiment; the source built in isolation and six read-only fixture screenshots documented its interaction model.