Audit
Per-channel readiness score
Every product gets a separate readiness score for each AI shopping channel — ChatGPT (ACP spec), Google AI Mode, Microsoft Copilot (Bing Shopping), Perplexity (UCP). A product can be 95% ready for ChatGPT but only 60% ready for AI Mode because the field requirements differ slightly.
Example — Product X — ChatGPT: 94% · AI Mode: 78% · Copilot: 67% · Perplexity: 89%. AI Mode flag: title is too short.
Audit
ACP/UCP feed-spec compliance
Checks every product field against the published Agentic Commerce Protocol and Universal Commerce Protocol feed specs — title length, description length, GTIN/MPN/brand presence, image dimensions and alt text, structured Product schema, category mapping, condition, availability.
Example — Product missing GTIN. Recommendation: add via metafield or product variants tab. Pro: AI rewrite can suggest based on similar products.
Audit
Read-only Free tier
Free-plan audits are 100% read-only. We read your catalog state, run checks, produce a report — nothing is modified, no scopes granted to write. Free tier is the right starting point if you want to see your AI shopping readiness before committing to changes.
Example — Install free → audit runs → you see Health Score and prioritized issues → nothing in your catalog changed.
AI Rewrites (Basic)
Brand-voice product copy rewrites
Basic plan uses Moonshot AI's Kimi K2.6 model to rewrite failing fields. The voice extraction step pulls your existing top-performing product copy as training exemplars, so rewrites match your tone — formal vs casual, feature-led vs benefit-led, short vs long sentences. Field-scoped: we only rewrite what's failing, not what's already good.
Example — Failing field: description (too short, 67 chars). Rewrite: 240-char description in your brand tone, includes 'wool blend', 'tailored', 'Italian-made'. Reversible.
AI Rewrites (Basic)
Voice profile training
You teach GEO your brand voice by selecting 5-10 exemplar product descriptions you love. The model uses those as a few-shot prompt for every rewrite, so the output reads like you, not like a generic AI assistant.
Example — Pick your 5 best-performing product descriptions → GEO trains a voice profile → next rewrite reads like you wrote it.
AI Rewrites (Basic)
Autopilot — scheduled audits + auto-rewrites
Autopilot runs scheduled audits (daily by default) and automatically rewrites any product that drops below your target score. You don't need to remember to come back and re-audit when you add new products. Pausable anytime in Settings.
Example — You add a new product → next day autopilot scans → if score < target, AI rewrites are applied → logged to Work Ledger.
Safety
Work Ledger reversibility
Every AI rewrite is logged to the Work Ledger with the prior field value. Click 'Revert' on any entry and the prior value is written back via Shopify's GraphQL Admin API. No state lost; full history preserved forever.
Example — Autopilot rewrote product description yesterday. You don't love it. Open Work Ledger → Revert → prior description restored.
Safety
Field-scoped writes only
GEO only writes to product fields it's been instructed to rewrite. We never touch variants, prices, inventory, or any field outside the rewrite scope. The write_products scope is used exclusively for description/title/metafield writes.
Example — Rewrite scope: description only → GEO updates description, leaves price, inventory, variants, tags, collections untouched.
Insights
Channel-level performance trends
Track how your per-channel readiness score changes over time. See whether a recent catalog refresh improved or hurt your AI shopping readiness, and which channels need attention.
Example — Last 30 days: ChatGPT readiness +12pts (catalog cleanup paid off). AI Mode readiness -3pts (new products added without GTINs).
Insights
Issue prioritization by impact
Issues are sorted by impact — how many products fail this check × how heavily this check weighs in the per-channel score. Fix the top 3 issues and your channel readiness usually jumps 10-15 points.
Example — Top issue: 47 products missing GTIN → fixing all 47 raises ChatGPT score by 14 points.