02Case studyAmerican Eagle
American Eagle. When AI told shoppers the jeans cost $42.
The Strigid Ripped Mom Jean retails at $75. Three months after a sitewide promo ended, ChatGPT, Perplexity and Gemini were still quoting $27.96 to $41.96 to shoppers. Shopthru.ai caught it the first time it ran.
- Industry
- Apparel
- Headline outcome
- $42 vs $75
Measured catalog truth, query by query.
The team ran 1,161 LLM executions across OpenAI, Perplexity and Gemini, then diffed every quoted product fact against the live AE catalog. 800 of 1,411 ground-truth checks (57%) came back as a mismatch. 625 of those were pricing.
- 625price mismatches caught
- 57%of all GT checks were wrong
- $33avg price gap (AI says vs actual)
Simulated the cost: every wrong quote modelled as a lost-cart event.
A shopper who hears "$42" and shows up to find $75 abandons. The team modelled the gap on the AE Strigid Mom Jean and the AE Next Level High-Waisted Jegging (both jeans appearing in ~30 high-intent purchase queries each) and ranked them P1 for the catalog team.
Shipped the fix list: 206 product CONTENT_FIX tickets, with the diff.
The team opened 206 product CONTENT_FIX items, 23 TRUST_SIGNAL items, and 12 GT_TECHNICAL_READINESS items. Each ticket carries the live URL, the wrong AI quote, the correct value, and the recommended schema/feed change.
- 206 product-level CONTENT_FIX items with diff payloads
- 175 availability mismatches (also flagged for checkout fixes)
- 14 FEATURE_HIGHLIGHT items where AI missed a current discount
Compounded into brand memory: the "promo-as-live" pattern now flags automatically.
AE was the first audit to surface the "AI quotes ended-promo prices as live prices" pattern at scale. The team now warns when any quoted price sits more than 15% below the live feed. The rule fires on every catalog audit, regardless of vertical.