OVERVIEW

AI Image Recognition automatically reviews shopper uploaded photos during a return and provides structured assessments of the item’s condition. These assessments help your team quickly identify damage, determine resell eligibility, and reduce time spent manually reviewing images.

When a shopper uploads photos in the return flow, each image is analyzed by Loop’s AI Image analysis service. The model produces a set of standardized outputs for every photo:

• Description

  A short summary of visible damage.

  If no damage is detected, the description will be “no-damage.”

• Item condition

  One of the following labels: in-package, perfect, good, fair, poor, damaged

• Has tags

  Yes/No - whether tags or stickers appear attached.

• Is damaged

  Yes/No - whether the item appears to have visible damage.

• In package

  Yes/No - whether the item appears inside its original packaging.

• Is correct package/product

  Yes/No - whether the item appears to be the correct product.

These outputs appear in the Return Details page when hovering over the ✨ icon positioned top left of each uploaded photo.

To generate accurate assessments, the system uses:

• Return details (ID, product IDs, SKU, return reason text)

• Product data from Shopify

• Workflow-specific rules (e.g., required tag photos)

These help the model contextualize the uploaded images and evaluate them more accurately.

No setup is required.

AI Image Protection is enabled for all Loop merchants and works automatically on any return with photo uploads.

In the Return Details page, merchants will see:

• Description (e.g., “no-damage”, “dirt on outsole”)

• Item condition

• Yes/No indicators for:

  • Tags attached

  • Visible damage

  • Correct product

  • In original packaging

• Example “no-damage” output

• Example “damaged” output and incorrect package/product flag

• If the image is unclear or the model cannot assess confidently, the merchant should manually review.

• Merchants may override or disregard AI outputs at any time.

• The AI evaluates whatever the shopper submits — image quality enforcement is not handled by the model.

• No changes to shopper experience.

• Customers upload images according to your existing workflow rules.

• The portal does not evaluate or enforce photo quality.

What assessments does the AI provide? Damage description, item condition label, presence of tags, visible damage indicator, packaging indicator, and correct product indicator

Can the AI confirm exact product identity? It provides an estimation based on the image. If the product is unclear or partially shown, merchants should manually validate.

What if the photo is low quality? The model still attempts an assessment, but unclear or blurry images should be manually reviewed.

Does the AI replace manual review entirely? No. It can significantly reduce manual workload, but merchants maintain final judgment.

How does this differ from fraud tools? Image analysis and Fraud Tools are separate features. We don't recommend using image analysis alone for fraud prevention; merchants concerned about fraud should use our existing fraud tools service.

For additional questions, please reach out to support@loopreturns.com.