As of 13 August 2026, AI can only partly detect product defects from photos.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open your current quality-control specification, inspection checklist and product acceptance rules, and copy the relevant pass, fail and acceptable-variation criteria into a working document.
   2. Gather representative photographs for each item or batch, including an overall view, close-ups of the suspected defect and any required angle, label or component view; record the item or batch reference for every image.
   3. Add examples of known acceptable products and known defects where you have them, removing customer or employee personal data from images before uploading.
   4. Paste the prompt into an image-capable chatbot, replace each bracketed section with your product information and rules, then attach the labelled photographs.
   5. Check every reported observation against the original photograph and compare each possible defect with the copied inspection rule, marking uncertain cases for a trained quality-control colleague.
   6. Take the additional photographs or physical measurements requested by the model, then have the colleague make the final accept, reject, rework or escalation decision and record it in your normal quality system.

   Prompt

   You are assisting with a quality-control inspection of a product from photographs. Do not claim that a defect exists unless it is visible in the supplied images. Do not infer hidden, internal, electrical, chemical, safety or dimensional defects from appearance alone.
   
   Product: [product name and model]
   Inspection standard or specification: [paste the relevant pass or fail rules]
   Known acceptable variation: [describe what is allowed]
   Known defect examples: [describe or attach examples if available]
   Photos: [attach clear, well-lit photographs and identify the product or batch]
   
   Inspect each image and return a table with these columns: image or item reference, observed feature, possible defect, confidence as high, medium or low, applicable rule, and recommended human check. Separate visible observations from inferences. If the image is blurred, poorly lit, obstructed or taken from an unsuitable angle, say exactly what cannot be assessed and request the specific additional photograph or measurement needed. Do not give a final release decision. End with a short list of items that a trained quality-control colleague must inspect in person before the product is accepted, rejected or shipped.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.