Part 3: Using AI to make seller-generated product listings discoverable
Originally posted on LinkedIn in Jul’25
AI in Product Management workflows. Not Someday. Today.
Part 3: Making seller-generated product listings discoverable
In Part 2, I shared how AI can catch broken taxonomy before it starts hurting UX.
This time, I’m looking at a different kind of discovery problem: when product listings don’t reach the right audience.
I listed a wrap-around summer skirt on a secondhand marketplace. The photos were fine. The title was decent. But the listing barely got any views.
People were probably searching for “boho wrap skirt”, “beachwear midi”, or “flowy cotton skirt”. The system didn’t know those were close enough, so my listing stayed invisible.
Here’s how I’d solve that as a #PM using AI inside an internal tool:
1. Smart tagging assistant for sellers
- As the seller types in a title and description, AI suggests richer, trend-aware tags
- Built using OpenAI’s Function Calling API
- Auto-extracts:
• Style keywords: boho, vacation-ready
• Garment types: wrap skirt, A-line midi
• Occasion tags: resortwear, brunch outfit
- Seller gets a quick “Add suggested tags” button before hitting publish
2. Bulk metadata enrichment for internal teams
- CX or merchandising teams drop in underperforming listings
- GPT-4 or Claude flags:
• Metadata gaps
• Misaligned tags
• Broken discovery paths
- Outputs flow into your CMS or internal bulk editor
3. In-context seller education
- Show top-performing tags from similar listings
- Offer Gen AI-powered listing tips when input quality is low
- Optional: “Boost my visibility” toggle that runs suggestions before upload
No new UI needed. Just better inputs, better matching, better discovery.
(And yes, this was the skirt. A very cute wrap-around that deserved more love than 3 views and zero likes.)
Part 4 drops soon.
#GenAIProductManagement #MarketplaceUX #PeerToPeerProducts #ProductOps #AppliedAI #PMThinking
