Skip to content

What the AI seller learns from

Swishy’s AI seller sounds convincing on your PDP because it’s grounded in your data. It doesn’t guess. Every answer it gives is backed by something it has indexed from one of the sources below.

  • Product titles, descriptions, tags, vendor, product type
  • Variants, SKUs, prices, inventory
  • Product images and alt text
  • Collections and metafields

Synced in real time via Shopify webhooks. When you update a product, the AI knows within seconds.

  • Transcripts of reels and videos
  • Visual analysis of clips and photos
  • Captions and comments from imported posts
  • Written reviews pulled from connected review apps

Only content that is approved in the review queue and has creator consent is used. See Consent.

  • Shipping and returns policies (pulled from your Shopify settings)
  • Size guides and product-specific FAQs you upload
  • Custom knowledge you add in AI Seller → Knowledge
  • Prior chats on your own store (used to improve relevance, not shared with other merchants)
  • Patterns in what shoppers ask and what closes sales
  • Other merchants’ data. Your catalog and UGC are isolated. Nothing you feed Swishy trains a shared model.
  • Unapproved UGC. Content pending review or pending rights is invisible to the seller.
  • Personally identifiable shopper data beyond what’s needed to complete the conversation.
  • Retrieval-augmented generation — for every shopper message, the AI retrieves the most relevant catalog items and UGC snippets before answering.
  • Citation-aware — the AI can cite which review or clip it’s paraphrasing.
  • Guardrails — if it doesn’t have the information, it says so and offers to connect the shopper with your team.

In AI Seller → Settings you control:

  • Tone and voice — presets plus custom brand voice prompt.
  • Topics to avoid — e.g. medical claims, competitor names.
  • Escalation triggers — returns, pricing negotiation, multi-order issues.
  • UGC preferences — prefer video over images, cap clip length, etc.