Conversion optimization with AI chat
Forget persuasion scripts. What actually moves conversion in chat is whether the AI can take a shopper from a vague need to a specific, purchasable recommendation with as little friction as possible — and whether you can measure that against real orders afterward. Here's how ShopGuide's real features map to that.
Guided Discovery turns browsing into a decision
When a shopper describes a goal instead of naming a product — "something for downhill riding," "help me pick a gift" — GUIDED_DISCOVERY kicks in instead of handing back a generic list. The AI asks up to three short questions, three options each, then searches the catalog using the answers. That converts an open-ended browsing session into a narrowed recommendation, right at the point where undecided shoppers usually give up and leave.
How well this works depends entirely on your catalog data. If your descriptions don't capture the attributes that would actually change a recommendation — skill level, use case, size category — the questions can't narrow anything meaningfully. See Chat optimization for what to check first.
A discount code that closes, not a banner that's always there
Turn on the AI Discount Code Generator (Settings → AI Behavior) and the AI can build a personalized, time-limited Shopify discount mid-conversation — something like SARAH-LASHES-A3X, drawn from context in the chat plus a random suffix for uniqueness. It's scoped in Shopify to whatever products or collections you've marked eligible, so it's a real, enforced discount rather than text the AI is making up. The system prompt tells it flatly never to invent a code without calling the underlying tool — so anything shown in chat traces back to an actual Shopify discount record.
That's a genuinely different mechanism from a sitewide banner: contextual, offered when it's relevant rather than to everyone, and personalized rather than generic. It only works, though, if you keep the eligible list current — see Common mistakes.
Recognizing when someone's ready to buy
ADD_TO_CART and CHECKOUT are handled as their own intents, separate from general product questions, each triggering a different tool set. A shopper who says "add this to my cart" gets a cart action, not another round of product details. That distinction — browsing vs. comparing vs. ready to buy — gets made by the intent router before the AI even starts generating a reply, rather than left for the AI to infer mid-answer.
Measuring what actually converts
ShopGuide's A/B testing doesn't stop at engagement — it attributes completed Shopify orders back to each variant, so a launch message, prompt, or discount configuration gets compared against real revenue rather than a proxy like click-through rate. Set one up under A/B testing, read results in A/B test results, and give it a full business cycle before acting — early reads are unreliable; see Common mistakes for why.
Chat-attributed revenue itself works by linking a completed order back to a shopper's chat thread whenever a purchase follows a chat interaction. That's the number that shows up as chat orders and chat revenue on your dashboard. For how to read those figures without treating them as a generic "conversion lift" percentage, see Success metrics.