Chat optimization playbook
Chat quality in ShopGuide has less to do with wordsmithing a welcome message and more to do with three things you actually control: how completely your catalog is indexed, whether your product data matches how customers phrase things, and how quickly you act on what the AI couldn't resolve. This is about ShopGuide's real architecture, not generic chatbot advice.
How ShopGuide decides what to say
Every incoming message gets classified into an intent — PRODUCT_SEARCH, PRODUCT_QUESTION, PRODUCT_COMPARISON, GUIDED_DISCOVERY, BROWSE_COLLECTIONS, INFORMATION, ORDER_MANAGEMENT, SUPPORT, ADD_TO_CART, CHECKOUT, SPECIAL, CLARIFICATION, plus DISCOUNT_CODE if you've turned that feature on — before the AI writes a word. Each one activates a different tool set and prompt section, not just a different tone.
What that means in practice: if customers describe products using phrasing your catalog doesn't contain, they land in BROWSE_COLLECTIONS or CLARIFICATION instead of PRODUCT_SEARCH, and get a generic category response instead of a direct answer. The fix isn't a better welcome message. It's tightening product titles, collection handles, and descriptions to match how shoppers actually talk about what you sell.
Catalog coverage is the real lever here
Product search and product questions run on Qdrant vector search over your catalog plus RAG over your shop's knowledge content — semantic matching, not keyword matching. A shopper typing "warm winter hat" should land on a product titled "Merino Wool Beanie" because the match is on meaning.
Which also means the AI can only recommend or answer questions about what's actually indexed. If a product line isn't showing up in results, check sync and content completeness — full descriptions, current variants, populated metafields — before touching prompt wording. An unindexed or thinly-described product won't surface no matter how you instruct the AI.
When Guided Discovery earns its keep
GUIDED_DISCOVERY covers vague-need requests — "help me choose," "what do I need for X" — where a keyword search would return too broad a list to be useful. Instead of searching right away, the AI asks up to three short questions, three options each, shown as an in-chat overlay, then searches using the answers.
This flow is only as good as the distinctions your catalog supports. If two products differ in ways your descriptions don't capture — skill level, use case, fit, material — the follow-up questions can't meaningfully change the outcome. If these conversations keep ending in generic browsing rather than a specific pick, that's a catalog-detail problem, not a prompting one.
Closing the loop with Information Gaps
Every response gets checked afterward by Information Gaps Detection, which sorts anything the AI couldn't fully resolve into knowledge_gap (your content doesn't cover it), bug (something's broken in your store), live_data (needs real-time info the AI has no tool for), system_error (the app itself misbehaved), or other. Recurring phrasings of the same question get stacked into one row instead of fragmenting into duplicates, so the count you see is a real frequency signal.
Check the Knowledge Gaps tab regularly — see Chat management. It's the most direct answer to "what is my AI actually failing at," and a shop-specific spike in system_error is worth treating as a regression to fix, not a content gap to write around.
Test, don't guess
Rather than assuming which welcome message or prompt variant performs better, run it as an A/B test. ShopGuide splits shoppers into variants and attributes both engagement and completed orders back to each one, so you're comparing real outcomes instead of a proxy engagement number. Set one up under A/B testing and read results via A/B test results — see Common mistakes for how long to let a test run before acting on it.
Language coverage
The widget's static chrome — buttons, guided-discovery labels, footer text, disclaimer — ships in English, German, and Dutch only. The AI's own replies aren't limited to those three: it detects and responds in whatever language the shopper is using, regardless of the UI setting. Serve a market outside en/de/nl, and expect fluent AI prose next to English buttons and labels. See Multi-language support for what's configurable and what isn't.