E-commerce
June 28, 2026
The customer arrives at /collections/summer-dresses. One hundred and forty products, twelve filters, no clicks for one minute. They do not know whether they should filter by size, open a sub-collection, or compare two similar models.
A generic "How can I help?" chatbot is not enough. On a collection page, the assistant must know the current collection, active filters, and the visible grid to guide between categories, filters, and product pages.
This guide #196 covers the AI chatbot on collection pages: contextualized discovery flows, deep link filters, short comparisons, and PLP triggers. Distinct from collection CRO optimization (#23), large catalog assistant (#110), and the comparison chatbot vs search (#14).
Summary
Why deploy a chatbot on collection pages rather than globally?
The collection page (PLP) is the moment when the intent is semi-structured: more precise than the homepage, hazier than a product page.
PLP Friction Signals
Scroll without click: long journey, 0 ATC
Combined filters with 0 results: immediate abandonment
Collection bounce: SEO or ads entry, quick exit
Repeated questions: "what is the difference between these two models?"
Fast Simon analyzes nearly 50,000 shoppers in 2026: interactive discovery dialogues drive conversion to 15-22% in fashion and footwear, nearly triple that of keyword search alone (Fast Simon, 2026 Shopper AI Agents).
Collection Page Specificity
The PLP bot starts from the displayed collection, not the entire catalog. It suggests sub-collections, relevant filters, and 2-3 products from the grid, without re-imposing a global funnel of 800 SKUs.
How does it differ from the CRO collection and the catalog assistant?
Four related contents, four roles.
Collection optimization (#23)
Shopify collection CRO: layout, native filters, product cards, sorting, mobile. #196 adds the contextualized conversational layer on top of the existing PLP.
Large catalog assistant (#110)
Catalog assistant (#110) navigates 500+ SKUs across all categories. #196 remains anchored on /collections/X: jackets, face creams, dining chairs.
Conversion filters (#30)
Filters guide (#30) configures the facet. #196 teaches the customer which filter to use and generates deep links.
Chatbot vs search (#14)
Comparison (#14) routes search vs bot. #196 details bot execution on PLP with live collection context.
Which customer intents should be prioritized on a collection?
Map PLP intents from analytics and chats for 60 days.
Top 8 collection intents
plp_narrow: "too much choice, help me narrow it down"
plp_subcategory: "I want a midi dress, not a mini"
plp_filter_help: "which filter for sensitive skin?"
plp_compare: "difference between model A and B?"
plp_budget: "under €80 fast delivery"
plp_zero_result: active filters → 0 products
plp_gift: "gift in this range"
plp_stock: "available in L this week?"
Prioritization
Address plp_narrow, plp_filter_help, and plp_zero_result first: high volume, quick resolution via filter deep link or sub-collection. Cross-reference with segment purchase intent.
What context should the bot read on the collection page?
A contextualized collection chatbot without PLP context asks the same questions as a homepage bot.
Minimum data (Shopify sync)
Collection handle + title: summer-dresses, Waterproof jackets
Number of products: total and after URL filters
Active filters: size M, color black, price 50-100
Linked sub-collections: menu navigation and tags
Top 12 visible products: titles, price, stock, reviews
Available facets: Search & Discovery metafields
Shopify Search & Discovery exposes filters on options, standard fields, and metafields (Shopify, filters). The bot maps customer language ("dry skin") to the business facet ("skin_type:dry").
Contextualized opening message
"You are browsing our Summer Dresses (142 models). Are you looking for a specific occasion, size, or budget?" No generic greetings.
How to guide to the right sub-collection or category?
The plp_subcategory flow avoids filtering 400 products when a child collection exists.
4-step workflow
Customer expresses a need: "summer wedding guest dress"
Bot detects a mismatch with the broad "Dresses" collection
Suggests 2-3 sub-collections: Midi dresses, Guest dresses, Linen dresses
Direct link + summary: "34 models in Guest dresses, average budget €89"
Business rules
Never redirect outside the parent collection without explaining
Display the number of products per sub-collection
If only one sub-collection matches: direct redirect
DTC fashion example
"Running shoes" collection (218 SKUs). Customer: "muddy trail, beginner". Bot routes to /collections/trail-running-debutant (41 SKUs) instead of applying 5 filters.
How to help the customer choose and apply the right filters?
The plp_filter_help flow transforms the shop taxonomy into customer language.
Typical dialogue
Customer: "I don't see my budget". Bot reads active filters (no price). "What is your max budget?" Quick reply: < €60, €60-100, €100-150. Generates URL: /collections/cremes-visage?filter.p.price=0-100 and confirms: "28 creams under €100, sorted by bestsellers."
Zero-result rescue
0-product grid after filters: bot analyzes dead combination. "Size L + Navy blue: 0 results. I suggest L + Blue (8 models) or M + Navy blue (3 models)." Baymard points out that poorly filtered product lists lead to abandoned comparisons (Baymard, product list UX).
Filter guide link
Completes optimize filters (#30) without duplicating the facet config.
How do I compare two or three products from the grid in the chat?
The plp_compare flow addresses product hesitation among visible products.
Practical method
Customer mentions names or clicks "Compare" on in-chat cards
Bot pulls metafieldables: material, weight, battery life, review rating
Table with 3 columns max: attribute | Product A | Product B
Recommendation: "For a 3-day hike, A (650 g) vs B (420 g): B is lighter"
CTA: winning PDP link + "See both in detail"
Bot limitations
Compare 2-3 SKUs from the current collection, not 15 products. Beyond that: redirect to a pre-filtered collection or guided selling. Do not invent specs: source = Shopify catalog only.
In-chat product cards
Image, price, 1 differentiating line, "See product sheet" button. Gorgias Shopping Assistant presents recommendations based on browsing and cart (Gorgias, Shopping Assistant).
Which proactive triggers should be activated on a collection page?
The proactive PLP bot triggers on objective signals, not a second-2 popup.
Recommended triggers
Inactivity of 45-60 s without a product card click
Zero-result filters: URL or DOM detection
Scroll > 70% without click: discreet chip "Refine your choice?"
Back button from PDP back to same collection
Collection > 80 SKUs + first visit of the session
Interlocks
No proactivity if a conversation is already open. No full-screen mobile popup masking filters. A maximum of one intervention per collection / session. See contextual help vs pop-ups.
Widget placement
Sticky bottom right on mobile; discreet banner under the desktop collection title "Need help choosing in this range?". Aligned with collection CRO (#23) without pushing back the product grid.
How do you orchestrate bots, native filters, and internal search?
The collection bot complements the discovery stack; it does not replace it.
Bidirectional handoffs
Search 0 result on collection → plp_narrow bot
Bot shortlist → filtered collection deep link
Bot compare → PDP + return to collection with filters preserved
Exact reference intent → redirect to search bar
Recommended hybrid stack
Search for known SKUs. Filters for expert refinement. Collection bot for vague intent on PLP. ScaleWise notes that the conversational product finder reduces "filter fatigue" when the facet menu exceeds 8 criteria (ScaleWise, product finder 2026).
GA4 Events
plp_bot_open, plp_bot_filter_link, plp_bot_pdp_click, plp_bot_subcollection. Collection attribution in Looker report.
Which KPIs should be measured specifically on collection pages?
Bot collection KPIs differ from global bot metrics.
PLP Engagement
PLP bot open rate: opens / collection sessions
PLP Intent resolution: deep link or PDP without human handoff
Zero-rescue rate: % zero-results converted into alternative filters
Conversion
PLP CTR lift: product card click with vs without bot (A/B 21 days)
Assist-to-ATC collection: target 20-30% of PLP bot sessions
Collection Bounce: before-after bot delta
Quality
Repeat visit to the same collection without a purchase. "I can't find it" tickets tagged as collection. Complements chatbot KPIs (#11) with a dedicated PLP view.
How does Qstomy guide users on Shopify collection pages?
Qstomy reads the live PLP context and executes the plp_subcategory, plp_filter_help, and plp_compare flows.
Collection features
Collection context API: handle, count, URL filters
Deep link filters: Search & Discovery URL generation
Sub-collections: routing menu + Shopify tags
Compare in-chat: 2-3 products visible in the grid
PLP triggers: zero-result, inactivity, large catalog
Quantified DTC scenario
Cosmetics, "Facial care" collection of 186 SKUs, PLP bounce rate 52%, product card CTR 8.4%, 31 tickets/month on "which product to choose". Deployment of Qstomy PLP context + narrow/filter/compare flows on top 3 traffic collections. After 9 weeks of A/B testing: PLP bounce -19 points, product card CTR +24%, assist-to-ATC 27%, zero-rescue 68% of dead filters, discovery tickets -41%.
Explore AI sales agent, Shopify integration, request a demo.
Which operational playbooks should you deploy on your collections?
Playbook 1: map PLP intents (2 h)
Export chats + zero-result search on /collections/ URLs. Top 5 section 3 intents per priority collection.
Playbook 2: bot context config (3 h)
Sync collection handle, filters, top products. Section 4 contextualized opening message. Test 5 collections.
Playbook 3: narrow + filter flows (4 h)
Draft sections 5-6 for 1 vertical (fashion or beauty). 10 deep link tests for filters + sub-collections.
Playbook 4: PLP triggers (1 h)
Activate zero-result + 60 s inactivity. Section 8 interlocks. Verify mobile does not hide filters.
Playbook 5: A/B W+3 (ongoing)
Section 10 KPIs on traffic collection #1. Adjust flow for intent #1 by volume.
Useful links
A well-optimized collection page converts. A chatbot that knows this collection transforms passive scrolling into a guided choice between the right sub-category, the right filter, and the right product.
Enzo
June 28, 2026
