The Shopify AI Product Description Factory is a production-grade n8n workflow that converts product images and metadata into refined, SEO-aware descriptions—fully automated and region-agnostic. It blends GPT-4o vision for visible attribute extraction, Claude 3.5 Sonnet for premium copy, Perplexity research for verified brand context, Google Sheets for orchestration and audit trails, plus automated daily sales analytics enrichment. Link-header pagination and structured output enforcement ensure reliable scale.
To refine according to your usecase connect via my profile
@connect
Uses gpt-4o to identify only visible physical attributes (closure, heel, materials, sole) from product images—no guesses.
anthropic/claude-3.5-sonnet crafts concise, benefit-led descriptions with consistent tone, length control, and clean formatting.
perplexityTool verifies vendor/brand context from official sources to avoid speculation or fabricated claims.
Automates Shopify REST pagination via Link headers and persists page_info
for resumable runs.
Centralized staging, status tracking, and QA in Products, with ProcessingState for batch/page markers, and Error_log for diagnostics.
errorTrigger + AI diagnosis logs clear, non-technical and technical explanations to Error_log for fast recovery.
Daily sales tracking automatically captures and enriches total sales data for comprehensive business intelligence and performance monitoring.
httpRequest
fetches /admin/api/2024-04/products.json?limit=200&{page_info}
code
filters only items with:
body_html
currSeas:SS2025
tagx-styleCode
, country_of_origin
, and gender
when availablecode
parses Link headers for rel="next"
and extracts page_info
googleSheets
updates ProcessingState with page_info_next
and increments the batch number for resumable pollinggoogleSheets
pulls rows with Status = Ready for AI Description; limit throttles batch sizeopenAi
Analyze image (model gpt-4o) returns strictly visible featureslmChatOpenRouter
(Claude 3.5) composes the SEO description, optionally blending verified vendor context from perplexityTool
outputParserStructured
guarantees strict JSON: product_id
, product_title
(normalized), generated_description
, status
googleSheets
writes results back to Products for review/publishhttpRequest
fetches paid orders from Shopify REST API with date range filteringsplitOut
and summarize
nodes calculate total daily saleserrorTrigger
routes failures to an AI agent that explains the root cause and appends a concise note to Error_log.
httpRequest
(Shopify REST 2024-04 for products and orders)googleSheets
(multiple sheet operations)googleSheetsTool
(error logging)openAi
(gpt-4o vision analysis)lmChatOpenRouter
(anthropic/claude-3.5-sonnet for content generation)splitOut
(order data processing)summarize
(sales totals calculation)set
nodes (data field mapping)perplexityTool
(brand research)outputParserStructured
(JSON validation)memoryBufferWindow
(conversation context)scheduleTrigger
(multiple time-based triggers)cron
(periodic execution)limit
(batch size control)if
(conditional logic)code
(custom filtering and pagination logic)errorTrigger
+ AI diagnosis to Error_logx-styleCode
is presentgender
tag is presentpage_info_next
Auto-paginate Shopify; stage eligible products in Sheets with reasons and timestamps.
Descriptions reflect only visible attributes plus verified brand context; concise, mobile-friendly structure with gender-aware tone.
Auto-injects x-styleCode
, country_of_origin
, and gender
when present; natural SEO for brand and product type.
Automated daily sales tracking with Melbourne timezone support, handles zero-sale days, and maintains complete historical records.
Layman + technical diagnosis logged to Error_log to shorten MTTR.
Structured JSON via outputParserStructured
for predictable row updates.
page_info
persistence in ProcessingStatememoryBufferWindow
) to keep multi-step prompts consistentThe workflow includes two independent schedules:
For globally distributed teams, schedule triggers and timestamps can be standardized on UTC to avoid regional drift.
Start with small batches (limit
set to 10 or fewer) to validate both copy generation and sales tracking flows. The workflow handles dual operations independently - content generation failures won't affect sales analytics and vice versa. Monitor the Error_log sheet for any issues and use the ProcessingState sheet to track pagination progress.