Automate e-commerce product listings at scale with descriptions, images and SEO metadata

Imagine launching a new product line of 500 items. You've got photography sorted, but now your team faces weeks of manual work: writing unique descriptions, optimising images for different platforms, creating SEO metadata, and uploading everything one by one. By the time you're halfway through, market conditions have shifted, your competitors have already launched, and your margin timeline has compressed. This scenario plays out across e-commerce businesses weekly, and it's entirely avoidable. The cost isn't just time; it's opportunity. Every hour spent manually editing images or rewriting descriptions is an hour not spent on strategy, customer research, or product sourcing. At scale, this bottleneck strangles growth. The good news: modern AI tools can eliminate nearly all of this work. With the right orchestration, you can feed product data into a workflow that generates descriptions, creates optimised images, and outputs SEO metadata in minutes rather than weeks. This Alchemy walks you through building an automated e-commerce workflow that takes raw product information and delivers a complete, ready-to-publish catalogue entry: professional images, SEO-optimised descriptions, and structured metadata, all with zero manual handoffs.

The Automated Workflow

The workflow chains four key steps: product data ingestion, image enhancement, description generation, and SEO metadata creation. We'll use n8n as the orchestration engine because it handles file uploads, conditional logic, and multi-step API calls without vendor lock-in.

How data flows

Your source is a CSV or spreadsheet containing product names, basic descriptions, raw images, categories, and prices. The workflow: 1. Reads each product row from your CSV.

2. Uploads the raw product image to AI Boost for enhancement, upscaling, and background optimisation.

3. Sends the product information to Claude Opus 4.6 to generate a detailed, SEO-aware description.

4. Generates structured metadata (titles, keywords, alt text) using GPT-4o mini.

5. Writes the complete product record (images, descriptions, metadata) back to your e-commerce database or exports to a new CSV.

Setting up n8n

Start by creating a new workflow in n8n. You'll need API keys for: - OpenAI (for description and metadata generation)

• Anthropic (for Claude Opus 4.6)
• AI Boost (for image enhancement)
• Your e-commerce platform (Shopify, WooCommerce, custom API)

Step 1: Ingest product data

Add a CSV file read node or connect directly to a Google Sheet. Map your columns: product_name, category, base_description, image_url, price.

Copy{ "product_name": "Wool Blend Overcoat", "category": "Outerwear", "base_description": "Winter coat, warm", "image_url": "https://your-cdn.com/raw-images/coat-001.jpg", "price": "149.99"
}

Step 2: Image enhancement with AI Boost

AI Boost accepts image URLs and can upscale, enhance, remove backgrounds, and adjust lighting. Use the HTTP Request node to call their API.

CopyPOST /api/v2/image/enhance
Content-Type: application/json
Authorization: Bearer YOUR_AI_BOOST_API_KEY { "image_url": "https://your-cdn.com/raw-images/coat-001.jpg", "operations": [ { "type": "upscale", "scale": 2 }, { "type": "background_remove" }, { "type": "enhance", "brightness": 10, "contrast": 5 } ], "output_format": "webp"
}

Capture the returned enhanced_image_url for use in later steps.

Step 3: Generate product description with Claude Opus 4.6

Use the Anthropic API to write compelling, detailed descriptions. Claude excels at understanding product context and generating human-sounding copy that performs well with both search engines and readers.

CopyPOST https://api.anthropic.com/v1/messages
Authorization: x-api-key YOUR_ANTHROPIC_API_KEY
Content-Type: application/json { "model": "claude-opus-4.6", "max_tokens": 500, "messages": [ { "role": "user", "content": "Write a detailed product description for an e-commerce listing. Product: Wool Blend Overcoat, Category: Outerwear, Base info: Winter coat, warm, Price: £149.99. Include material benefits, care instructions, and style guidance. Optimise for both readability and search engines. Keep under 200 words." } ]
}

Parse the response and store the generated description.

Step 4: Generate SEO metadata with GPT-4o mini

SEO metadata requires structured output: meta title, meta description, keywords, and image alt text. GPT-4o mini is cost-effective for this task since the input is straightforward and output format is predictable.

CopyPOST https://api.openai.com/v1/chat/completions
Authorization: Bearer YOUR_OPENAI_API_KEY
Content-Type: application/json { "model": "gpt-4o-mini", "messages": [ { "role": "user", "content": "Generate SEO metadata for this product. Name: Wool Blend Overcoat, Category: Outerwear, Description: [GENERATED_DESCRIPTION_FROM_CLAUDE]. Return JSON with: meta_title (max 60 chars), meta_description (max 155 chars), keywords (list of 5-8), image_alt_text (max 125 chars). Ensure keywords target mid-funnel search intent." } ], "response_format": { "type": "json_schema", "json_schema": { "name": "seo_metadata", "schema": { "type": "object", "properties": { "meta_title": {"type": "string"}, "meta_description": {"type": "string"}, "keywords": {"type": "array", "items": {"type": "string"}}, "image_alt_text": {"type": "string"} } } } }
}

Step 5: Output and save

Combine all outputs (enhanced image URL, description, metadata) into a final record and either: - Write directly to your e-commerce platform via API (Shopify REST, WooCommerce, custom endpoints).

• Export to a CSV file for review or batch import.

• Push to a database table for manual review before publishing. Here's an example final record structure:

Copy{ "product_name": "Wool Blend Overcoat", "category": "Outerwear", "price": "149.99", "description": "[CLAUDE_GENERATED_TEXT]", "enhanced_image_url": "https://ai-boost-cdn.com/enhanced/coat-001-upscaled.webp", "meta_title": "Premium Wool Blend Overcoat for Men | Warm Winter Coat", "meta_description": "Luxurious wool blend overcoat. Warm, durable, and timeless. Shop now.", "keywords": ["wool overcoat", "winter coat men", "warm outerwear", "premium coat"], "image_alt_text": "Grey wool blend overcoat, front view, displayed on model"
}

Add a final step in n8n to POST this to your e-commerce API or write to a results CSV.

Rate limiting and batching

Both OpenAI and Anthropic enforce rate limits. Set n8n to process products in batches of 10 per minute. Add a wait node between API calls to avoid throttling.

Copy{ "type": "Wait", "duration": "6 seconds"
}

AI Boost image processing is asynchronous; use polling or webhooks to check when enhanced images are ready before proceeding to the next step.

The Manual Alternative

If you prefer human oversight at certain stages, insert approval nodes into your n8n workflow. After generating descriptions, save them to a spreadsheet with a "Review Status" column. Team members can edit and mark as approved, then the workflow resumes, uploading only approved items. This preserves quality control without manual image editing or metadata entry.

Pro Tips

Error handling and image failures.

Not all product images upscale equally.

AI Boost may reject low-resolution or complex images. Set up fallback logic: if upscaling fails, use the original image or trigger a Slack notification for manual intervention. Log failures to a separate error CSV so you can revisit problem images.

Cost per product.

At scale, costs add up. Use GPT-4o mini for metadata (cheaper) and Claude Opus 4.6 only for descriptions where nuance matters. Batch API calls where possible. AI Boost charges per operation; combining upscale and background removal in one request is more efficient than two separate calls.

Review before publishing.

Generate 50 products first, manually inspect them, and iterate your prompts. Small improvements to the Claude and GPT prompts yield dramatic quality gains. Once you're satisfied, run the full catalogue.

Test with a subset.

Run the workflow on 20 products before launching against your entire catalogue. This catches integration issues, API authentication problems, and format mismatches without wasting quota or generating unusable data.

Image sizing for multiple platforms.

After enhancement, resize the image for different channels (thumbnail, mobile, desktop). Add Pixelcut AI to your workflow for one-click platform-specific optimisation. This ensures consistency across Shopify, Instagram, and your website without additional manual steps.

Cost Breakdown