Generate and transform content with AI data services
Use Productsup AI data services to generate and transform product content — titles, descriptions, product types, and SEO text — plus translate feeds and edit product images.
Productsup AI data services use AI to generate, rewrite, translate, and enrich your product data automatically as it comes into the platform. Instead of writing content by hand or maintaining it per product, you describe what you want, and Productsup sends the relevant attributes to an AI provider and imports the result back into your feed.
Common jobs you can hand to AI include generating SEO-friendly product titles, writing product descriptions, classifying products into a product type or Google product category, generating SEO keywords or product highlights, shortening or lengthening existing text, translating a feed into another language, and removing product-image backgrounds.
Choose the right AI data service
| I want to… | Use this data service |
|---|---|
| Generate a product title from existing attributes | Custom Prompt With Dynamic Values |
| Generate a product description | Custom Prompt With Dynamic Values |
| Generate a product type or Google product category | Custom Prompt With Dynamic Values |
| Generate SEO keywords or product highlights | Custom Prompt With Dynamic Values |
| Use Anthropic's Claude models for any of the above | Use Claude with the data service Custom Prompt With Dynamic Values |
| Shorten, lengthen, or create text from existing attributes | Shorten, lengthen, or create text from attributes with AI data services |
| Translate a feed into another language | Translate your feed with the Google Translate and Google Translate (Advanced) data services |
| Remove the background from product images | Remove image background with the Image Background Removal service |
Use Custom Prompt With Dynamic Values whenever you want AI to write something new from your attributes. Use Shorten, lengthen, or create text when you already have text and only need to reshape its length or tone.
How AI data services work
All AI data services share the same basic behavior, so the details below apply across the services listed above unless an individual article notes an exception.
Where they run. AI data services run at the import or intermediate stage, before your data reaches mapping and export.
Which AI provider you use. Depending on the service, you can connect to OpenAI, Microsoft Copilot (Azure OpenAI), or Anthropic's Claude. You provide your own provider account and API key, and you accept the provider's terms during setup.
Where the output goes. Each service writes its result into a new attribute whose name starts with three underscores, for example ___service_ai_custom_prompt. Attributes starting with ___ are working attributes: the platform doesn't send them to your export channels unless you map their values to an exported attribute yourself. This lets you review AI output before it reaches any channel.
Caching and cost. Productsup caches the AI result per product, so identical inputs aren't re-sent to the provider on every run. This keeps output stable between runs and controls cost. The cache is tied to the exact model, context, prompt, and the values of the attributes you reference. Changing any of these — even by one character — invalidates the cache and regenerates the affected products on the next run. Productsup's integration is free; the AI provider may charge for usage based on their own pricing.
What you need. A provider account and API key, and a unique item identifier set on your site.
Best practices
Test on a small sample first. Before running an AI service across your full catalog, confirm the output on a small set of products. Large catalogs can take a long time and consume provider tokens, so validate the prompt and output format before scaling up.
Generate at parent level and inherit to variants. If your feed has variant products (size, color) sharing a parent, run the AI service only for parent SKUs and propagate the result to variants by matching on Item Group ID. This cuts token cost dramatically and keeps every variant of the same product consistent, instead of generating a slightly different value per variant.
Map AI output to export deliberately. AI results stay in ___-prefixed working attributes so nothing reaches a channel by accident. Review the values in Data View, then map them to the attribute you actually export.
For prompt-writing tips specific to Custom Prompt With Dynamic Values — such as writing precise instructions, keeping variant attributes consistent, and avoiding conditions that lose context after attribute substitution — see Best practices on the Custom Prompt article.
FAQ
How do I generate new product titles or descriptions with AI? Use the Custom Prompt With Dynamic Values data service. See Generate or transform values with the data service Custom Prompt With Dynamic Values for setup steps and example prompts.
Can Productsup write SEO text or product content for me? Yes. The AI data services can generate SEO-friendly titles, descriptions, product types, and keywords from your existing data. Start with the Custom Prompt With Dynamic Values service.
Which AI models can I use? OpenAI models, Microsoft Copilot (Azure OpenAI), and Anthropic's Claude, depending on the service. You bring your own provider account and API key.
Does using AI data services cost extra? Productsup's integration with the AI providers is free. The AI provider itself may charge for usage based on your prompts and data volume. Review your provider's pricing before running at scale.
Caution
You are responsible for monitoring and controlling your own AI provider costs. Set a spending limit or usage budget directly on your API key with your provider (for example, in your OpenAI or Azure account) so usage can't exceed what you expect.
Will my AI-generated values change on every run? No. Results are cached per product, so the same inputs return the same output run over run. Values regenerate only when you clear the cache or change the model, context, prompt, or a referenced attribute.
Caution
The cache is temporary storage, not a permanent record. While we do our best to preserve it, we do not guarantee it. If an AI-generated value is business-critical, export it to an external system rather than relying on the cache as your only copy.
See also
How is this guide?