Phraza is a product data and content platform for webshops. It enriches every product with structured, verified facts, then writes the content from those facts in 18+ languages. One file in. Better data and finished content out.
Most AI tools give you text. Phraza gives you the product data behind the text as well. Analysis & enrichment is the foundation and is included for every product. Then you choose the content you need.
Analysis & enrichment + a meta-title = €0.04 per product. Analysis & enrichment is required for every product, because all content is written from it. Add more content types or languages later. The enriched data is reused and never paid twice.
The AI model writes the sentences. What makes them correct, consistent and ready to import is the work before and after.
| Generic AI wrapper | Phraza | |
|---|---|---|
| What you get | Text | Enriched product data and finished content |
| What the product is | Guessed from the name | Classified in the Google Product Taxonomy (5,595 categories) |
| Product facts | What you give it, plus what the model fills in | Read from text and images, conflicts resolved, stored as core data |
| Missing facts | Often invented | Left empty and logged |
| Instructions | Generic prompts, or prompts you write and maintain | Versioned rules per content type, language and category |
| Languages | Often written in one language, then translated | Written natively in each language from the same facts |
| Scale | Quality drifts as the list grows | Thousands of products per run, in the background |
| Output | Text, with or without a file, but no fixed data structure | CSV or JSON in a fixed structure, validated and ready to import |
Every product is placed in the Google Product Taxonomy. The category decides which of 449 attributes matter. Phraza reads them from your text and your product images, resolves conflicts, and leaves a field empty rather than guessing.
Search engines, marketplaces and AI agents that shop for consumers rely on complete, structured attributes. Phraza delivers them as data you can use anywhere.
Phraza translates facts, not text. Each language is written from scratch with its own rules for grammar, punctuation and number formats.
Each text follows a versioned recipe for its content type, language and tone, with title formulas per industry and length limits built in.
Runs process thousands of products in parallel in the background. Every product is handled on its own. Test a sample first, then run the rest.
Each text can be traced to the facts, the recipe version and the model call that produced it.
Every response is checked against a fixed output format before it is saved, so thousands of products export cleanly as CSV or JSON.
The built-in AI Assistant works on your own product data. Ask which products need review, organize collections or start an enrichment. It asks before it saves.
Each customer's product data sits in its own database, hosted in Germany. Phraza runs on Google Gemini via Vertex AI in the EU region.
Product content for clients takes time and is hard to keep consistent across a team.
No. A generic AI wrapper sends your text to a language model and returns what it writes. Phraza first classifies every product in the Google Product Taxonomy, extracts facts from your text and product images, resolves conflicts and stores the result as structured core data. Only then does it write content, using versioned instructions for each content type, language and product category.
ChatGPT, Gemini and generic AI wrappers write text from what you give them. Phraza delivers two things from one CSV file: enriched, structured product data and finished product content in 18+ languages. Facts that cannot be derived from your data are left empty rather than invented, and every language is written natively from the same facts.
Yes. Enriched product data, with the Google product category and attributes read from text and images, can be exported as CSV or JSON on its own, with or without content.
No. Phraza translates facts, not text. Each language is written from scratch with rules for that language's grammar, punctuation and number formats, based on the same core data. The content reads as native, and the facts are identical in every language.
Phraza runs on Google Gemini via Google Vertex AI in the EU region. Customer data is never used to train AI models.
Start with a sample of 5 products and check the data and the content.
