How We Built a Shopify SEO Automation System for WMK Berlin

The SEO team maps primary and secondary keywords to each collection and maintains a central keyword register. Proposed terms are checked against keywords already assigned elsewhere on the site.

💡 Why it matters: This reduces the risk of two collections competing for the same search intent.

For each main keyword, the workflow reviews leading competitors and related queries. Those findings are inputs, not automatic instructions.

💡 Why it matters: An SEO specialist decides what is relevant before a term becomes part of the working strategy.

The workflow scans all live products inside each collection, using both product-page text and all available product images. This matters especially for vintage furniture, where useful details are often visible in photos even when they were never included in the original description.

Using the Claude API, the system analyses those images together with the written product data and builds a richer picture of what is actually present in each collection. The model can therefore work with specific product evidence instead of falling back to broad or generic category language.

These product observations are saved and reused, so the same catalogue does not need to be analysed from scratch every time the collection content is updated.

This also creates a strong base for the next steps in Shopify product automation. The same product data can later be used to improve individual product descriptions and build more accurate, descriptive ALT text for product images.

Each generated collection page includes a section with selected top products from the current assortment.

This list is connected to the live catalogue. When a product is sold, removed or no longer belongs in the collection, it is automatically dropped from the list. New relevant products can then take its place.

That means the collection content does not stay frozen after publication. It continues to reflect the real inventory as the store changes.

For users, this keeps the page useful and accurate. For SEO, it also means the collection is regularly refreshed with current product information instead of becoming outdated over time.

Because the product selection is handled as a dynamic module rather than hard-coded into the surrounding collection copy, sold items do not leave stale references behind. The page keeps the same collection structure while the products shown inside it can change with the catalogue, so new inventory can be reflected without rewriting the full description.

The generation layer is deliberately restrictive. Keywords are used only where the assortment supports them. Product-specific facts are kept in the parts of the page where they can be updated safely. General claims about an entire collection are not allowed unless the available product data supports them across the relevant set.

FAQ content follows the same logic. If there is enough real information for four useful questions, the block contains four. It is not stretched to ten with generic filler.

Before content is published, a specialist checks accuracy, keyword logic, internal linking and whether the generated text reflects the real catalogue. The system handles the repetitive work; the team keeps control of strategic and publishing decisions.

💡 Result: The workflow removes repetitive work without removing editorial responsibility.