By the time you reach the generate step, the hard planning is already done. You have a template that reflects the five-part structure, and you have loaded and prioritized a batch of products worth working on. Generation is where those two pieces come together and actually produce a draft description you can read, edit, and eventually publish.
This article walks through the practical decisions in the generate step: which products to select, which template to apply, how brand voice and rewrite controls affect the result, and what to expect from the drafts that come out the other side.
Generation works from the batch of products you loaded and prioritized in the previous step. For a first run, keep the batch small, somewhere in the range of five to twenty products, so you can read every draft closely rather than skimming a hundred at once. As you build confidence in your template and instructions, larger batches make sense, but a small first batch is how you catch problems in your template before they repeat across your whole catalog.
Choose products that represent the range you actually sell. If your catalog spans multiple categories, include at least one from each in your first batch so you can see how your template performs across different kinds of products, not just the easiest case.
If you have built more than one template, for example separate structures for different product categories or brands, this is the point where you confirm the right template is applied to the right batch. A template built around outdoor gear specifications will not automatically make sense for a completely different product category. Matching the template to the batch is a quick check worth doing every time, not just the first time.
Beyond structure, generation typically lets you influence tone and how aggressively the existing text is rewritten versus replaced with newly researched content. Brand voice settings shape word choice and tone: direct and technical, warm and conversational, or whatever matches how your store actually talks to customers. Rewrite controls influence how much of the original supplier text survives versus how much is replaced by new, researched content.
A useful starting approach is to lean toward a fuller rewrite for products with thin or generic supplier copy, and a lighter touch for products that already have solid, brand-appropriate descriptions you mostly want to restructure rather than replace. The exact names, ranges, and defaults for these controls can change between releases, so confirm the current options in your live account rather than assuming they match an older screenshot or tutorial.
Generation produces a structured draft that follows your template's five parts: opening, summary, feature and benefit list, detailed description, and technical specifications. It is built from your product's existing data plus researched information about the product, brand, and likely use case, shaped by whatever global and section-level instructions your template carries.
The result is a draft, not an automatically approved final version. Nothing is published to your live Shopify store during generation. You are looking at a proposed listing that you still get to read, edit, approve, or reject before anything reaches a customer.
You can generate a description for a single product to test a template change, or run a batch across your prioritized list once you are satisfied with the results. Exact batch limits and generation behavior depend on your plan and the current product capabilities, so check your account for current limits before planning a large run. A large batch takes real processing time to complete, not a fixed or instantaneous duration, so plan your workflow around the actual behavior you observe rather than assuming a specific speed.
For your first real generation run, this sequence works well:
That last step matters more than it seems. A template instruction that is slightly off will produce the same small flaw across every product in a batch. Catching it on five products and fixing the instruction saves you from catching it on five hundred products later.
Generation is not a guarantee that every fact in the draft is correct. It is not a substitute for review, and it should not be treated as the final step in the workflow. Enrichio supports research and generation, but merchants should verify facts, specifications, and any claims before publishing, the same way you would check a new hire's first draft before it goes to a customer.
Once you have generated a batch of drafts you are reasonably happy with, the next step is the review and publish stage, where you confirm accuracy, check formatting, and push the approved descriptions back to Shopify. Generation and review work together: a good template makes generation more consistent, and a careful review process makes sure that consistency never comes at the cost of accuracy.
A few patterns show up often enough in a first generation run that they are worth naming directly. The first is generating a large batch before validating the template on a handful of products. It feels efficient to run everything at once, but if the template has a flaw, you now have that flaw repeated across the whole batch instead of caught early on five products.
The second is skipping the brand voice setup because the defaults seem fine. Defaults are a reasonable starting point, but they were not written for your specific store, your specific customers, or your specific tone. A few minutes spent describing how your brand actually talks pays off across every product you generate afterward, not just the current batch.
The third is treating a generated draft as finished because it looks polished. Polished and accurate are not the same thing. A draft can read smoothly and still contain a specification that does not match the actual product, especially if the underlying Shopify data was incomplete or outdated when you loaded it. Reading, not skimming, each draft before it reaches the review and publish stage is what catches this.
The first few generation runs will feel slower than they eventually become, mostly because you are still calibrating your template's instructions and learning what a strong draft looks like for your specific catalog. That calibration period is normal and worth the time. Once your template reliably produces drafts you only need to lightly edit, generation becomes the fast part of the workflow it was designed to be, and the review step becomes your main remaining checkpoint before publishing.