How do you audit an ecommerce site with millions of product pages?

Templates first

Catalogue sites are generated. A five-million-URL site is usually a handful of templates repeated: product, category, filtered category, paginated category, brand, search results, editorial. Defects live in templates, not in individual URLs, so the unit of analysis is the template.

Enumerate them, sample within each with enough volume to catch conditional behaviour (out of stock, no image, single variant, long title, missing attribute), and audit the sample properly rather than everything superficially.

Let the logs set the priority

Crawl budget on large catalogues is usually consumed by URL patterns nobody intended to publish: faceted combinations, sort parameters, session artefacts, pagination running past the point of usefulness. Logs show where it goes. That distribution decides what to fix first, because a template that consumes a third of the budget and generates no revenue is the highest-value fix on the site regardless of how it scores in isolation.

Entity consistency at scale

On large catalogues the AEO problem is rarely per-page quality. It is inconsistency: the same product described differently across variants, structured data that disagrees with the visible page on price or availability, brand entities that resolve differently in different categories.

Models corroborate across sources. Internal contradiction at scale is a reason not to cite you, and it is invisible in any per-page audit score.

What the deliverable looks like

• Template inventory with defect findings attached per template, ranked by URLs affected multiplied by commercial value
• Crawl budget distribution with the patterns to suppress and the mechanism for each
• Structured data conformance across the catalogue rather than on a sample of ten products
• An implementation sequence your engineering team can put into a sprint without further interpretation