Premium Fruit, Invisible Online

The client was preparing to expand from its home market into Taiwan, Hong Kong, Malaysia, and Singapore, and wanted to know whether the site could carry that expansion before committing to an ongoing SEO retainer. Rather than start the retainer on faith, they asked for a standalone technical review first: what is actually wrong, what is actually fine, and what has to happen before spending real budget on four new markets at once. That is a harder brief than it sounds, because a pre-contract audit sits under a real incentive to lean one way or the other. Pad it and the bigger contract looks more necessary. Soften it and the relationship stays comfortable a little longer. Neither is honest, and neither was the brief here.

Four questions

Every one of the 93 products was checked against a standard, not sampled. The site's console and network activity were inspected in a real browser session, not just its source code, which is how a tag-management tool that was installed and never finished being wired up was caught at all. A public search for the site and the company name stood in for Search Console access, reported as a signal rather than a certainty. AEO was evaluated as its own category rather than folded into SEO, including a live Lighthouse "Agentic Browsing" test. And when the client independently ran a Google PageSpeed test mid-engagement and shared it, that data was credited as client-supplied and cross-checked against the audit's own findings, rather than quietly absorbed as if it were Reidify's own measurement.

• Metadata6
• Structured data3
• Site architecture2
• Heading structure2
• Content2
• Indexability2
• Performance2
• Visual design2

17 further categories carried exactly one finding each, from security headers to competitor benchmarking. 25 distinct categories in total.

The site was still running on Shopify's own placeholder subdomain, eighteen months into selling produce internationally, the kind of detail that reads as a real signal to an international buyer paying up front for perishable fruit from an unfamiliar overseas supplier. Twenty of the 93 products, all sharing one naming pattern, carried an identical copy-pasted line declining to describe the product at all, not thin content in the abstract but the same sentence twenty times, traceable to one batch. A separate pair of listings turned out to be the same product at two live web addresses, each quietly working against the other's ranking. A public search for the site and the company name returned nothing, the finding that makes every other finding conditional, since a site that is not indexed has nothing else worth sequencing until that is confirmed. And the one test built for what comes after search, a live Lighthouse Agentic Browsing run, found seven forms an AI shopping agent could not read or fill in, on a platform whose own robots.txt had just started shipping agent-facing directives by default. The infrastructure for being useful to an agent had arrived. Nothing on the site was written to use it.

Implementation has not happened, so there is no post-fix number to report, and no quote exists to run. Both would have to be invented to include, and this page does not do that. What is provable is the audit itself, so the audit itself is the whole case study: a site that had never been indexed, mapped precisely enough that the next twelve months of work has a sequence instead of a guess.