The 2009 stack is a 2027 liability

The storefront still takes orders. That is the trap.

Every quarterly review for the last four years has ended the same way: revenue is flat, sessions are flat, the platform is stable, and therefore the platform is fine. The people defending it are not lying. From inside the CMS, nothing looks broken. The catalog renders, the cart works, the checkout completes.

Then you open ChatGPT, ask it which brake kit fits a 2016 Silverado with the tow package, and it names three marketplaces and two competitors you have never heard of. Not you. Never you.

That is not a ranking problem. It is a legibility problem, and a 2009 stack is very good at being illegible.

What actually broke

Nothing broke. The interface between you and the buyer changed underneath you.

For twenty years the deal was simple: a human types a query, a results page lists ten links, the human picks one. Your job was to be in the ten. Every dollar of SEO and SEM spend was aimed at that list, and every agency contract was written around that list.

Now a large share of product research ends in a generated answer. The buyer never sees a results page. They see a paragraph with three citations, and the paragraph is assembled by a model that had to understand your catalog well enough to recommend a specific part number.

A model can only recommend what it can parse. If your product data arrives as a JavaScript payload behind a filter widget, it gets parsed as a blank page with a nice banner.

The four things a 2009 stack cannot do

1. Serve product data in the HTML. If a crawler with no JavaScript execution sees an empty <div id="app">, that is the entire page as far as most retrieval systems are concerned. Product name, price, availability, fitment, application notes: gone.

2. Treat fitment as a first-class filter. Fitment is not a tag you bolt on at the end. It is the primary key of the catalog. Any architecture where a part's vehicle applications live in a separate table that only the internal search knows about is invisible from the outside.

3. Answer a question in a paragraph. Legacy templates are built for category browsing. They have no place to put a 40-word direct answer, no question-shaped headings, no FAQ, no spec table a model can lift cleanly. So the model lifts its answer from a forum instead.

4. Publish machine-readable trust signals. Structured data, consistent entity naming, a real llms.txt policy, documented crawler access. Most 2009 stacks strip all of this out of the templates because it was never part of the design.

What we would do in 90 days, without a replatform

We have watched three replatform proposals die in committee. Do not wait for the fourth. Almost all of this is additive.

• Server-render the product and category templates. Same CMS, same URLs, same design. The difference is that the HTML contains the product facts before the JavaScript runs.
• Rebuild the feed as the source of truth. If your Google Merchant Center feed is a nightly export from a spreadsheet, it is already stale. Feed the product feed from the same data that renders the page.
• Add one structured layer. Product schema with sku, brand, offers, and vehicle applications expressed as properties. One template change, applied everywhere.
• Put an answer block at the top of every category page. Two sentences: what this category is, what it fits, what to check before buying. Written by a human, edited by a machine.
• Instrument the new surface. Search Console will not tell you when an assistant cites you. You have to sample prompts monthly and log the answers manually. Ugly, but it is the only honest baseline you will get for free.

The part nobody says out loud

All of the above is a two-quarter project with one competent engineer and one merchandiser who actually knows the catalog.

It does not get done because no single person owns the surface where the customer now starts. SEO owns the old list. Paid owns the auction. Merchandising owns the feed. IT owns the platform. The generated answer belongs to nobody, so it gets budget from nobody.

That is the failure. Not the platform. The org chart.

Your store did not get worse. The doorway moved, and the department that used to hold the door was never told.

If you are inside one of these groups and you cannot say any of this on a call with your name attached, you are not alone. That is exactly why this thing exists.