Technical optimization is the work of removing everything that stops search engines and AI tools from reaching, reading and understanding your store. On a typical store, between fifteen and forty per cent of the catalogue is effectively invisible before this work starts.
Those products are live. They are in stock. A customer with the direct link can buy them. But no search engine has ever indexed them, so nobody arrives at them from search, and no AI tool can quote them.
This is the least exciting work we do and usually the most profitable, because you are not building new demand. You are recovering pages that already exist and already convert.
What Is Crawl Waste and Why Does It Cost You Money?
Search engines give every site a limited amount of attention. On a store, filters and sorting options can create tens of thousands of near identical pages that soak up that attention before your real products get looked at.
Every filter combination on a collection page can generate its own web address. Colour, size, price band, brand, sort order and page number multiply together. A collection with six filters can produce more addresses than you have products.
None of those pages are useful to a buyer arriving from search, and none of them will ever rank. But a crawler does not know that until it has visited them, and every visit it spends there is a visit it does not spend on a product you actually want found.
The fix is a set of rules about which combinations are worth keeping, which should be blocked, and which should point back at a single main version. It is careful work, because blocking too much removes pages that were earning, and blocking too little leaves the problem in place.
Why Do Your Own Product Pages Compete Against Each Other?
When the same product is reachable at several addresses, search engines have to guess which one is the real one. They usually guess badly, and the strength that should sit on one page is split across four.
The fix is to choose one main version for every product, tell search engines clearly which one it is, and make sure every internal link points at it. Done across a full catalogue this is one of the largest single gains available to a store.
Can ChatGPT and Perplexity Actually Reach Your Store?
Many stores block the crawlers that AI tools use without knowing it. If those crawlers cannot read your site, being named in an AI answer is impossible no matter how good your content is.
AI tools use their own crawlers, with their own names, separate from the ones search engines use. Blocking rules written years ago, or added by a security tool or a firewall setting, often stop them at the door.
We check each one individually and report which are allowed, which are blocked, and which are being slowed down so heavily that they give up. Then we make a recommendation, because the answer is not automatically to allow everything.
There is a real decision here. Allowing these crawlers means your content can be used in answers that may not link back to you. Blocking them means you cannot be named at all. For a store selling products, being named is almost always worth more than the traffic you might lose, but it is your decision and we will lay out both sides rather than deciding for you.
How Fast Does a Product Page Need to Be?
Fast enough that a shopper on a phone with average signal sees the product and can act on it within about two and a half seconds. Most stores are well outside that on their collection pages.
Speed is measured wrong almost everywhere. Teams test the home page on a laptop on office broadband and see a good score. Buyers arrive on collection and product pages, on a phone, often on mobile data, where the same store behaves completely differently.
On a store, slowness usually comes from a small number of predictable places. Apps and scripts that were added for a trial and never removed. Product images uploaded at full size and resized in the browser. Review and chat widgets loading before the product does. Themes carrying code for features the store does not use.
We measure the pages that matter, find which of those causes applies, and fix them in the order of what each one is costing. Speed work is worth doing well and it is not worth doing forever, and knowing where to stop is part of the job.
What Product Data Do Machines Need From You?
Price, availability, condition, brand, unique product code, shipping cost and reviews, marked up so a machine can read them without guessing from the page layout.
This structured data is what puts your products into free product listings, into rich results with prices and star ratings attached, and increasingly into the shopping results appearing inside AI tools. Without it, your product is a page of text that a machine has to interpret. With it, your product is a record a machine can use directly.
Most stores have some of this and not all of it. Price and name are usually present. Unique product codes, shipping information and correct availability are usually missing or wrong, and wrong is worse than missing, because it gets your listings rejected.
We audit what exists, fix what is broken, add what is missing, and set it up so new products get it automatically instead of needing a person to remember.
What Does Technical Work Look Like Month to Month?
Heavy in months one and two, then steady monitoring with occasional projects. Technical problems do not stay fixed on a store, because the catalogue keeps changing.
The first two months are a clearing exercise. We work down the audit list, highest cost first, and most of the recoverable traffic comes back in this window.
After that it becomes monitoring. A store adds products, retires products, installs an app, changes a theme, runs a sale that creates ten thousand new pages. Any of those can undo work that was finished in month one, and without monitoring nobody notices until traffic falls.
So we run a scheduled crawl every month, compare it against the last one, and flag what changed. Most months this is quiet. The months it is not, we catch a problem in days rather than in the quarterly report.