RIPT
A few years ago, finding a shirt went like this. You typed “retro Batman t-shirt” into Google, you scrolled, you clicked, you found us. Maybe.
Now a chunk of people do something else. They open ChatGPT or Perplexity, describe the shirt they want, and get back a short answer with a few places to buy it. No scrolling. No page two. Three or four names, and everyone else may as well not exist.
That change sent us down a rabbit hole this year, and what we found in our own data was not flattering.
The part where we were invisible to ourselves
Here is the uncomfortable bit. We had thousands of designs on this site that search engines had never properly filed away. Not ranked badly. Not filed at all.
When we actually looked, Google’s own reports showed 102,935 pages on riptapparel.com sitting in the “not indexed” pile. A shirt that is not indexed cannot be found by anybody, by any method, ever. It is a t-shirt in a warehouse with the lights off.
Some of that was our own doing. Every colour and size combination was quietly generating its own web address, so a single design could spawn dozens of near-identical pages. Search engines saw the mess and shrugged. Some of it was old template text repeated across thousands of products, which tells a search engine nothing about whether a shirt has Ghostbusters on it.
And the pages that were showing up were wearing the wrong labels. Our page titles, the blue line you actually click in search results, were generic. Shoppers scrolled past them.
What we changed
Nothing glamorous. Over 18 days in April we rewrote 30 page titles on our most important pages, cleaned up the duplicate web addresses, replaced template product text with descriptions of the actual design, mapped over a thousand customer reviews onto the individual products they were written about, and pushed the pages that had never been filed to be looked at again.
Then we watched Google Search Console every day, which is less exciting than it sounds and more addictive than it should be.
What happened
Over those 18 days, monthly organic clicks went from 2,170 to 8,450. The click-through rate, meaning the share of people who saw us in results and actually clicked, went from 3.3% to 18.3%. Of the 30 pages we rewrote, 22 improved. Eight did not, which is worth saying out loud, because anyone telling you they go 30 for 30 is selling something.
Some individual pages were absurd. Our Daily Deals page went from 8 clicks to 661. Trending Designs went from 3 to 239. The Graveyard collection, where retired designs go, went from 11 to 103.
Six months on, the picture held up rather than snapping back. Clicks over the period went from 11.3K to 25.8K, and the “not indexed” pile fell from 102,935 pages to 36,400. Still too many. Better than it was.
What this has to do with robots
Here is the link between the boring housekeeping and the shiny AI question.
The assistants people now ask for recommendations do not have a secret separate catalogue of t-shirt shops. They read the web. If a page is not indexed, not readable, or describes itself in template language that could belong to any shirt, an assistant has nothing to work with and will happily recommend somebody else who wrote a clear sentence about their Ghostbusters design.
Being findable by a robot and being findable by a person turn out to be mostly the same job, done properly.
So when we say we spent the spring fixing page titles, that is the unglamorous answer to a question we get asked in a shinier form: how do you show up when someone asks an AI where to buy a retro tee?
Who did the work
We did not guess at any of this. We brought in CiteVantage, who went and measured it, published the full breakdown including the pages that did not improve, and now works on exactly this problem for other stores.
We will keep chipping away at that 36,400. There are a lot of shirts in the dark back there, and some of them are really good.