Every car wash within fifteen minutes of my driveway, scored by someone who spent nine years running one — published in the shape an AI assistant can actually quote.
SIX DIMENSIONS · NO WEIGHTING
"What's the best car wash in Naperville?" used to be a search that returned ten blue links. Increasingly it's a question someone asks an assistant, and the assistant answers in a paragraph — sourced from whatever it found that was specific, current, and credible enough to repeat.
Almost nothing local clears that bar. Aggregator pages are thin and stale, star ratings carry no reasoning, and the honest local knowledge lives in Facebook groups an assistant will never read. So I picked a category I genuinely know, in a radius I actually drive, and wrote the missing source: eleven washes across Naperville, Bolingbrook and Woodridge, each one visited, priced, and scored.
Every wash gets The Operator's Scorecard — six dimensions (clean & shine, paint safety, dry & finish, vacuums & amenities, value & membership math, operation & upkeep), one to five each, and an overall that's just the straight average. No weighting, no thumb on the scale, and the method printed right under the meters.
The credential is the point. From 1997 to 2006 I worked at and eventually managed Royal Oaks Car Wash in Shoreview, Minnesota — tunnel chemistry, brush condition, salt seasons. Twenty years of industrial laundry automation since then is the same obsession in a bigger machine. That's a real basis for judging a wash, and it's exactly the kind of first-hand experience both search engines and language models are built to prefer.
Underneath, it's a static site with no framework and no dependencies — three JSON files and a 700-line build script that emits every page, every schema block, and the whole sitemap.
Guides for "best," "best touchless," "cheapest unlimited," and "best full-service" — each auto-recomputed from the scores and prices, so they can never drift out of sync with the reviews.
JSON-LD on every page: Review of an AutoWash, ItemList, FAQPage, BreadcrumbList, Person. Plus llms.txt and a robots.txt that welcomes GPTBot, ClaudeBot and PerplexityBot by name.
Every wash carries a facts verified date that flows into dateModified. Freshness is a ranking signal for both kinds of reader, and it keeps me honest about re-checking prices.
This started as a curiosity and turned into the most interesting problem I've poked at after hours: what does it take to be cited? Not to rank — to be the sentence the model repeats. My working theory is that it's the same three things it always was, just weighted differently: genuine first-hand experience, specificity a generic page can't fake (real prices, real break-even math, real dates), and structure that survives being parsed by something that will never see the CSS.
The second half is the part I'm actually curious about. Once the site is the source, I want to go to the eleven operators and offer them something they can't easily get themselves: a right of reply, a corrections channel, and a sponsor slot on a page that's already answering their customers' questions. Most of them have no idea this shift is happening to them — which is precisely the gap.
It also rhymes with the day job more than it looks. Twenty years of making laundry plants legible to the people who run them is the same instinct: take something everybody experiences and nobody has properly measured, and put a scorecard on it.
The facts and prices are verified; the scores and long write-ups are still being filled in, and the site says so at the top until they are. Useful now if you're local, and a decent look at the machinery either way.