Technical SEO
Building a search program for an online coach from public data
An online coach has no storefront, so there is no address to put on a map and no neighborhood to be "near." I helped build the site and search program for Diogo Fitness, an online fitness coach in South Florida, through Ascent Vision, the marketing company I run with Grayson Sorrentino. The hard question was what a page about Chandler, Arizona could honestly say to someone who will never meet the coach.
The answer I settled on: it can say something true about Chandler, from a public source, or it should not exist.
The doorway-page problem
The obvious play for a service with no location is to generate a page per place: "online fitness coach in [city]," same paragraph, city name swapped. Search engines have a name for that, doorway pages, and a long record of discounting them. An online service makes it worse, because there is no real local fact to hide behind. Every page would be the same page.
So the rule for the place pages was blunt: a place page ships only if it carries real local data. If the data is not there, the page does not get built.
What the place pages are made of
Today the site has 51 state pages (50 states plus DC), 43 metro pages and 27 city pages under /online-coaching/. Each is assembled from three public sources, cited on the page itself:
- Activity and health behavior: CDC BRFSS and PLACES.
- Commuting: Census American Community Survey.
- Climate: NOAA 1991-2020 climate normals.
The point is not to decorate a page with statistics. Commute time, summer heat and local activity rates are the things that decide whether someone can keep a training habit, which is exactly what a coach works on. A coach who plans around a long commute and a hot August is saying something specific. The data lets the page say that for one place and not another. You can see how the state pages are organized from the index.
Alongside those sit about 35 articles under /learn/ that cite peer-reviewed sources. The one that does the most work is the article on what online fitness coaching is, because it answers the question most visitors arrive with before any place page matters.
The rules that keep it honest
A coach's site is exactly where marketing habits go wrong, so I wrote the prohibitions down and made the machine enforce them:
- No page shows a price. The build runs a check that fails if it finds a dollar amount.
- No invented testimonials, client counts or results. If it is not real and checkable, it is not on the page.
- Applications go to Instagram DMs or email. The site stores nothing; the apply page builds a message in your browser, and what you type does not leave your device until you send it.
The first rule is a build check rather than a guideline on purpose. Guidelines get forgotten in the fortieth batch of pages. A failing build does not.
A second agent re-fetches every source
I draft pages with agents, and agents are fluent in a way that outruns their evidence. So no page ships on the drafter's word. A second agent, with no stake in the draft, re-fetches every cited source and checks each sentence against what the source says.
It earns its keep on wording more than on numbers. The typical catch is a sentence that reads a little stronger than the source supports. The figure can be right and the claim still wrong by a degree, and a reader would never notice. A checker with the source open does.
What I would tell the next person
An online service has no local address, so place pages earn their place only with place-specific data. Everything else follows from that. If you cannot name the local fact a page contains, you are building a doorway page with a nicer template.
New domains get crawled fast and indexed slowly. Crawlers found the new domain quickly and read the whole site. Being added to the index is a separate decision, made later, and it is the one that matters. I would rather say that plainly than promise a timeline. It is also why I am not publishing results here: the program is young, and I would only be guessing.
Verification catches a different class of error than testing does. Builds catch broken links and a stray price. Only a second reader with the sources open catches a claim that is well formed, linked and wrong by a degree. I have written before about checks that passed while the work was wrong, and this is the same lesson at page scale.
Constraints are the product. "No price, no testimonial, no unsourced claim" sounds like a handicap for a marketing site. In practice it is why every page can be defended line by line. When a page is built that way, you can show someone the source for any sentence on it.
The whole site is checkable by anyone with a browser: open a city page, find the source it cites, and see whether the sentence matches.
Clayton Camera
Technical founder and software engineer in Orlando, Florida. Builds AI systems, autonomous agents, and full-stack software — including OrbitRoute and KnockFiber. Client work at workwithclayton.com.