The roofer Googlecouldn’t see.

A tag telling Google to stay away. The new site was already built and passing its checks when we read the old site’s pages, and every page that loaded normally carried a robots tag reading “noindex, nofollow” in its code.

Google is plain about what that tag does. When its crawler reads one, its guide to blocking indexing says Google will “drop that page entirely from Google Search results”, and links from other sites don’t save it.

So we looked from the outside. A search for site:threestone.com came back empty, twice. Google says that search doesn’t always list everything, which is why we trust the page source first, but an empty answer for a whole domain tells you plenty. Search the company’s own name and Yelp came first. The one threestone.com result was the Smyrna page, down at #4.

Meanwhile their Business Profile still held #1 in the map pack for “roofers nashville tn”, so from the outside nothing looked broken.

Nobody could say who did it or when. Archived copies of the home page show no such tag as late as June 11, 2026, so it arrived some time in the three months after that. The tag, together with a dead sitemap at /wp-sitemap.xml, is what one checkbox in WordPress produces: Discourage search engines from indexing this site, on the Reading Settings screen. WordPress’s sitemap code reads that same setting. We can’t prove that box is what happened here. We can tell you it’s one click, and that nothing rings when somebody makes it.

We took corrective measures and fixed it right away. Nobody knew about it or how it happened.Michael Blair, co-owner

Unticking that box would have fixed one setting. It wouldn’t have fixed the site. The old site was on a clunky, slow, outdated WordPress platform, so we started from scratch for a better design, faster speeds, and a site that’s more reliable and safer from hackers.

There’s no WordPress to update now and no plugins to patch: it’s hand-written HTML and CSS on Cloudflare’s network. Plugins accounted for 91% of the 11,334 new WordPress vulnerabilities Patchstack logged in 2025, so having none removes that whole category.

It’s also AI friendly now. OpenAI’s GPTBot, ClaudeBot and PerplexityBot are allowed in, and there’s an llms.txt for the language models to read.

Google came back a page at a time, and the work kept going. The launch wasn’t the end of it: we keep doing Three Stone’s on-page and off-site SEO and AIO, the work that gets a business into AI answers.

On October 7, two weeks after launch, Google had indexed 20 of the site’s 72 pages. We sent the core pages in for indexing, and by November 4 it was 69 of 72.

Google clicks went from 193 in the old site’s last four weeks to 241 in the latest four full weeks, and we’re still working to pass the 278 the old site had in the summer before it slid. How much of that slide came from the tag, we can’t tell you. The timing fits, and timing isn’t proof. Clicks on the website’s own Google listings went from 79 to 148 in those same weeks. We also audited all 119 domains linking to them, one by one.

Those are search numbers, and I’d read them as only that: who found the site and clicked.