The Charleston AI Visibility Index.

What the index measures

Every site was crawled by the same engine Standin uses on paid audits, identifying itself as an AI retrieval bot and reading only the HTML the server sends, with no JavaScript. That is how the bots behind ChatGPT search, Perplexity, and Google's AI answers read the web. The crawl covered the first twelve to twenty-five pages of each site, money pages first.

Each site was then scored out of 100 against a fixed rubric with five categories: crawlability (can the bot get in and read the content), structured data (does the site describe itself in a form a machine can trust), content and metadata (are the pages answerable and labelled), authority signals (who wrote this and who else points to it), and technical health. The rubric is deterministic. The same site scores the same way twice, and every finding points at something measured on the site.

Two things the index does not measure. It does not ask the assistants what they say about a business, which is a separate benchmark that costs real money per site. And it is not a judgment of the business. A fifty-year-old firm with an excellent reputation can score an F because its website was rebuilt on a JavaScript framework in 2021. The score describes the website's readability to machines, nothing more.

Scores by industry

Med spas, the trades, and retail boutiques lead. Those are industries that buy paid search and sell online, so the agencies and platforms serving them have kept the markup current. Restaurants are the largest group in the index and sit at the bottom, alongside fitness studios and galleries: beautiful sites, often on page builders that put the words in JavaScript and leave the HTML empty, with the menu as a PDF or an image. Law firms and dental practices land in the upper half, which fits how much both spend on being found. Industries with fewer than three scored sites are folded into the nearest group.

The Charleston 50: the most AI-ready business websites in the area

Named because they did well. Every site on this list can be read, understood, and correctly described by an AI assistant today, which is more than most of their competitors can say. Scores are out of 100 as of the crawl date; a site that changes will score differently in the next edition.

If you are on this list, you are welcome to say so. The index link and the crawl date are the citation. The free report below runs the same engine on your site and gives you the score and your five biggest findings; the full audit is a separate, paid piece of work.

The bottom of the index, without names

Nobody below the top 50 is named here, and that is permanent policy. The patterns at the bottom are worth describing, though, because they are the same three problems over and over.

• A local business, 19/100. The one reachable page served zero words of text to the bot. A visitor sees a full site; the crawler sees an empty JavaScript shell. No usable sitemap. No Organization schema, so an assistant has to guess what the business is.
• A local business, 19/100. The homepage delivers 7 words of HTML text; the rest only exists after JavaScript runs. No usable sitemap. No Organization schema, so an assistant has to guess what the business is.
• A local business, 19/100. The one reachable page served zero words of text to the bot. A visitor sees a full site; the crawler sees an empty JavaScript shell. No usable sitemap. No Organization schema, so an assistant has to guess what the business is.
• A local business, 21/100. The one reachable page served zero words of text to the bot. A visitor sees a full site; the crawler sees an empty JavaScript shell. No usable sitemap. No Organization schema, so an assistant has to guess what the business is. Titles and descriptions missing or duplicated across pages.

80 further sites refused the crawler outright: a normal browser gets the page, and a request that identifies itself as an AI retrieval bot gets a 403 or a rate limit. Usually that is a hosting firewall or a bot-protection default nobody chose. They are not scored, because a score would be fiction. They are also, today, invisible to every AI assistant, which is the finding. A further 115 businesses listed on public maps with a website had one that no longer resolves, has a broken certificate, or returns an error; those are excluded from every number on this page.

How to read a band

Method notes

• Scoring model v4.2 (August 2026). Crawls between August and October 2026. Sites crawled in August were not re-crawled for this edition; their scores are as of that date.
• Sites came from OpenStreetMap listings inside the Charleston tri-county area, plus businesses we had scanned for earlier work. Each site either sits at a mapped Charleston address or states on its own pages that it serves the area. National and regional chains were excluded wherever we could identify them; a few may remain in the aggregates, and any found in the named list will be removed on request.
• No Standin client, and no site Standin has built or rebuilt, appears in the index.
• Sites that answered the crawler with a 403 or a rate limit are reported as refusing it and excluded from every average. Sites that were dead, broken, or timed out are excluded entirely. Single-page sites were scored on the one page that exists.
• The index is free to cite with a link. Businesses may dispute a score by requesting a re-scan; the next edition carries the new number.