What we checked, and what we did not.

We took the S&P 500 constituent list and resolved each company to its primary public website. For every one we requested the home page and the small set of addresses an assistant checks before answering a question about a business.

Each site was then classified once. “Readable” means a file that exists and returns usable content. “Not an answer” means the address resolves but returns a web page, a redirect to marketing, or an error. It is a surprisingly common failure, because a site that serves its 404 page with a 200 status looks fine to a person and is useless to a machine.

“Blocked” means the site names AI crawlers in robots.txt and disallows them. We count it separately because it is the one result that means an assistant genuinely cannot read the site, and because the remedy is a conversation rather than a build.

What this does not measure: whether a page can be fetched at all. Almost every site can be. An assistant with browsing reads ordinary HTML perfectly well, and a company publishing nothing here is not invisible. It is illegible. The difference matters, and we would rather say it than let a number imply otherwise.

Of the 500, 494 served a home page. The remaining 27 did not resolve, timed out, or returned a hard block at the network edge from a region we could not work around; those are excluded from every percentage rather than counted as failures.