Vouch for Your Website’s Humanity with “human.json”

An up-and-coming protocol proposes an interesting method of battling AI-generated slop.

As AI-generated slop becomes increasingly prevalent, it’s all the more important to know whether the websites you’re visiting are maintained by actual human beings, or if they’re just the output of AI chatbots. But how, exactly, can you determine that? While AI sites do possess some telltale signs, that sort of verification is more difficult and time-consuming than it should be.

One potential solution is the nascent “human.json” protocol, which allows “humans to assert authorship of their website content and vouch for the humanity of others.” Website owners upload a human.json file that contains some basic information about their website followed by a list of websites that they vouch for as human-generated:

{
	"version": "0.1.1",
	"url": "https://domain.com/",
	"vouches": [
		{
			"url": "https://human-website-1.com",
			"vouched_at": "2026-04-06"
		},
		{
			"url": "https://human-website-2.com",
			"vouched_at": "2026-04-06"
		}
	]
}

This file, which “indicates that the content of your site was generated by you, not AI,” is then linked to in the website’s <head> like so:

<link rel="human-json" href="/human.json">

For what it’s worth, here’s Opus’s human.json file, which vouches for several websites, including several by friends and family as well as some other sites that I trust. The file’s purpose, according to the protocol’s documentation, is to create “a network of trusted sites that allows people to discover other real humans by following that web of vouches.” But just how do you discover and take advantage of this network?

Enter browser extensions for Chrome and Firefox that identify if the website you’re visiting has a human.json file. If one’s detected, then the extension will notify you and display a list of other websites that are vouched for by the website you’re currently visiting. If you choose to trust that website, it’s then added to your “trust graph.” As you visit other websites, even ones without human.json files, the extension indicates if they’re connected to a website in your “trust graph” as well as how many “hops” they are from that website, thus giving you a better idea of who can be trusted.

The protocol is far from foolproof — for starters, there’s nothing that prevents an AI slop blog from having a human.json file of its own, thus falsely claiming that its content is human-generated — but I still think it’s an interesting step in the right direction. Meanwhile, the browser extensions enable and encourage users to take a closer look at the websites they’re visiting, and determine just how trustworthy they might be.

As I mentioned earlier, the protocol is still in its early stages. The proposed v0.2.0 release refines the standard and sets the stage for future enhancements, like the ability for website owners to specify an AI policy URL (such as an /ai slash page). Other proposed enhancements, like a scoring system, strike me as adding unnecessary complexity, though. I think human.json will ultimately work best if it’s kept as simple and direct as possible to encourage easier and wider adoption. By humans, of course.

Enjoy reading Opus? Want to support my writing? Become a subscriber for just $5/month or $50/year.
Subscribe Today
Return to the Opus homepage