{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreifff3jrfbq2aok34r32s6igrd2f5kjkhnkvjp47usc642rxpucg6e",
    "uri": "at://did:plc:s67opkie37qfdnrjagqiarmh/app.bsky.feed.post/3mhnmbmre2ar2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreie6znrx2vlwexyrtpnsqfgizj5memsckawnpftpcahvdwg4y45hnq"
    },
    "mimeType": "image/jpeg",
    "size": 167440
  },
  "path": "/maintaining-a-human-web-with-humans-json-aiblacklist/",
  "publishedAt": "2026-03-21T12:48:00.000Z",
  "site": "https://tzovar.as",
  "tags": [
    "AI",
    "artificial intelligence",
    "slop",
    "indieweb",
    "human.json",
    "small web",
    "is",
    "well",
    "documented",
    "‘removing friction’",
    "1",
    "ai-ublock-blacklist",
    "“AI” uBlock blacklist",
    "Ploum",
    "2",
    "human.json project and browser extension",
    "Mastodon post by Evan Hahn",
    "3",
    "have a look at mine",
    "some knowledge commons pushing back against this",
    "↩",
    "Codeberg",
    "humans.txt",
    "mine"
  ],
  "textContent": "\n\n\nMy ~~dislike~~ strong opposition to engaging with “AI” generated things, be that texts, images, code or anything else is well-documented. But avoiding the mindless slop is still getting harder, as it seems that more and more developers, writers – and of course SEO-scum trying to make quick buck – can’t avoid the siren song of ‘removing friction’, all negative consequences be damned.1 Technical solutions of course can’t alone help us solve the problem, they can offer at least some help in minimizing our exposure, by either positively highlighting human-made things or by actively blocking out the slop. I’ve recently started using two tools to help me with that: The ai-ublock-blacklist and the human.json approach.\n\nI came across the “AI” uBlock blacklist thanks to a post by Ploum, and it does exactly what you might expect: You can add it as an extra filter list to your adblocker and from then on, instead of visiting ‘AI’-generated websites, you’ll just get a big warning. I find this particularly useful when searching for things, as these days all search engine results are just full of low-quality results that come straight out of the text-extruder and at best just SEO-spam that’s overly verbose or at worst completely wrong. Which is what you really don’t want for any technical or factual questions you might have. The blocklist itself is collectively maintained, and if you find missing “AI”-generated slop pages, you can add them via a pull request.2\n\nOn the other side of the problem, to identify human-created websites, the human.json project and browser extension can help. I found it via a Mastodon post by Evan Hahn, and it works by adding a `human.json` file to your website root, to self-declare as being human-created.3 Of course, as any page could self-proclaim to not use ‘AI’, you can also vouch for other websites that you know are human-created. In that way, you create a “web of humanity”, similar to the “web of trust” in the old-school _PGP_ framework.\n\nThrough the browser extension of `human.json` you can then trust some _“seed”_ -pages that you know are human-created. From those you can get a network of pages that those pages trust etc, and which are then indicated as such in the browser extension, giving you at a glance an idea if a website is vouched for as human, based on your own trust-network. E.g. If you trust website A, and website A vouches for websites B, C, and D, those will appear as vouched via one hop in your browser too. If you want to see a real example, you can have a look at mine.\n\nThe `human.json` approach is still quite new, but seems to have been picked up quite a bit in my small corner of the small web, which is heartening to see. And between it and the AI blocklist, it’s at least gotten a bit easier to avoid the worst parts of the machine-generated web. Give it a try if you also want to avoid reading slop.\n\n## Footnotes\n\n  1. Luckily, there are still some knowledge commons pushing back against this ↩\n\n  2. Yes, it would be nicer if the list was maintained on Codeberg instead of GitHub, but oh well, we can’t have it all. ↩\n\n  3. While superficially similar, the much older humans.txt is mostly about celebrating the people who made a website and giving some human-readable details. See mine for an example. ↩\n\n\n",
  "title": "Maintaining a human web: human.json & the ‘AI’ blacklist",
  "updatedAt": "2026-03-21T12:48:00.000Z"
}