• asudox@lemmy.world
    link
    fedilink
    arrow-up
    108
    ·
    4 months ago

    Block? Nope, robots.txt does not block the bots. It’s just a text file that says: “Hey robot X, please do not crawl my website. Thanks :>”

    • ɐɥO@lemmy.ohaa.xyz
      link
      fedilink
      arrow-up
      58
      ·
      4 months ago

      I disallow a page in my robots.txt and ip-ban everyone who goes there. Thats pretty effective.

    • Cynicus Rex@lemmy.mlOP
      link
      fedilink
      arrow-up
      13
      ·
      4 months ago

      Unfortunate indeed.

      “Can AI bots ignore my robots.txt file? Well-established companies such as Google and OpenAI typically adhere to robots.txt protocols. But some poorly designed AI bots will ignore your robots.txt.”

      • breadsmasher@lemmy.world
        link
        fedilink
        English
        arrow-up
        23
        ·
        4 months ago

        typically adhere. but they don’t have to follow it.

        poorly designed AI bots

        Is it a poor design if its explicitly a design choice to ignore it entirely to scrape as much data as possible? Id argue its more AI bots designed to scrape everything regardless of robots.txt. That’s the intention. Asshole design vs poor design.

    • majestictechie@lemmy.fosshost.com
      link
      fedilink
      English
      arrow-up
      7
      ·
      4 months ago

      This is why I block in a htaccess:

      # Bot Agent Block Rule
      RewriteEngine On
      RewriteCond %{HTTP_USER_AGENT} (BOTNAME|BOTNAME2|BOTNAME3) [NC]
      RewriteRule (.*) - [F,L]
      
      • drkt@lemmy.dbzer0.com
        link
        fedilink
        arrow-up
        19
        ·
        4 months ago

        This is still relying on the bot being nice enough to tell you that it’s a bot; it could just not.

        • IphtashuFitz@lemmy.world
          link
          fedilink
          English
          arrow-up
          7
          ·
          4 months ago

          Exactly. The only truly effectively way I’ve ever found to block bots is to use a service like Akamai. They have an add-on called Bot Manager that identifies requests as bots in real time. They have a library of over 1000 known bots and can also identify unknown bots built on different frameworks, bots that impersonate well known bots like Googlebot, etc. This service is expensive, but effective…

          • poVoq@slrpnk.net
            link
            fedilink
            arrow-up
            5
            ·
            edit-2
            4 months ago

            I wonder if there is an AI scraper block list I could add to Suricata 🤔

            • IphtashuFitz@lemmy.world
              link
              fedilink
              English
              arrow-up
              8
              ·
              edit-2
              4 months ago

              When any browser, app, etc. makes an HTTP request, the request consists of a series of lines (headers) that define the details of the request, and what is expected in the response. For example:

              
              GET /home.html HTTP/1.1
              Host: developer.mozilla.org
              User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:50.0) Gecko/20100101 Firefox/50.0
              Accept: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
              Accept-Language: en-US,en;q=0.5
              Accept-Encoding: gzip, deflate, br
              Referer: https://developer.mozilla.org/testpage.html
              Connection: keep-alive
              Upgrade-Insecure-Requests: 1
              Cache-Control: max-age=0
              
              

              The thing is, many of these headers are optional, and there’s no requirement regarding their order. As a result, virtually every web browser, every programming framework, etc. sends different headers and/or orders them differently. So by looking at what headers are included in a request, the order of the headers, and in some cases the values of some headers, it’s possible to tell if a person is using Firefox or Chrome, even if you use a plug-in to spoof your User-Agent to look like you’re using Safari.

              Then there’s what is known as TLS fingerprinting, which can also be used to help identify a browser/app/programming language. Since so many sites use/require HTTPS these days it provides another way to collect details of an end user. Before the HTTP request is sent, the client & server have to negotiate the encryption to use. Similar to the HTTP headers, there are a number of optional encryption protocols & ciphers that can be used. Once again, different browsers, etc. will offer different ciphers & in different orders. The TLS fingerprint for Googlebot is likely very different than the one for Firefox, or for the Java HTTP library or the Python requests package, etc.

              On top of all this Akamai uses other knowledge & tricks to determine bots vs. humans, not all of which is public knowledge. One thing they know, for example, is the set of IP addresses that Google’s bots operate out of. (Google likely publishes it somewhere) So if they see a User-Agent identifying itself as Googlebot they know it’s fake if it didn’t come from one of Google’s IP’s. Akamai also occasionally injects JavaScript, cookies, etc. into a request to see how the client responds. Lots of bots don’t process JavaScript, or only support a subset of it. Some bots also ignore cookies, and others even modify cookies to try to trick servers.

              It’s through a combination of all the above plus other sorts of analysis that Akamai doesn’t publicize that they can identify bot vs human traffic pretty reliably.

              • Daemon Silverstein@thelemmy.club
                link
                fedilink
                English
                arrow-up
                1
                ·
                4 months ago

                What if a bot/crawler Puppeteers a Chromium browser instead of sending a direct HTTP requisition and, somehow, it managed to set navigator.webdriver = false so that the browser will seem not to be automated? It’d be tricky to identify this as a bot/crawler.

                • IphtashuFitz@lemmy.world
                  link
                  fedilink
                  English
                  arrow-up
                  3
                  ·
                  4 months ago

                  Oh there are definitely ways to circumvent many bot protections if you really want to work at it. Like a lot of web protection tools/systems, it’s largely about frustrating the attacker to the point that they give up and move on.

                  Having said that, I know Akamai can detect at least some instances where browsers are controlled as you suggested. My employer (which is an Akamai customer and why I know a bit about all this) uses tools from a company called Saucelabs for some automated testing. My understanding is that our QA teams can create tests that launch Chrome (or other browsers) and script their behavior to log into our website, navigate around, test different functionality, etc. I know that Akamai can recognize this traffic as potentially malicious because we have to configure the Akamai WAF to explicitly allow this traffic to our sites. I believe Akamai classifies this traffic as a “headless” Chrome impersonator bot.

    • Cynicus Rex@lemmy.mlOP
      link
      fedilink
      arrow-up
      7
      arrow-down
      8
      ·
      edit-2
      4 months ago

      Lies, as in that it’s not really “blocking” but a mere unenforceable request? If you meant something else could you please point it out?

      • Da Bald Eagul@feddit.nl
        link
        fedilink
        arrow-up
        37
        ·
        4 months ago

        That is what they meant, yes. The title promises a block, completely preventing crawlers from accessing the site. That is not what is delivered.

  • Onno (VK6FLAB)@lemmy.radio
    link
    fedilink
    arrow-up
    24
    ·
    4 months ago

    This does not block anything at all.

    It’s a 1994 “standard” that requires voluntary compliance and the user-agent is a string set by the operator of the tool used to access your site.

    https://en.m.wikipedia.org/wiki/Robots.txt

    https://en.m.wikipedia.org/wiki/User-Agent_header

    In other words, the bot operator can ignore your robots.txt file and if you check your webserver logs, they can set their user-agent to whatever they like, so you cannot tell if they are ignoring you.

  • DigitalDilemma@lemmy.ml
    link
    fedilink
    English
    arrow-up
    23
    ·
    4 months ago

    robots.txt does not work. I don’t think it ever has - it’s an honour system with no penalty for ignoring it.

    I have a few low traffic sites hosted at home, and when a crawler takes an interest they can totally flood my connection. I’m using cloudflare and being incredibly aggressive with my filtering but so many bots are ignoring robots.txt as well as lying about who they are with humanesque UAs that it’s having a real impact on my ability to provide the sites for humans.

    Over the past year it’s got around ten times worse. I woke up this morning to find my connection at a crawl and on checking the logs, AmazonBot has been hitting one site 12000 times an hour, and that’s one of the more well-behaved bots. But there’s thousands and thousands of them.

  • NullPointer@programming.dev
    link
    fedilink
    arrow-up
    17
    ·
    4 months ago

    robots.txt will not block a bad bot, but you can use it to lure the bad bots into a “bot-trap” so you can ban them in an automated fashion.

    • Dave.@aussie.zone
      link
      fedilink
      arrow-up
      8
      ·
      4 months ago

      I’m guessing something like:

      Robots.txt: Do not index this particular area.

      Main page: invisible link to particular area at top of page, with alt text of “don’t follow this, it’s just a bot trap” for screen readers and such.

      Result: any access to said particular area equals insta-ban for that IP. Maybe just for 24 hours so nosy humans can get back to enjoying your site.

          • doodledup@lemmy.world
            link
            fedilink
            arrow-up
            1
            arrow-down
            3
            ·
            4 months ago

            You misunderstand. Sometimes you want your public website to be indexed by search engines but not scraped for the next LLM model. If you disallow scraping alltogether, then you won’t be indexed on the internet. That can be a problem.

            • ɐɥO@lemmy.ohaa.xyz
              link
              fedilink
              arrow-up
              7
              ·
              4 months ago

              I know that. Thats why I dont ban everyone but only those who dont follow the rules inside my robots.txt. All “sane” search engine crawlers should follow those so its no problem

        • mox@lemmy.sdf.org
          link
          fedilink
          arrow-up
          5
          ·
          edit-2
          4 months ago

          Robots.txt: Do not index this particular area.

          Problem is that you’re also blocking search engines to index your site, no?

          No. That’s why they wrote “this particular area”.

          The point is to have an area of the site that serves no purpose other than to catch bots that ignore the rules in robots.txt. Legit search engine indexers will respect directives in robots.txt to avoid that area; they will still index everything else. Bad bots will ignore the directives, index the forbidden area anyway, and by doing so, reveal themselves in the server logs.

          That’s the trap, aka honeypot.

  • 5opn0o30@lemmy.world
    link
    fedilink
    arrow-up
    20
    arrow-down
    5
    ·
    4 months ago

    Wow. A lot of cynicism here. The AI bots are (currently) honoring robots.txt so this is an easy way to say go away. Honeypot urls can be a second line of defense as well as blocking published IP ranges. They’re no different than other bots that have existed for years.

    • DigitalDilemma@lemmy.ml
      link
      fedilink
      English
      arrow-up
      9
      ·
      edit-2
      4 months ago

      In my experience, the AI bots are absolutely not honoring robots.txt - and there are literally hundreds of unique ones. Everyone and their dog has unleashed AI/LLM harvesters over the past year without much thought to the impact to low bandwidth sites.

      Many of them aren’t even identifying themselves as AI bots, but faking human user-agents.

  • breadsmasher@lemmy.world
    link
    fedilink
    English
    arrow-up
    10
    ·
    4 months ago

    It isn’t an enforceable solution. robots.txt and similar are just please bots dont index these pages. Doesn’t mean any bots will respect it

  • Cynicus Rex@lemmy.mlOP
    link
    fedilink
    arrow-up
    12
    arrow-down
    3
    ·
    4 months ago

    #TL;DR:

    User-agent: GPTBot
    Disallow: /
    User-agent: ChatGPT-User
    Disallow: /
    User-agent: Google-Extended
    Disallow: /
    User-agent: PerplexityBot
    Disallow: /
    User-agent: Amazonbot
    Disallow: /
    User-agent: ClaudeBot
    Disallow: /
    User-agent: Omgilibot
    Disallow: /
    User-Agent: FacebookBot
    Disallow: /
    User-Agent: Applebot
    Disallow: /
    User-agent: anthropic-ai
    Disallow: /
    User-agent: Bytespider
    Disallow: /
    User-agent: Claude-Web
    Disallow: /
    User-agent: Diffbot
    Disallow: /
    User-agent: ImagesiftBot
    Disallow: /
    User-agent: Omgilibot
    Disallow: /
    User-agent: Omgili
    Disallow: /
    User-agent: YouBot
    Disallow: /