Is there a simple way to severly impede webscraping and LLM data collection of my website?

Speaking from experience, be careful you don't become over-zealous in your anti-scraping efforts.

I often buy parts and equipment from a particular online supplier. I also use custom inventory software to catalog my parts. In the past, I could use cURL to pull from their website, and my software would parse the website and save part specifications to my local database.

They have since enacted intense anti-scraping measures, to the point that cURL no longer works. I've had to resort to having the software launch Firefox to load the web page, then the software extracts the HTML from Firefox.

I doubt that their goal was to block customers from accessing data for items they purchased, but that's exactly what they did in my case. I've bought thousands of dollars of product from them in the past, but this was enough of a pain in the ass to make me consider switching to a supplier with a decent API or at least a less restrictive website.

Simple rate limiting may have been a better choice.

I did this a while back for blocking LLMs and there are more methods discussed in that threads comments.

https://lemmy.world/post/14767952

Scrape a bunch of Onion articles, link them together in an index, then post an invsible link from your home page that spiders will follow but humans can't see.

Write a script to randomize the words on all the articles and link them in too. Then change the image tags to point to random wikimedia files.

If there's one thing we've learned, it's that there's very little quality control. Channel your inner Ken Kesey / Merry Prankster. Have fun.

Why not add a basic http Auth to the site? Most web hosts provide a simple way to protect a site or directory.

You can have a simple username and pass for humans, but it will stop scrapers as they won't get past the Auth challenge unless they know the details. I'm pretty sure you can even show login details in the Auth dialog, if you wanted to, rather than pre sharing them.

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

Should cover any polite web crawlers but it is voluntary.

https://platform.openai.com/docs/gptbot

Might have to put it behind a captcha or other type to severely limit automated access.

It's not realistic to assume it won't get scraped eventually. Such as someone paying people to bypass capatcha or web crawlers that don't respect robots.txt. I also don't know if Google and Microsoft bundle their AI data collection that doesn't also remove your site from web search.

Put each character in a spans with random classes, intersperse other random characters all over the place also with random classes, then make the unwanted characters hidden.

Bonus points if you use css to shuffle the order of letters too.

Accessibility? Pffffft.

I'm just gonna address the domain question.

ccTLD's for countries that are members of the EU usually have pretty strong privacy protection, especially if you are buying as an individual.

.de (Germany) is probably the cheapest (3-4€) but if you're not a resident you will need the registrar to arrange a mailing address for you for a small fee (another 3€ or so). Still going to be a pretty low price.

.nl is another cheap option, without any residency requirement.

The only issue with both is that you can only buy for one year at a time.

The owner's details in the registry are never published. Legit requests in case of abuse etc. need to go through the registry.

Any attempts to mangle the body of the pages or obscure in in JS are moot. Most competent stealthy scrapers have visionai as a fallback, so even if you reduce the ability to programmatically parse the page body, the bot can just snatch an image of the page and OCR the contents.

No, not really as they best way would be making it totally private.

Edit: I see you edited the title. You might be able to slow down LLM training. However, your content is such a small percentage in the whole that I doubt it would matter.

The simplest way might be to add a artificial delay to the page load. You could create a simple loading page that is just long enough to cause bots to move on. However, this will completely break search indexing assuming that this method works.

Wasn't there some new robots.txt rule for this?

Possibly, see https://github.com/ai-robots-txt/ai.robots.txt but I just discovered it myself while looking for a Robots.txt a la CrowdSec/AdBlocking lists, so feedback appreciated!

Blocking non-Mozilla user agents has eliminated 99% of scraping in my experience. I've seen a few larger sites do it as well but not many.

If you look in your access logs, or /var/log/nginx/access.log and look for user agents in the log file that indicate things like chatgptbot, etc. Then add if ($http_user_agent ~* "useragent1|useragent2|... useragents") { return 403; } to the server block of your websites config file in /etc/nginx/sites-enabled/. You can also add a robots.txt that forbids scraping. Chatgpt generally checks and respects that... for now. This paired with some of the stuff above should work.

Use a special/custom font where the letter shown differs from the character used.

use javascript 🤣

I'm a bit confused by your question: it sounds like you want to advertise yourself and your work. Why don't you let AI scrape your information? If I were you, I'd want a chatbot to spit out my details when someone asks it to name the name of someone who does what I do.

I'm violently anti-AI, but this is the one use case I would happily feed it information: to use it as an amplifier to spread public information I want to broadcast as far and as wide as possible.