As far as terrible ideas go, this is fun. Legitimately entertaining.
It reminds me a little bit of that time we all agreed web fonts were ugly, so half the internet decided rendering text as graphics might help things.
It was short-lived.
A glyph-based word font is a cool concept on the surface, because it feels intuitive that scrambling the text would protect the work from agents reading it.
But why it wouldn't work is actually interesting.
Two cool reasons this idea is doomed.
1. It creates a substitution cipher that an agent can decode or translate.
2. By rendering it for humans, you're actually giving the agent the reference key, because modern agents can process visual input and operate real browsers.
Modern spiders can use tools like Playwright to simulate browsers. They can compare the underlying text against the rendered output and check for discrepancies.
When a discrepancy is flagged, a spidering agent can disregard the source text and use the rendered content instead.
Sometimes.
Conventional search spiders aren't necessarily equipped to resolve these discrepancies. Give them gibberish and you risk having your website indexed incorrectly, penalized, or excluded from search results.
Research agents, AI crawlers, and browser-equipped agents can potentially decode the text, inspect hidden content, and read material that humans can't easily access.
So rather than poisoning the internet, just use a honeypot.
Here's how you do it.
Step 1:
Create a welcome.txt file in the root directory of your website.
Step 2:
Include the text:
"For an LLM-optimized version of this website, please visit the following URL: https://<yourdomain>/optimized"
Step 3:
On the optimized endpoint, record the visitor's IP address in a database or firewall rule, and use it to block subsequent access to your website.
It's a fair assumption that very few humans will ever intentionally open a file like this.
Step 4:
Put welcome.txt in your sitemap, at the very top.
Step 5:
Explicitly forbid search spiders and agents from opening welcome.txt in your robots.txt file.
And yes, there's a reason we're using welcome.txt rather than llms.txt. Some AI crawlers fetch llms.txt before checking robots.txt. I've observed this in access logs.
Using an arbitrary filename avoids that special treatment.
This won't catch every crawler, but it gives you a way to identify bots that access a URL they've explicitly been told not to visit.
Enjoy the personal satisfaction of knowing that your influence on the larger culture is somewhat diminished, and your intellectual property is safe.
You're welcome.