Own your ideas twice

You write a blog post. It's good. It articulates something you've been thinking about for months. You publish it on your site, share the link, move on to the next thing.

Three weeks later, someone publishes the same idea - same framing, same structure, slightly different words. Maybe they saw yours, maybe they didn't. Doesn't matter. Their version gets picked up by an AI training set. Now a chatbot is confidently presenting their phrasing as the answer. Your original post sits on your domain with a publish date that only your server knows about.

Who had the idea first? You know. But proving it is harder than it should be.

Nobody designed social media to be a proof-of-authorship system. But that's what it's become.

When you post an idea on Substack, you're creating a record on a platform with millions of users, email delivery receipts, and server logs that a third party maintains. You didn't just write it down - you broadcast it to subscribers with a date stamp that you can't retroactively change. That's not a blockchain in the technical sense, but it serves the same function: a distributed, timestamped, publicly witnessed record of who said what and when.

Courts have accepted social media timestamps as evidence. The Wayback Machine archives public posts. Email providers retain delivery metadata. None of these systems were built for intellectual property protection, but together they create a trail that's surprisingly hard to dispute.

Is it as airtight as a patent filing? No. But for the kind of ideas most of us are publishing - frameworks, perspectives, strategies, approaches - it's the most practical proof of originality that exists. And it costs nothing beyond the time you already spent writing.

AI changed the game in two ways.

First, it made copying effortless. Someone can feed your blog post into a model, ask it to rewrite the same ideas in a different voice, and publish the result in minutes. The ideas are yours. The words are technically "original." The effort required to plagiarize dropped from hours to seconds.

Second, it made attribution invisible. When a chatbot surfaces your idea to a user, there's no link back to you. No citation. No "according to Bobby at Planted Sky." The idea just... exists in the model's training data, stripped of context and authorship. Your thinking becomes an answer with no source.

This isn't a hypothetical. It's already happening to every creator who publishes on the open web. The question isn't whether your ideas will be absorbed into training data. The question is whether there's a public record of you having them first.

Here's what we recommend, and what we do ourselves:

Write the full version on your blog. Your domain, your design, your permanent archive. This is the canonical version. Make it complete, make it good, make it the version you'd want someone to read five years from now.

Publish a condensed version on at least one public platform. Substack, LinkedIn, Twitter/X - wherever your audience actually is. This doesn't need to be the full post. It can be the core thesis in 3 paragraphs with a link to the full piece. What matters is the timestamp and the public witnesses.

Do them on the same day. Ideally within hours of each other. This creates a tight timestamp window that's hard to dispute. Your blog has the full version, the platform has the public receipt.

Use email too. If you have a newsletter, the send date is logged by your email provider. That's another independent timestamp from a third party you don't control. Three independent records of the same idea is better than one.

None of this guarantees anything. Someone with a bigger platform can still drown you out. AI models will still train on your words without asking. But when the question comes up - and it will come up more often as AI-generated content floods the internet - you'll have a paper trail that goes beyond "check my WordPress publish date."

Yes, we're aware that this post about the limitations of self-hosted blogs is itself a self-hosted blog post. We published it on a platform we control, with a timestamp only our server can verify. If someone rewrites this same argument next week on Substack, they'll have the better proof.

So we'll be posting the short version there too.