Did I predict GEO in 2013?

I was cruising around YouTube, finding all the videos, podcasts, and what-have-you that I’ve been on to add to the Appearances page on this site. Back in January 2013, I was on Higher Ed Live: Looking Ahead — Marketing & Web Predictions for 2013, with Seth Odell, Mike Petroff, and Ma’ayan Plaut. The topic was predictions for 2013, and I think I accidentally foreshadowed Generative Engine Optimization, or GEO, the practice of trying to get your content cited in AI search platforms.

If you squint your eyes and shake your head a bit, anyway.

Back then we were in the middle of the responsive web design shift. We were all grappling with our traditions of long-form content writing and what that meant for reading on tiny screens. I can’t even count how many times I had to tell my higher ed clients not to hide content on small screens. Instead, they should rethink — and rewrite — their content for this new medium.

“People have to be able to access and derive the same — or a very similar — meaning from every form that [the content] takes on. And I hope we get past thinking about the specific devices to just thinking about what the content is itself.”

Content should be ubiquitous

When I think about content strategy (and I’ve thought about it a lot over the years), I tend to fall back on hospitable design principles. Everyone should get the complete experience no matter how they’re experiencing it. As the creators, it’s really a shift in focus from the mode to the creation itself.

Even in 2013, when I said content should be ubiquitous, I was thinking about the value of the content itself. And that’s exactly what I’m seeing today with the shift to AI search and GEO. Where it shows up, though, is now dependent on the content. The quality of the content, its completeness, and its usefulness are what get it into citations.

That’s sort of a reverse from how we thought about the mobile web and responsive design.

Function now dictates form

The functional quality of content and its reach via references, links, and quotations controls where it lands. That’s function dictating form. It should be the ultimate achievement of content-driven design, but unfortunately the form it takes is generic, overly wordy, and hyper-mediated.

Most of that is down to how so many people use generative AI. GenAI has a word vomit problem. That’s why I dislike using AI search — nobody has time to read all of that. But that’s the output. What goes in works differently. The agents doing the research want to spend fewer tokens (cheaper), run fewer agents (cheaper), and do it fast (cheaper), so they favour content whose meaning is obvious.

Clean, clear, simple content gets trusted more by humans and machines. It gets picked up, processed, dressed up with (way too many) new words by the model, and spat out into a response or on a page.

And you don’t know where or how it’ll be presented.

The morphability of content

How many layers of mediation happen between the words I’m writing here and their eventual resting states? I’ll publish this post here on my blog, then I’ll copy and paste part of it into my newsletter. A couple of days later, I’ll paraphrase it and put it on LinkedIn, hoping for clicks. I might even make a video out of part of this one.

If it eventually makes it to a citation in ChatGPT or Claude or something, only the stripped-down literal meaning will be pulled in, and then an LLM will dress it up in its own words. That’s 3 or 4 layers of mediation. At least.

This is classic Marshall McLuhan. In Understanding Media (1964), McLuhan writes, “the content of any medium is always another medium.” As long as that holds, we have to be thinking about the forms our content might eventually take. Or ignore the forms entirely and focus on the content quality itself.

How malleable is the content you’ve created? Can the meaning you intended survive the inevitable hyper-mediation it’ll go through?

Keep what control you have

I think this is why original thought and original writing are still so important. Years ago, Douglas Rushkoff wrote this book, Program or Be Programmed, and while the form of programming has shifted, I think it’s still apt here. Original content might be the only way to influence the new models. It’s certainly going to be one of the few ways you’ll stand out, even if that’s only apparent when someone actually interacts with the original content.

What we create will have to morph into different forms and still retain its meaning. And that’s hard to do. Maybe even impossible. We probably won’t be able to control everything about where or how our content appears in the future. What you can control is the source material and how well its crafted.