#57: Chronicles of the Generative Internet

How GEO is changing marketing

Your social feed is a personalized stream of content. Websites show you custom product descriptions optimized to drive a purchase. Chatbots give you a single answer without a click. The internet is generating a user experience based on your past behaviors. We are in the age of the generative internet.

With hundreds of millions of ChatGPT weekly users and millions more on other LLMs, the generative internet is certainly catching on. Searching and discovering information works fundamentally differently on ChatGPT compared to Google Search. Prompts are longer and people tend to write additional context. Also, it’s commonplace to receive a single answer as the output in generative engine results. This differs from the typical Google Search results page, which includes ten blue links, some from advertisers, and leaves it up to you to sort through whether the results answer your question. Why would you sift through ten links when you could receive a single answer? While both engines can produce wrong answers, the efficiency of a single output is often a game-changer.

Lately, I’ve been paying attention to the concept of generative engine optimization (GEO). In simpler terms, GEO is marketing for the generative internet.

Similar to search engine optimization (SEO) where one tries to rank for specific phrases or “keywords” on a traditional search engine (like Google or Bing), GEO is all about how often your brand shows up in prompt responses for specific categories, as well as the general sentiment of how your brand is perceived. And by “show up”, I mean how often a brand name is mentioned in relevant prompt responses, as well as the frequency the brand’s website is used as an information source. But there’s more. Because LLMs provide detailed output, answers tend to have a bias that’s meant to provide the user guidance. This concept is called sentiment, and the sentiment of your brand matters immensely when it comes to awareness and perception.

While this might sound familiar to SEO, the rules of the game are shifting. GEO places an emphasis on understanding the inner workings of where each LLM receives its information from. LLMs are typically trained on a mix of publicly available internet material (such as Wikipedia), licensed sources (such as Reddit), and data from partners or human trainers. A key component of a GEO strategy is creating content that ends up appearing in specific Reddit threads, on a Wikipedia page, or in a “Best of” section of The New York Times. With that said, the use of some publisher content is disputed, as in the case of the New York Times lawsuit.

Imagine this hypothetical: you gave niche sunscreen advice to a group of strangers on a Reddit thread a few years ago. It turns out that ChatGPT has licensed access to Reddit data. Your comment may have been included in the model’s training data as it’s being cited when people ask questions about sunscreen today. ChatGPT (and other LLMs) prioritize high-authority content, and high-authority content doesn’t necessarily need to come from an M.D. Community validation, like upvotes, can help bubble up content that models are more likely to reflect.

Now, this isn’t that unlike SEO, however the types of content that win in GEO are a bit different. FAQ guides and step-by-step instructions perform well because they're structured in a way that's easy for LLMs to understand.

There’s a lot of anxiety around keeping up with the constantly changing news, products, and players in the AI world. Also, there are plenty of concerns about AI becoming too powerful. I find it reassuring that much of the content that is cited frequently comes from sources that we as online communities produce, whether through Wikipedia edits, Reddit discussions, or publisher lists.

I see the winners of AI marketing falling into two buckets. First off, GEO analytics companies provide insights into how your brand is perceived in generative engine results, thus are well situated. Any marketer (or even business owner) who doesn’t want to be left behind will soon find a GEO software subscription essential. And of those GEO analytics firms, the ones that can turn insight into action in the way of automated content to appeal to a specific prompt category are better positioned. Companies like Profound and AthenaHQ have built impressive software around the insight to action step.

Moreover, there are many other GEO software providers, and I see this as an industry that will quickly consolidate. The analytical interfaces aren’t drastically differentiated from one another, which suggests consolidation will come down to customer acquisition speed rather than product differentiation.

Furthermore, marketers who have a knack for understanding the bigger picture will thrive in the generative internet age. By leveraging GEO software, a marketer can pinpoint the exact Reddit thread or “Best of” list that drives either brand mentions, citations, or sentiment. They can then look to build out an organic content strategy on Reddit based on their learnings. Or strike up the proper publisher partnership to make sure their brand stays in a top spot of that “Best of” list.

Consider this. You learn the things people like and don’t like regarding your brand via sentiment score analysis. From there, you can create or source content to help shift the narrative in your favor (or double down with the existing perception). Maybe there’s a new type of creator you end up working with that can best highlight a product use case that LLMs continue to cite. Or another scenario. LLMs regularly cite how your product is confusing to use. Tweak your website copy and add an instructional video. You’ll have more happy customers because you acted on the insight provided by GEO.

Learnings from GEO impact both your paid and organic marketing efforts, creating a flywheel that will ultimately affect the sources that LLMs train on. Rarely do new acquisition channels pop up so quickly and it’s even rarer do they reach the sheer scale that generative engines have. The ability to influence share of voice, frequency of mention, and sentiment will separate the brand winners from those that fall behind in the AI marketing realm.

I’m staying close to GEO, as I believe we are witnessing a transformational shift in how humans discover information. The question isn’t whether GEO is here to stay, but how you’ll get ahead of the curve.