It’s a tough time to be a web marketer.
In just a few months, change has dominated how web content is discovered and consumed. For years, the pipeline for organic traffic was simple: you run a search on one of the major providers, like Google, you take a look at the top few results, and you click the one that seems best.
But with the rise of AI conversation engines, this model is adapting quickly, and the pace is speeding up.
The result? Direct traffic to web sites has declined year over year. Paid advertising is more expensive than before, and conversion rates are lower. Google tends to return Gemini results — their own foray into AI-assisted search — with most searches, and it’s estimated that around 60% of searches last year ended with no clickthrough, and that trend is increasing. The value of having a top search result is under half of what it was a few years ago.
So, what do we do now?
The rise of the GEO.
The simple fact is that people simply aren’t seeing links anymore. And, really, this shouldn’t be a surprise — while we’ve been trained to think that traditional search engines are designed to bring us to the right pages, in reality they were designed with a much simpler task: to bring us to the right answers.
Which means, to users on Claude or ChatGPT, the links were never the purpose. Not only has AI-assisted search removed the need for direct page links, when we do see these links users are becoming accustomed to ignoring them. In fact, Google's own recent introduction of "AI Mode" means that Google is heading in the same direction for many searches.
I spent a recent Saturday helping my daughter shop for a used car. Almost all of our searches about features and reliability reports were simply conversations with Claude. In past years, we would have been to dozens of sites for this information. Just to be sure, I checked the results afterwards: Claude may not know everything, but it seems to know what to look for in a used Subaru.
This is what we mean when we say that users are increasingly trusting the results from LLMs to get their answers. SEO is no longer the only strategy to attract organic traffic — a growing portion of the customer journey is going to happen outside of the realm of brand experiences.
So if this discovery and research process is happening in a GPT, how can we have any influence on what customers are being recommended? Or even if those recommendations are right?
Well, the good news is that a lot of what makes good, authoritative content for search engines still applies: writing good content, making your site accessible, and providing good metadata. In other words, a reminder that search engines — whether traditional or AI-supported — still need context and structure to understand what they’re crawling. Just like we need to worry about how our content is structured for SEO, we now need to worry about how our content is perceived and indexed by AI — a discipline that's so new that we’re not even all agreed on what to call it yet.
I’ve heard AIEO, or AEO, or LEO, for LLMs, but the one that seems to be sticking is GEO, Generative Engine Optimization.
Digestible Content
While a lot of what makes content understandable and valuable to a person also applies to LLMs, the key difference is one of reasoning. The balance is shifted because, as good as an LLM might get, it is still interacting with words and patterns, not actual thinking. What LLMs need is context.
When I'm shopping for a car, I know what a site means when it mentions "AWD.” I know from experience what's meant when someone says the car is "a manual." LLMs lack that context, so these things are just words. Is a “manual” a book? And what's an "awd?"
Granted, there is already some of this context within an LLM's training data. But as a marketer you can make its job easier by including some content on your site that provides explanation for some of these things that you might ordinarily take for granted. Corey Vilhauer, Blend's director of strategy and lead content strategist, has been a proponent of including extensive glossaries on sites for years now. (Corey's always been smart, but with the rise of LLMs he's looking like a prophet — don't tell him I said that.)
There’s also renewed emphasis on really good FAQs — not just questions you want people to ask, but the ones that they might really ask the LLM. Makes sense, right? If you give the machine an answer to a question it's trying to help with, it’s more likely to use it.
Metadata Becomes Even More Useful
The idea of using schema to help expose understandable data to search engines is not a new idea. It’s always been a good idea to use schema and embedded semantics like microdata and JSON-LD to help crawlers and systems better understand your content.
And, schema is ready to get to work. If you look on schema.org, there are hundreds of schema available for all sorts of use cases. In the past — and, in the past, we've already recommended focusing on the small subset that Google focuses on to generate their Rich Results cards in search results.
But with AI now also absorbing this content, there's a broader range of understanding, since an LLM can infer meaning from the structure without necessarily understanding the exact schema specification. Therefore, it becomes helpful to use a broader array of the available schema, not limiting yourself to the closest thing Google understands, but the closest schema to match your actual use case.
A new and emerging pseudo-standard is the llms.txt file. While a sitemaps.txt file is designed to helpfully list all of the pages a traditional search engine should index, llms.txt is intended to provide a concierge tour of your site for any learning model that encounters it, including key notes and links to major sections. Written in Markdown (LLMs seem to love Markdown), this file is intended to give an LLM a leg up in understanding your organization, its jargon, site structure, and major features.
Build Authority By Being Authoritative
It’s also helpful to demonstrate that you’re an authoritative source on the things that you want to talk about. More than just providing good content, it’s helpful to get your name out there. Are you working with other sites, providing articles, videos? Are you in the news, and linking to your earned media? Do you have a Wikipedia entry — and is it accurate? All of these are signals that your content is worth quoting and referencing because the rest of the internet is already doing that.
Your authority is determined not just by what you write, but how often other content refers to what you write. If other humans are linking to what your organization has published about a given topic, AI will be more likely to do so as well.
Everything Old is New Again
A lot of these things were already important with basic SEO, but they’re even more important now, because while you’re still getting your content out there, you’re trying to get it to people that may never even come to your website.
So your message has to be really good, your structure, your data, all have to be really good. There are no shortcuts.
But the truth is, there never have been.