This is part two of a two-part series on AI and digital marketing.

The way we search for information is changing, and the way Google and other search engines display answers is changing. Some people are moving away from traditional search and going straight to large language models (LLMs) like ChatGPT and Google Gemini to answer their queries. Many still rely on Google, but Google serves up AI answers with certain queries, and you’ve seen it – AI-generated summaries and overviews to certain types of searches.

At EVG Media, we know a thing or two about SEO and helping our partners rank for crucial keywords and appear in SERP features (think: knowledge panel, people also ask, shopping results). But how do you make your brand or business show up in AI answers? Does your current SEO-optimized content cover your bases? (Spoiler: no, it doesn’t.)

First, let’s take a brief crash course on how LLMs produce answers.

How do LLMs work?

In much the same way we teach toddlers to comprehend their first language, LLMs have to be trained. This is accomplished by feeding the model relevant data. For example, if your goal is to train an LLM to generate descriptions of houses for real estate listings, you want to feed your model pre-existing house descriptions.

From there, the LLM carefully analyzes the words that appear in your data, paying special attention to the order and frequency of those words (those two words will be important going forward). The more data it receives, the better it will become at recognizing patterns and understanding how all of the information works together. It can then start to predict words in sequences based on what it has learned from its training data.

Using our real estate example, the more often the LLM sees the words “spacious backyard” together while in training, the more it comes to associate those words with one another. And if it sees “spacious backyard” together more often than “fenced-in backyard,” it will determine that the former is preferable to the latter. So when the time comes for the LLM to write a description itself, it will be more likely to use the phrase “spacious backyard” rather than “fenced-in backyard.” The key term here is more likely. The LLM could still write a description using “fenced-in backyard,” but it will generally use that phrase less often.

For a more in-depth explanation of the order and frequency concept, check out Stephen Wolfram’s fascinating blog post on the subject.

How to optimize your content for AI

So now that we know a little more about how LLMs produce their answers, how do we get our content to appear in them? Unfortunately, it’s not as straightforward as it might seem.

The world of SEO revolves around the E.E.A.T. principle, which focuses on an individual website’s experience, expertise, authoritativeness, and trustworthiness. The better the content on your website is at conveying that you are an authority in your field speaking from well-informed experience, the better your content will rank. But, as we’ve established, LLMs learn from the order and frequency of words, and it doesn’t crawl every single indexed website on the internet for data.

That means your business has to appear on a page that the LLM uses for training.

Time for a new example. Let’s say that I ask Google Gemini to give me a list of the best roofers in Upstate SC. Unless Gemini has used every roofer in the area’s website for training, it isn’t going to be able to tell me about all of my options. Instead, it’s going to give me the business names it has seen most often (frequency) alongside the words “best roofers in Upstate SC” (order) in its training data.

So how do you find out where LLMs get their data? In one of his recent blog posts, Rand Fishkin used the audience research tool SparkToro and keyword research tool BuzzSumo to compile a list of websites featuring content related to a specific keyword. Then, he input that list into ChatGPT and asked it to rank the likelihood of those sites being used for training for LLMs. While he cautioned that the response wasn’t necessarily 100-percent accurate, it gave a good idea of which websites AI gathers data from. And those are the websites businesses have to appear on in order to make it into an LLM response.

As you probably already know, getting your business’ name onto a website that isn’t yours is challenging. It’s even more difficult to make it onto high-authority websites that LLMs pull their data from. However, a bit of PR, pitching a sponsored post idea to an editor, and good old-fashioned networking can help you get the job done. Thoughtful, non-spammy link building can also play a big role here. It can be time-consuming, and it involves reaching out to sites and presenting a case for your inclusion, but a great link from an authoritative source can be a huge boost to your site’s credibility and reach. There are plenty of tools out there (Ahrefs, SEMRush, and others) that can help simplify this process. And a good CRM (think HubSpot) can automate many tedious processes and track your results.

Although your website’s content doesn’t play a direct role in helping your brand appear in an AI answer, high quality copy with relevant internal and external links helps prove that you know your stuff, which in turn helps representatives from other websites see the value in including your business. Not to mention, it’s great for organic search.

Want to make sure your web content continues to be optimized for the ever-changing search landscape? Let’s talk.