Prepare for a world where AI is your primary audience

Pam says: “Start preparing for an AI-first future, where the AIs are your primary target audience, not humans.”

Is AI going to be buying your stuff in the future?

“Absolutely. Obviously, we're a bit out from that, but I do think things are heading in a direction where humans are going to rely on AI agents to do a lot of things for them, especially when it comes to searching for, digesting, and distilling down information, and turning that information into a recommendation.

We're already seeing it going in that direction. Think about how much click-through rates have sunk in the past six months. That shows people are starting to rely on the recommendations that AI overviews give them, and they are not doing as much of the research themselves anymore.”

Will AI be consuming Google search, or will it have its own list of websites/search engines and make a decision based on that?

“Right now, AI chatbots are using traditional search engines.

Those two things have been merging. Traditional search engines are adding AI chat features, and AI chatbots are adding traditional web search features. As it stands, AI chatbots are doing the searching for you, on a traditional search engine. For example, Microsoft Copilot and ChatGPT use Bing, Gemini uses Google, Perplexity uses a mix, and Claude uses the Brave search engine

At the moment, the AI chatbot is using a traditional search engine on behalf of the human, but that's going to change in the future. I can even imagine not having websites at some point, at least not in the form that we know them now. Everything will have to be optimized for machine-readability, as opposed to humans.

We focus so much on the design, the prettiness, and the polished look and feel now. I don't think that will matter in the future. I envision websites being more like an XML sitemap, where it's just completely machine-optimized.

Right now, we have one foot in each world. You need to start preparing for an AI-first future, and there are some things that everyone should definitely start doing now, and some things to look ahead to as well.”

Can you just optimize for the slightly different way that AI browses and consumes the web, or should you optimize for what's coming in the future?

“Just because AI chatbots are still relying on traditional search engine databases and indexes right now, that doesn't mean that you shouldn't change your approach to how you put content out there.

There are certain things that make content a lot more easily digestible and understood by AI chatbots. If you have old school content on your website – which is pretty in-depth, with long paragraphs that each have five full sentences, and each subheading has several paragraphs beneath it – a human has no problem digesting that. LLMs, however, digest information in very small chunks.

If you think about it, every time you ask an AI chatbot for something, you get bulleted bits of information in the answer. That's because the way that LLMs digest and store information is in tiny little chunks. When they learn about something, they take the different tidbits and break them up into those chunks. A five-sentence paragraph might be stored as five different tidbits in five different places in the AI's vector database.

As it stands right now, they're relying on traditional search engines and their indexes, but they're absolutely doing their own crawling and building their own indexes as well. As they do that, they are not storing a web page the way traditional search engines do. A web page is not a thing to an LLM. A web page is filled with tonnes of tiny tidbits of information, and they're all stored separately.

Think of it like an Excel spreadsheet, with all kinds of different columns and cells. The information gets split up all over the place, and then it reassembles it as it starts to understand things. To make your content more easily understood and indexed by LLM chatbots, you can break things down into chunks.

Gone are the days of writing five full sentences in a paragraph and three paragraphs under a subheading. You want to get super granular, with one to three sentences in a paragraph and many, many subheadings – more than you would ever feel is natural to use. Obviously, you still want your content to look natural to humans, just way more chunked up.

I've been calling it ‘chunk SEO’. Think of it almost like you're writing in bullets: being so succinct in each section, and getting right to the point. It's like putting your conclusion sentence first. Typically, in a paragraph, the conclusion is the fifth of five sentences. Now, you want to open with that answer and then give a bit more information.”

Is a 2000-word blog article not as effective as it used to be?

“Not necessarily. The beefiness of the content still very much matters, and should still be there. It's really just about chunking that information up.

We have two clients who very quickly accepted our recommendations for chunk SEO. They are both websites of a very similar age and other metrics, but the difference is that one has content that is much longer than the other. Although they've both been rewriting their content and restructuring it in this very granular chunked-up way, the one with the longer content is outperforming the other in AI search.

They're both doing better than other websites that are not embracing that yet, but the length still matters a lot.”

How do you know that your chunking is as optimized as possible?

“With the way we're doing it right now, we're almost reverse-engineering the LLM's answer.

Google is still the primary market shareholder of search, so we focus on Gemini the most. Gemini's the LLM that powers the AI overviews at the top, but you get a more in-depth answer if you go to Gemini directly and chat with it like ChatGPT.

You can go and ask Gemini to search the web about the topic you want to write about. First, Gemini will give you an answer, and then it will go out to the web and search for websites that state facts backing up what it already said.

That's actually the way that AI-powered search works. It's the reverse of what we're used to thinking. We think that it goes and finds the answer, and it gives us what it found. It actually answers from the general knowledge that it's been trained on first (which is pretty much the entirety of the internet), then it goes and looks for websites to back up what it's saying.

Go to Gemini and look for tidbits of information that it didn't find a website to cite to back up that statement. That's what you need to include and drill down into within the content that you're developing.

Other than your overall presence in AI overviews and the traffic that you're getting from AI-powered search engines, it’s hard to know if you're doing the chunking right, but you can definitely reverse-engineer the chunked answers that you get from an LLM before you even work on your content.”

What is the ideal way to write FAQs nowadays?

“LLMs absolutely love both FAQs and tables of contents. They help them understand things, and that's the format they like to store information in and answer with, because it’s very chunked-up by nature.

An FAQ, with a typical short question and answer, is a great structure to add to your existing content – particularly product/service-based, higher-intent, commercial content. For longer content, adding a table of contents with anchor links that hop down to the subheading in the article is very good for AI-powered search.

Having one full blog article as the answer to a single FAQ can very much still apply, but there are always going to be sub-FAQs within that. You can think of it like a parent question with child questions under it.

Basically, think about People Also Ask. You type a question into Google, and you go to the People Also Ask section. That's how LLMs like to do things, too. They throw in those extra tidbits, or they prompt you at the end with a question to drill down further.

Having a single question in the title and a bunch of paragraphs as the answer is probably not the best structure anymore, unless you chunk that down. If you have a bunch of articles that are just a single FAQ with a long answer, break it up into sub-FAQs, and add the table of contents at the top. Having both of those together is even better.”

You also say brands must optimize not just for visibility, but for influencing AI decision-making. How do you optimize for AI decision-making?

“First, I think it's important to point out the difference between an AI chatbot and an AI agent. These terms get thrown around a lot and can sound very similar in meaning, but they're actually quite different.

AI chatbots just chat with you, and they distil and summarise information and knowledge. An AI agent can actually be autonomous in that it can make decisions and take actions on its own. That's the big difference.

If you go to ChatGPT and you tell it to do something, and you're not in agent mode, it really can't do much for you. If you told it to research the best flights to X country next September and book them, ChatGPT cannot do that for you, but an AI agent could. It can not only distil and summarise information, but it can also take another two steps: it can make decisions and perform actions.

We're heading towards a world where AI will be making decisions and purchases for humans. Therefore, websites won't look the way that they do now. They would be more of a portal for feeding information to an AI agent in a machine-readable format and helping to influence its decision-making.

That includes everything we said about making content machine-readable, but it’s also more than that. I think that the big difference for influencing an AI agent's decision-making is going to be the off-page factors.

Those factors are very important now (inbound links being the main one), but they will include more of the softer side of that as well: signals, recommendations, reviews, PR coverage, getting mentioned in news articles, and even people just talking to each other on social media. All of those brand equity-building things are going to be seen by the agents and taken into account, in the same way humans do now.”

Pam, what's the key takeaway from the tip you shared today?

“Start thinking differently about your online target audience. Although humans are still doing their own research and clicking, start to think about optimizing for robot decision-making.

Basically, there will be an intermediary between you and your target human. Think of it like going to book a doctor's appointment. You don't call up and speak to the doctor directly; you have to go through the staff to get an appointment.

That's how everyone should start thinking about their human target audience. You will have to go through an AI agent to get to them, so you need to start optimizing for recommendations from those agents.”

Pam Aungst Cronin is President and Founder of Pam Ann Marketing and Stealth Search and Analytics. Find out more over at PamAnnMarketing.com.