AEO: Optimize for the follow up
When someone interacts with an AI, they are not searching for a “result” but are problem-solving through a dialogue that evolves with every response from the LLM.
This week’s newsletter is sponsored by North Star Inbound and Peec AI
Paid subscribers of this newsletter can download the accompanying asset for this post: customer journey workbook
We have spent the last two decades using SEO best practices for optimizing for a very specific, robotic interaction: a user types a keyword into a box, and we attempt to satisfy that isolated request with a single click. In traditional SEO (Pre-2021), the focus was always on treating the query as a singular action, with the user’s problem as a query to be solved immediately.
[Sponsored by Peec AI]
2026 is the year to take AI visibility seriously.
Track and improve your brand’s visibility in AI search with Peec AI.
AI search analytics built for modern marketing teams:
Track how often your brand appears in AI search results
Monitor visibility, ranking position, and sentiment in one dashboard
Discover the prompts and sources that influence AI answers
LLMs fundamentally break this model because they facilitate conversations rather than transactions. When someone interacts with an AI, they are not searching for a “result” but are problem-solving through a dialogue that evolves with every response from the LLM. We often forget that humans are messy thinkers; we rarely know exactly what we want upfront. Think about your own LLM conversations and how often they end in a place vastly different than where you started.
A user might ask a question and get an answer, but that initial response almost always triggers subsequent questions. The LLMs are designed to always end with a question to keep the conversation going, so it does. If your content serves them that first answer but fails to anticipate the questions that inevitably follow, you have already lost them to a competitor who thought further ahead and maintained mind share throughout the journey.
SEO needs to broaden
Most SEO strategies are woefully unprepared for this shift because many marketers are still building dead-end landing pages designed to capture a single query phrase. While some marketers have adapted slightly to rebrand these target phrases as prompts rather than keywords, the outcome is identical if the experience ends at the landing page.
In a conversational search environment, that single-page victory can not be the end result because the user has more questions, and if your marketing approach doesn’t answer them, the AI will simply pull those answers from someone else. Even worse, if you force the user to hit the ‘back’ button to find the next part of the answer, you have told the algorithm that your site is incomplete.
Think of this flow like a hardware store rather than a library. If a customer asks a store assistant for a specific drill bit, a bad experience is simply pointing them to the correct aisle. A solution is to ask what they are drilling into (in my upcoming book, Customer Intelligence, I refer to this as the WHY), because if they are working with concrete, the standard bit they requested will fail. The customer asked a question, but the actual need was for something else. That is only uncovered with further questions and then answers. That is the principle we need to apply to SEO today, where the initial query is just the opening line of a much longer dialogue.
Stop building pillar pages
I have built dozens of massive pillar pages in my career, where the process was always to identify a high-volume keyword and write thousands of words to cover every possible angle. In the past, this was the gold standard for ranking on a competitive keyword, but in a conversational search environment, this format is obsolete. The old journey of learn, navigating pages, and clicking on a CTA no longer happens on a website; it happens in the LLM.
When a user asks an AI how to form an LLC, the model might summarize your pillar page to deliver a clean and accurate answer. The LLM has essentially stolen your work to bring that user to a new place in the journey. This user’s reaction to the LLC formation response is rarely to close the tab and move on; instead, they immediately wonder about the costs, structure, and a myriad of logistical questions. In traditional SEO, that was more searches. Now, that is a follow-up question.
If your product is nothing more than a static wall of text, the AI extracts what it needs and moves on to cite the follow-up answers from other sites. Your content may have helped the user in the LLM, but they aren’t ready to click on to your site. They have more questions, and only once they get those answers will they enter your funnel
You might wonder how you can predict what users will ask next without access to the private chat logs of millions of users that some AEO tools claim to have. The answer is that you have something significantly better: your actual customers and the data they generate every single day.
Stop relying solely on keyword research tools that show only aggregate search volume, and start looking at sources that reveal the conversational intent of a single human decision-maker. Tools will only tell you what people typed, but they never tell you what people were thinking. To get that, you should analyze your customer support tickets and user session recordings to identify the exact points where people get stuck. These sources reveal the actual question sequences your users follow and the specific concerns that prevent them from converting.
Internal linking as a guide
This approach requires rethinking your information architecture, because internal links can no longer be solely about passing authority to boost rankings. Your internal links should map that conversational flow, so when the LLMs extract your site, you stay relevant through the entire journey, and so your on-site journey matches the way a current user is being primed by LLMs to navigate content
If a user is reading about ‘best time to visit Japan,’ the very next logical step is not a weather chart; it is a pricing calendar. There isn’t a default here because it’s dependent on YOUR customer’s journey.
Maybe the weather chart IS the right answer for someone in the research phase. If you shove pricing at someone who’s six months away from booking, you might alienate them. The conversational model works both ways. You need to answer the question they asked AND anticipate the next one, but you can’t skip steps, or you’ll lose trust.
Zero-click is not the goal, but the teacher
Some are trying to make just showing up in an LLM response the end goal, but this is not SEO. This is a brand metric, this means using SEO budget for brand is a really bad idea. (Message me for a deeper explanation.) SEO is and will always be about driving acquisition from organic traffic. You can’t acquire if they don’t click. Marketers must accept that the top of the funnel is now rented land controlled by the AI.
I believe zero-click search is actually a filter that destroys low-value intermediaries and rewards real product owners. If your business model depends on arbitrage or getting someone to click just so you can show them ads, you are adding the exact kind of friction that AI is designed to remove. If you are selling cars, AI overviews/LLM responses remove the tire-kickers just looking for information, but they don’t change the TAM of people who want to buy cars online.
AI is a qualification system that handles initial research and works through surface-level objections before handing the educated user over to you. When a user is ready to take action and needs to actually do the thing rather than just learn about it, AI can’t solve that problem. AI is only a resource for information.
Start small, but start today.
You don’t need to overhaul your entire site overnight. Start by picking your most common customer journey (use this guide if you are a paid subscriber) and interviewing five recent customers to ask them what questions they had at each stage of their buying process. Map out that sequence, and you will likely find a consistent pattern of questions that appear in roughly the same order for most buyers.
Stop trying to out-publish the robot. Instead, use your content to build a bridge from the general answer the AI provides to the specific solution only you can offer. If you can anticipate the user’s second, third, and fourth questions better than the general model can, you earn the click that will eventually happen.
If you see results from this new flow, the results are not just LLM visibility, but conversion from your ICP; you know you are onto something. Don’t just adapt your old SEO tactics to new AI lingo. Everything has changed; you need to change with it.
[Sponsored by North Star Inbound ]
What happens when you combine best-in-class SEO with conversion optimization?
BigRentz’s traffic increased by 186% (85k), added 1950 conversions in 12 months.
Self Financial’s traffic increased by 50k/month, and added 685 new customers.
Lastly, Secure Data added 1968 phone calls.
North Star Inbound’s SEO strategies earn leads, conversions, and revenue..
Book a call for a free content audit and 10% off any engagement.
Intrigued? [Learn more here]
the content hub/cluster model suddenly makes way more sense through this lens. not as a keyword coverage play, but as a way to keep your brand present across the entire conversation arc.
what i keep wondering though: how do you even know when you've been "dropped" mid-conversation? like if someone starts asking about project management tools and you appear in answer 1, but by answer 4 they've narrowed down and you're gone... is that a win or a loss?
feels like we need entirely new attribution models for this.
The hardware store analogy hits home for me. Traditional content was built assuming users arrive with fully formed questions, but LLM interactions are inherently iterative. I've noticed this in my own usage - initial prompts are often vague and the real value comes from the refinement process. The challenge now is building content that stays relevant across an entire conversation thread, not just the entry point. Zero-click being a "filter" rather then a goal is a useful reframe.