AAO: What is Assistive Agent Optimization and How to Implement It

SEO has been following the same core logic for decades: making Google find your content, understand it, and show it when someone is looking for something relevant. That logic worked well because the process always had a human at the center. The search engine showed options. The user chose.

That's changing. And the change isn't gradual.

Today's AI systems, Chat GPT, Perplexity, Google AI Mode, Gemini, don't just show options. They evaluate, recommend and, in some cases, act directly. They book, they buy, they select suppliers. The user receives a response, not a list of links. And in many cases, accept that answer without reviewing alternatives.

This is the context in which a new concept was born: assistive agent optimization, or AAO.

What does AAO mean and where does it come from

AAO is a term proposed by Jason Barnard, a consultant who specializes in how brands are represented by artificial intelligence systems. The logic behind the name is simpler than it seems.

“Assistive” describes the purpose of the system: to assist the user. “Agent” describes the actor who performs that assistance, not a search engine that shows results, but an agent that makes decisions and acts. “Optimization” is what we do, working to make that agent choose us.

Barnard's argument is that the other terms that circulate in the industry, GEO, AEO, AIEO, LLM optimization, AI and SEO, describe parts of the problem but not the entire problem.

GEO focuses on the generative component but ignores knowledge graphs. Entity SEO covers knowledge graphs but requires explanation for anyone outside the technical world. LLM optimization covers a third of the system and ignores the rest.

AAO attempts to cover the three components behind any AI system that makes recommendations: language models, knowledge graphs, and traditional search. Barnard calls this the algorithmic trinity, and argues that any strategy that optimizes for just one of the three components will leave gaps.

Why the distinction between motor and agent matters in practice

The difference between a search engine and an agent isn't just semantic. It fundamentally changes what the system does and what it means to appear in it.

A search engine presents options. The user evaluates and decides. Your job as a brand was to appear on that list and make the user choose to click on you.

An agent evaluates and decides for the user. The user receives a response. In the best case, you receive a short list. More and more often, you receive only one recommendation. Barnard describes this as the “perfect click”, the zero-sum moment in AI where the system presents a solution and most users accept it.

That means that the full funnel, awareness, consideration, decision, happens inside the agent before the user sees anything. The agent learns that your brand exists, considers you alongside alternatives, and decides, all internally. If you are not present and well represented in that internal process, you do not arrive at the result.

For marketing teams, that changes the objective. For decades, the goal was to attract visitors to the site and convert there. Under the logic of AAO, the goal is to be the response that the agent delivers, before the user reaches your site, or without it.

The three structural changes that AAO involves

1. Brand identity becomes infrastructure

When an agent recommends a hotel, a software provider, or a consultant, they don't analyze individual pages looking for well-optimized title tags. Evaluate what you know about the brand: what it does, who it serves, why it would be a reliable solution for this specific case, and how sure you are of that data.

That trust is built from what Barnard calls the “entity home”, the page that anchors everything the system knows about your brand, and expands through every external source that corroborates that information. Media that mention you, directories that list you, reviews that describe you, social networks that identify you.

If the agent has a confusing or incomplete picture of your brand, they'll choose a brand they understand better. Not because that brand is better, but because uncertainty comes at a cost to AI systems and agents tend to minimize it.

2. The web index is no longer the only input channel

For two decades, the logic was simple: if Google didn't index it, it doesn't exist. That monopoly is being broken on two fronts.

On the one hand, agents are increasingly consulting sources that never go through a traditional web index: APIs, internal databases, structured feeds, reservation systems. If your information only lives on conventional websites, there are distribution channels that you are not using.

On the other hand, push mechanisms are emerging, such as IndexNow, MCP and other similar protocols, that allow you to send structured information directly to the systems that act, instead of waiting for them to come and get you.

Barnard describes it as going back to the 90s in certain aspects: actively sending information to the ecosystem instead of publishing and waiting.

There's one technical point worth mentioning for teams working with developers: most AI agent bots don't render JavaScript. If important content on your site depends on client-side JavaScript to be displayed, a growing number of agents simply don't see it. That's a visibility issue that doesn't show up in traditional SEO reports.

3. SEO skills are still valid, but the goal moves

AAO doesn't replace SEO. It contains it. The technical skills, the content structure, the link work, the schema markup, all of that is still relevant. What changes is the ultimate goal.

In SEO, the objective was to be found by the engine and chosen by the user. In AAO the objective is to be chosen by the agent when acting, recommended when the user investigates, and mentioned when the user asks.

There are three levels of visibility with varying degrees of system autonomy, and optimizing for all three requires thinking beyond traditional search rankings.

Why Conceptual Clarity Matters Now

There is a pragmatic argument behind adopting a clear conceptual framework for this work, beyond the discussion about which acronym is more accurate.

The data that Barnard cites about the concentration of visibility in AI systems are illustrative. In December 2024, the sites with the best citability performance within AI systems captured about 31% of the total. By February 2026, that percentage had risen to 59.5%. Within two months, the concentration nearly doubled.

This suggests that AI systems are consolidating their trust around an increasingly select group of sources, and that the cost of being late for this work is not static. Those starting now aren't starting at the same point as those who started six months ago.

The discussion about whether to call it AAO, GEO, AIEO, or anything else is secondary. What matters is to understand that optimizing for systems that recommend and act is a different job than optimizing for systems that show lists, that this work has specific technical and editorial components, and that the window for building advantage before concentration is consolidated is still open.

What this means for a marketing team today

Not everything has to change at the same time. But there are three questions worth asking now.

The first is whether the information about your brand is consistent and complete in the sources that AI systems consult. Not just your website, but media, directories, review platforms, social networks. If there are inconsistencies or gaps, agents work with partial information.

The second is if your most important content is accessible to bots that don't render JavaScript. If it's not, there's a visibility issue that traditional SEO isn't measuring.

The third is if you have visibility into how AI systems represent your brand today. What they say when someone asks about your category, who appears as an alternative, what attributes they associate with you. Without that information, it's hard to know what needs to be improved.

AAO as a discipline is still under construction. Complete methodological frameworks, such as the ten-stage pipeline that Barnard describes in his work, are now being documented. But the problem you're trying to solve already exists and it's already affecting which brands appear and which don't in the systems that more and more people use to make decisions.