In a recent briefing, Addy Osmani, director of engineering at Google Cloud AI, introduced a fresh framework called Agentic Engine Optimization (AEO). The goal is to make web content more accessible and useful to autonomous AI agents—software that fetches, parses, and acts on information without human interaction. AEO is not a replacement for traditional Search Engine Optimization (SEO); instead, it operates alongside it, addressing the unique needs of machine readers that can perform complex tasks in a single request.
What Agentic Engine Optimization Means
AEO is a set of guidelines that help creators design pages so that AI agents can quickly locate and understand the information they need. Unlike human users, agents do not scroll, click, or engage with the user interface. They send a request, receive a block of text, and then decide whether that text satisfies their goal. Because of this, conventional engagement metrics—time on page, bounce rate, scroll depth—become less relevant. Instead, the focus shifts to how efficiently an agent can extract the desired answer.
Why Token Limits Are the New SEO Metric
One of the core insights Osmani highlighted is the token constraint that many AI models impose. A token can be as short as a single character or as long as a word, but the total number of tokens that a model can process in one go is limited. When a page exceeds that limit, the agent may receive truncated content, skip entire sections, or even hallucinate facts that aren’t present in the source material.
These issues translate into three concrete problems for content creators:
- Truncated information: Key details are cut off, leaving the agent with an incomplete answer.
- Skipped pages: If a page is too large, the agent may ignore it altogether, assuming it’s irrelevant.
- Hallucinated outputs: The agent might generate plausible but incorrect statements because it has to fill gaps in the data.
Because of these challenges, token count has emerged as a primary optimization metric in the AEO framework. By keeping content within the token window, creators can ensure that agents receive the full context they need.
Practical Tips for AI-First Content
Osmani’s guidance offers a clear recipe for structuring pages that cater to AI agents:
- Answer first: Place the most important answer within the first 500 tokens of the page. This mirrors how search engines display featured snippets and ensures the agent can find the core information immediately.
- Compact and focused: Avoid sprawling narratives. Each page should tackle a single topic or question, reducing the risk of exceeding token limits and improving clarity for the agent.
- Short preambles: Long introductions or background sections can bury critical insights. Keep the opening concise and to the point.
- Clear headings and subheadings: Use descriptive headings that signal the content’s purpose. AI agents can use these cues to navigate the page quickly.
- Structured data:
Leave a Comment