GEO and AEO, and getting cited by AI

That last line is the whole practical difference. Search engines index documents. Answer engines chunk them, embed the chunks, and pull back whichever chunk best matches the question, which means a paragraph that only makes sense in the context of the four paragraphs above it gets retrieved and then thrown away as useless.

Writing for AEO means writing passages that stand alone. Name the subject in the sentence rather than relying on the heading. Resolve the question before you explain it. This page follows that rule, which is why every answer on konigi.com names Michael Angeles instead of saying “he.”

GEO and AEO get used almost interchangeably. The distinction people draw is mostly about which surface they mean: ChatGPT and Perplexity on one side, Google’s AI Overviews on the other. I don’t think the vocabulary is settled enough to be precious about it.

What matters is that the old lever, keyword density, does close to nothing here, while the things SEO treated as nice-to-have (a clear entity, a named method, third-party mentions) turn out to do most of the work.

The on-site half is mechanical and you can finish it in a week: JSON-LD structured data, stable entity identifiers, clear headings, explicit takeaways, signed bylines, a robots.txt that doesn’t accidentally block the AI crawlers. None of it is hard. None of it is optional.

The off-site half is slower and it decides the outcome. Writing for other publications, being quoted, having your work shared by people the models already trust. A perfectly marked-up site with no external signal is a well-organized nobody.

This section of konigi.com is that method applied to itself, which is the only honest way to sell it.

AEO work isn’t listed with a fixed rate on the Konigi services page because the shape varies more than the other work does. A single assessment is a different animal from a monitoring retainer, and a content engagement depends entirely on how much there is to write and how much of it already exists in some form.

Tell me what you’re working with. I’ll come back with a shape and a number.

Concretely, a Konigi assessment audits your robots.txt and crawler access, your structured data and whether the entities in it resolve to stable identifiers, your heading structure and whether answers appear before the reasoning, and your off-site footprint.

You won’t get a number out of a hundred. Those exist to make a report feel objective, and they collapse a dozen unrelated problems into one digit you can’t act on. You get a ranked list instead. It tells you what to do Monday.

The part of schema that genuinely earns its keep is entity disambiguation. A Person with a stable @id, a filled-in sameAs list, and knowsAbout terms that point at real definitions tells an engine which Michael Angeles wrote this and what he can be trusted to know. An FAQ block hoping for a rich result is a different thing entirely.

Worth knowing: Google restricted FAQ rich results to government and health sites back in 2023. That was three years ago. Anyone still selling you FAQ schema for the SERP accordion is selling something that doesn’t exist.

Server logs help too. The AI crawlers identify themselves, so you can see whether GPTBot and PerplexityBot are actually fetching your pages before you go looking for why ChatGPT isn’t quoting you.

Be honest about the limits of this. Answers vary between runs, personalize by account, and change when a model updates. What you get is a trend line and a competitive picture, not a rank you can chart to two decimal places. Anyone promising precision here is guessing.