Glossary
How SEO, GEO, AEO and AIO relate
There are only two axes: SEO and GEO. AEO is formatting work done inside those two axes, and AIO is an outcome that follows once the conditions are in place. Proposals often list all four side by side, which makes something no one can sell (AIO) look like a product.
- 01Why the four don’t belong on one line
- 02First-generation programmatic SEO vs. today
- 03Common misconceptions
Relationship
Why the four don’t belong on one line
Four boxes side by side make all four look like things you can buy. The last one isn’t sold, though; it follows once the others are in place.
1Only two axes
SEOCrawled, indexed and evaluated
Technical health and intent matching. A document that isn’t indexed can’t make it into any answer.
GEOCited by AI
Getting a brand recognized as a single entity, with original data, sourcing and outside evidence in place. It’s about eligibility, not phrasing.
This is what DATAPREPREP specializes in. The G stands for Generative, not Geographic.
Work done inside both axes
2Execution
AEOExcerpted into answer boxes
Question-style headings with a direct answer up front (100–160 characters on Korean pages). It’s where a document with SEO and GEO in place moves into answer placements.
It’s formatting work, so it costs almost nothing.
What follows once conditions are met
3Outcome, not directly buildable
AIOOutcome metric
The stage where a document that meets AEO requirements is cited as a source in generative AI answers such as ChatGPT search and AI Overviews.
An outcome that appears once the conditions are met. What DATAPREPREP builds is those conditions.
E-E-A-T Experience, Expertise, Authoritativeness and Trustworthiness isn’t an item in any one box; it runs through all four. SEO and GEO are the only axes, and documents reach AIO by way of AEO. DATAPREPREP specializes in GEO, and E-E-A-T runs through all four.
Question for choosing the axis
Where does your traffic come from today?
If it comes from search, SEO work comes first. Doing GEO work before pages are indexed puts the steps in the wrong order.
Question for choosing the execution
Is the document easy to excerpt?
Question-style subheadings and direct answers are formatting work, so they cost almost nothing. This is the first place to improve.
Question about the outcome
What is the proposal committing to?
A proposal that asserts placement in AI Overviews has nothing to back it up. What can be committed to is the conditions, not the result.
Generations
First-generation programmatic SEO vs. today
Both look like “making lots of pages,” but they optimize for different things. The first generation targeted keyword crawlers; today’s work targets citations in answer engines.
| Aspect | First-generation programmatic SEO | Today (graph- and GEO-based) |
|---|---|---|
| Core asset | Relational database + fixed templates | Entity relationship map + terminology system |
| How it works | Variable substitution and rule-based publishing | Entity relationship mapping + answer-sized composition |
| Optimizes for | Search engines’ keyword crawlers | Citations and indexing in answer engines |
| Unit of writing | One page per keyword | One direct answer per question (an excerptable paragraph) |
| What decides success | Page count and automation speed | Indexing rate and eligibility for citation (consistent naming, evidence) |
| How it fails | Thin pages get filtered out of the index | Scattered entities keep signals from adding up |
The first generation wasn’t wrong. As search moved toward answers, the target changed, and the assets and units of work changed with it. DATAPREPREP works from the right-hand column.
20 terms
Common misconceptions
The misconception column lists only what DATAPREPREP has actually seen in meetings and proposals.
| Term | Definition | Common misconception |
|---|---|---|
| SEOSearch Engine Optimization | The axis that makes a site crawlable, understandable and assessable for search engines. Most traffic still comes from here. | Saying SEO is over now that AI search has arrived. Crawling and indexing come before any answer, so they don’t go away. |
| GEOGenerative Engine Optimization | The axis that gets generative answers to cite a brand. It’s a question of eligibility for citation, not of phrasing. | Treating it as the GEO of geographic search. The letters match; the meaning doesn’t. |
| AEOAnswer Engine Optimization | Formatting work that lets answer boxes in search results pull a paragraph from your page: question-style subheadings plus a direct answer. | Calling generative AI answers such as ChatGPT’s AEO as well. Generative answers belong to GEO. |
| AIOAI Overviews | Google’s AI summary at the top of search results. An outcome that follows once the conditions are in place. | Selling AIO appearances as a product. No one can create them directly. |
| E-E-A-T | Experience, Expertise, Authoritativeness and Trustworthiness. An evaluation standard rather than a task, and it runs through all four terms above. | Assuming an author bio is enough. It needs evidence that can be verified outside your own site. |
| Indexing | A search engine storing a document in its index. It’s a separate step from crawling, so a crawled document can still be left out of the index. | Assuming pages are indexed because a sitemap was submitted. Submission is a request, not a result. |
| Indexing rate | The share of submitted pages that a search engine has actually indexed. Neither Google’s nor Naver’s documentation gives a normal range or a threshold. | Judging a whole site by that one ratio. Not every page gets indexed, so check first whether your key pages are in. |
| Entity-based design & expansion | Listing the entities a search engine should recognize, then building a separate page for each row of that list. It combines getting entities recognized and publishing pages into one workflow. | Reading it as mass-producing pages. The order is the reverse: entities and data come first, and pages are the result. |
| Entity SEO | Getting search engines to identify a brand, place or concept as a single entity and connect it to others. The target is a place in the Knowledge Graph, not a ranking. | Equating it with adding more keywords. The target is the entity, not the string. |
| Topical authority | Being trusted on a topic. It builds up on an entity once the entity is recognized, and E-E-A-T signals accumulate there too. | Thinking it grows with the number of posts. What counts is how much of the topic you answer, not the post count. |
| Entity | A single thing a search engine recognizes. If the business name, description and contact details vary across channels, they won’t merge into one entity. | Using a different brand name on each channel and expecting mentions to add up. |
| Zero-click | A search that ends on the results page without a click. That’s why the goal shifts from clicks to citations. | Reading fewer clicks only as a drop in rankings. |
| UGCUser-generated content | In Naver’s context: Blog, Cafe (community forums), Knowledge iN (Q&A) and Premium Contents (paid subscriptions). Naver Place, its local business listings, isn’t included. | Stretching UGC to mean online communities in general. The scope Naver described is those four services. |
| AI crawlers | Crawlers AI services use to collect documents. AI search crawlers that fetch sources for answers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) and AI training crawlers (GPTBot, ClaudeBot) have separate user agent names, and Google-Extended is a control token that decides whether content is used for training. | Allowing or disallowing AI training crawlers and AI search crawlers with a single rule. robots.txt lists the two types separately. |
| Structured dataSchema, JSON-LD | Markup that describes a page’s structure to search engines. It remains useful for rich results. | Counting it as a tactic for AI citations. Google has officially stated it isn’t required for generative AI search. |
| llms.txt | A proposed format a site can use to describe its structure to AI agents. | Selling it as a search visibility tactic. It’s meant for agents, not for search. |
| Uniqueness rate | Industry shorthand for how different a page is from other pages. Neither Google’s nor Naver’s documentation sets a percentage; what they describe is the value a page adds for readers. | Mass-producing pages that swap only synonyms or place names. Google calls this scaled content abuse, and Naver describes it as mass-publishing pages that change only parts of sentences. |
| Informational query | A query that asks for a definition, method or procedure. For this type, documents outside the top search results have been observed being cited. | Applying the same approach to commercial queries (recommendations, rankings, comparisons). Their citation sources overlap almost entirely with the top search results. |
| Mention | A brand named inside a sentence rather than linked. Mentions that build up elsewhere on the web become entity signals. | Committing to a number of mentions and manufacturing them. Google’s guide to generative AI features lists seeking inauthentic mentions among the things that don’t help. |
| Programmatic SEOpSEO | The industry name for building pages at scale from one template and a data list. The expansion side of our work falls here. | Reading it only as mass-producing pages. It works only when each page has its own real data. |
In DATAPREPREP documents, each statement shows whether it comes from official documentation we checked directly or from third-party data we cite.
We build to the standards written here. First, we check where your current website is getting stuck.
Free audit