> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trychainshift.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Understanding AEO/GEO

> Search optimization in the age of AI answer engines

This page is background reading for anyone new to AEO/GEO.

## Search is changing

Instead of searching Google for "used car prices", people have started asking ChatGPT questions like this.

> "Recommend a used SUV in the 20 million won range"

The AI does not give you ten links. It gives you **one answer**. It recommends specific models, mentions specific platforms, and cites specific web pages as evidence.

The shift shows up in the data too. On Naver, search volume for "used car" dropped by more than a third over the past year.
(Example of measured values from [Naver trend data](https://docs.trychainshift.ai/en/guides/keywords.md#naver-search-volume-trend))

## What AEO/GEO is

**AEO** (Answer Engine Optimization), or **GEO** (Generative Engine Optimization), is the work of getting your brand into that "one answer". There are two goals.

1. **Mention**: the AI names your brand or product directly in its answer
2. **Citation**: the AI links to your web page as evidence for its answer

If traditional SEO was about "getting onto page one of the search results", AEO is about "getting into the sentences of the AI's answer".

## How AI builds an answer

To measure anything, you first need to know how the AI generates an answer. Looking at real collected data:

1. The user asks in conversational language: *"Tell me what to watch out for when buying a used EV"*
2. The AI internally **creates and issues fan-out queries**: *"used EV purchase precautions battery health check"* (these are the AI's search terms, not a human's)
3. It reads a few of the web pages it found and builds an answer while **citing** them
4. Specific brands and products are **mentioned** in the answer

Each point in this pipeline is a measurement point, and each maps 1:1 to the structure of this API.

## What you measure: questions mapped to the API

| What you want to know | Measured example (used car market) | API to use |
| - | - | - |
| What answers the AI actually gives | ChatGPT recommends the Seltos and Tucson and cites 2 specific blogs | [Retrieving answers](https://docs.trychainshift.ai/en/guides/collecting-answers.md) |
| Which sites get cited (and do we?) | Of 800 citations, 264 are naver.com | [Cited domain stats](https://docs.trychainshift.ai/en/guides/analytics.md#cited-domain-ranking) |
| Which sentences get cited | "Tips for buying popular used cars by price range…" (magazine-style copy) | [Cited URL stats + quotes](https://docs.trychainshift.ai/en/guides/analytics.md#cited-url-ranking) |
| What search terms the AI issues (what should we create?) | "used EV purchase precautions battery health check" | [Fan-out stats](https://docs.trychainshift.ai/en/guides/analytics.md#fan-out-query-aggregation) |
| Is there demand or opportunity in that term | "used car prices": difficulty 18, +22% quarterly growth | [Keyword data](https://docs.trychainshift.ai/en/guides/keywords.md) |

## Why repeated measurement matters

AI answers are **non-deterministic**. Ask the same question twice and you can get different answers.\
So "is our brand mentioned?" is not a yes or no, it is a **probability**. This is why a run in this API can repeat the same prompt and model combination up to 100 times, and why visibility is measured as "how many mentions out of 100".

Models also behave differently. In one measured example, Perplexity cited YouTube content 38 times while ChatGPT cited it once. Your strategy depends on which model you want to be visible in. See [Supported models](https://docs.trychainshift.ai/en/models.md).

## Terminology

| Term | Meaning |
| - | - |
| Answer engine | A service like ChatGPT, Perplexity, or Google AI that gives answers instead of links |
| Prompt | A question a user might plausibly ask an AI. The unit of monitoring |
| Run | A collection job that runs prompts against multiple AI models |
| Citation | A web page the AI presented as evidence for its answer |
| Fan-out | Search queries the AI issued internally to build its answer |
| Mention | A brand or product name appearing in the answer body |

## Get started

> [card] [Quickstart](https://docs.trychainshift.ai/en/quickstart.md)
>
> Your first API call in a minute

> [card] [Running prompts](https://docs.trychainshift.ai/en/guides/submitting-runs.md)
>
> Run your first monitoring collection


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.