Your customers ask AI now. Can they find you?
Search split in two — Google ranks pages, AI engines pick sources to name and cite. Here is what that means and how to measure your side of it.
Something structural happened to search, and it didn't arrive with an announcement: your customers now ask ChatGPT, Claude, Gemini and Perplexity before — sometimes instead of — asking Google. And those engines don't return ten blue links. They compose an answer, name a few brands, and cite a handful of sources.
That changes what "being visible" means.
Two surfaces, two games
Classic SEO is a ranking game: get your page into the top results for a query, and you exist. AI answers are a selection game: the engine decides which brands to mention and which sources to cite while composing the answer. A page can rank #1 on Google and never be cited by an AI engine; another site can be quoted constantly without ranking at all.
Neither surface is going away. That's why we built AIEO.ee to measure both from one workspace — keywords for Google, prompts for AI answers, and a single Growth Score that reads both.
What you can actually measure
The AI side sounds fuzzy until you decompose it. For each question your buyers really ask, run against each engine, three numbers matter:
- Mention rate — of all the answers generated, how many name you at all.
- Citation rate — how many link your domain as a source. Mentioned ≠ cited: an answer can name you while citing a competitor's page.
- Share of voice — your mentions versus your tracked competitors', zero-sum. Someone's gain is someone's loss.
Engines overwhelmingly cite pages that directly answer the question — comparison tables, definitions, step-by-step guides — not homepages. That is good news: it's a content-structure problem, and content-structure problems have a playbook.
The honest part
Measuring AI answers means asking an LLM what an answer engine would say, and that has to be labeled. Three engines in AIEO.ee support a live backend (OpenAI web search, the Perplexity Sonar API, Gemini grounding) — and even for those, a run only counts as live when the backend actually returned real search evidence. Everything else is labeled sampled — on every run, so you can always check what a number is made of. We wrote a whole post about how that labeling works.
Directional? Yes. Directional and tracked over time beats anecdotal and guessed — especially when the numbers share one score with your Google side.
Start small
Add the five questions your buyers actually ask. Run them daily across a few engines. Read the citation table. In a week you'll know more about your AI visibility than most of your category learns this year.
The quickstart takes about ten minutes.