Live - tracking ChatGPT, Gemini & ClaudeNew - Google AI Overviews & DeepSeek on the roadmapFree audit - see how AI sees youEvery metric ships a 95% confidence range
Guides · Article

Discoverability vs. selectability: the two gates of an AI citation

Every AI citation clears two separate gates. Confusing them is the most expensive mistake in GEO. Here is how to tell which one is costing you answers.

When your brand is missing from an AI answer, there are two completely different reasons why — and the fix for one does nothing for the other. Getting this distinction right is the difference between effort that moves citations and months spent polishing the wrong thing.

Two gates, not one

Before an engine like ChatGPT, Gemini or Claude writes an answer, it runs a retrieval step: it pulls a pool of candidate pages, then reranks them and quotes a small handful. A citation has to clear both steps. We call them discoverability and selectability.

Gate 1

Discoverability

Is your page even in the retrieval pool? Driven by off-site authority and search-index coverage — the same forces that decide whether a source shows up at all. No on-page polish helps if you are never retrieved.

Gate 2

Selectability

Once retrieved, does your page survive the rerank and get quoted? Won on-page: clean passages, fact density, structured data, freshness, speed.

Selection is the real bottleneck

Most people assume getting retrieved is the hard part. It is not. Roughly 85% of the pages an engine retrieves are never cited (Ahrefs, via Quattr) — meaning selectability, not discoverability, is where most pages die. Being in the pool is necessary but nowhere near sufficient: the rerank throws out the overwhelming majority of what it considered.

~85%
of retrieved pages are never cited — the selectability gate is brutal (Ahrefs)
76% → 38%
of AI-Overview citations from top-10 organic, year over year — discoverability is decoupling from rank (Ahrefs)

And discoverability is changing fast in its own right: only 38% of Google AI-Overview citations now come from the top-10 organic results, down from 76% a year earlier (Ahrefs). Ranking #1 no longer guarantees you even make the pool. Both gates are moving — which is exactly why you have to know which one is failing. Treat these figures as correlation, not proven cause: domain authority and brand search volume are confounders, and the numbers swing month to month.

How to tell which gate is failing

Compare cited vs. not-cited sources that appeared for the same query. If strong, well-structured pages on your domain never enter the pool, you have a discoverability problem — earn third-party mentions and fix index coverage. If you are retrieved but never quoted, that is selectability — restructure passages and raise fact density. The two diagnoses lead to opposite fixes, which is why guessing is so expensive.

One product note, then back to the work: CitedOS's Source-Gap Analyzer is built around this split, labelling which gate each gap sits behind so you spend effort on the one that is actually failing. The diagnosis above stands on its own, though — for the on-page (selectability) side, see GEO tactics that work; for the wider research picture, what holds up in AI-search research.

Sources

See where you stand in AI search — in 30 seconds.