Not retrieved, retrieved but not cited, cited but not recommended. Almost every AEO problem is one of these, and treating them as one problem is why most fixes fail.
Why one label is not enough
"We are not visible in AI" is a symptom, not a diagnosis, and it covers three separate mechanical failures that happen at different stages and need opposite responses. Shipping a fix without knowing which one you have is guessing.
The three are sequential. Retrieval has to succeed before citation is possible; citation has to happen before recommendation is on the table. Find the earliest stage that is failing and fix that, because fixing a later stage while an earlier one is broken changes nothing.
Not retrieved — the engine never reached your content. Crawler access, rendering, or simply no content on the topic.
Retrieved but not cited — the engine reached it and did not use it. Usually a corroboration or answer-shape problem.
Cited but not recommended — the engine used you as a source and still endorsed someone else.
Localising which one you have
Not retrieved shows up as total absence — your domain appears nowhere in the citations, including on brand prompts where you should be the obvious source. Check crawler access first; content that requires JavaScript to render, or a robots configuration that excludes AI crawlers, produces exactly this pattern.
Retrieved but not cited looks like your domain appearing occasionally but never on the prompts you care about. The content exists and is reachable but is not what the engine wants for that question — usually because it does not answer the question in a directly extractable form, or because nothing else on the web corroborates the claim.
Cited but not recommended is the most advanced problem and the most commercially frustrating. Your content is good enough to inform the answer and something else is winning the endorsement — typically because the competitor has independent corroboration you lack.
Where this lives in SolCrys
Crawler access has a dedicated surface — check this first if you suspect a retrieval problem.
The citations view with the owned filter tells you whether your domain is being retrieved at all.
Recommendation is scored per response, separately from presence, which is how you see the third failure mode distinctly.
Lab: Name your failure mode
Needs a free SolCrys account (no credit card). No quota consumed.
Check whether your domain appears at all in your citations.
If it does not, check crawler access before anything else.
If it appears but not on your priority prompts, you are in mode two.
If it appears on prompts where a competitor is still recommended, you are in mode three.
Write down which mode you are in. The rest of the course assumes you know.
You rank well and still aren't cited in AI answers. Here are the five gaps SEO leaves — per-engine citation, corroboration, extractability, accuracy, and measurement — and how to close each.
Getting cited as a source and getting recommended as the pick are two different AI-search outcomes, and they are decoupling. Why most teams track the wrong one, what a 100-query study and our own 5-engine data show, and the KPI to measure instead.
Turn AI answer gaps into governed marketing execution.
Start free with a ChatGPT visibility read, then add multi-engine tracking, Corporate Context governance, and the action-to-result loop when you are ready.