Who AI Actually Cites for AEO Questions: 637 Answers, 980 Domains

If you sell AEO — or buy it — the first question is simple: when someone asks an AI engine about answer-engine optimization, which sources does it actually cite? Not in theory. In the answer.

So we measured it. We ran 90+ real AEO/GEO questions — "what is AEO", "how to get cited by ChatGPT", "best AI visibility tools", "AI citation service", "how to make my brand an entity for AI" — through ChatGPT Search, Claude (web search) and Google AI Overviews, with repetitions, and logged every citation in every answer. Here's what the engines cite.

The headline numbers

• 637 AI answers measured across 3 engines.
• 2,764 citations made, to 980 unique domains.
• The engines disagree sharply on whether to cite at all:

| Engine | Answers that cited ≥1 source | Citations per answer (avg) | |---|---|---| | ChatGPT Search | 95% (259 / 273) | ~4.9 | | Google AI Overviews | 90% (82 / 91) | ~10.0 | | Claude (web search) | 39% (107 / 273) | ~1.9 |

The single most important takeaway: ChatGPT and AI Overviews cite almost every time; Claude cites less than half the time. If your AEO strategy optimises for one engine's citation behaviour, you are mis-measuring the other two. They must be tracked separately — never collapsed into one "AI visibility" number.

Who gets cited for AEO

The long tail is long — 980 domains — but it is dominated by a small head. The most-cited domains across all 637 answers:

| Domain | Citations | Type | |---|---|---| | youtube.com | 140 | Video / UGC | | linkedin.com | 62 | Professional / UGC | | google.com | 58 | Platform | | reddit.com | 51 | Community / UGC | | aeoengine.ai | 30 | AEO vendor | | en.wikipedia.org | 26 | Knowledge graph | | techradar.com | 25 | High-DR publisher | | searchengineland.com | 23 | Industry publisher | | rankability.com | 23 | AEO vendor | | semrush.com | 21 | SEO platform | | xseek.io | 20 | AEO vendor |

Three patterns jump out:

1. User-generated and community media own the top. YouTube, LinkedIn, Reddit and Wikipedia are four of the top six. This matches the broader finding that ~90% of AI citations come from third-party media, not a vendor's own marketing site. If you are not present on those surfaces, you are absent from the answer.

2. A handful of high-DR publishers carry the editorial citations. TechRadar, Search Engine Land, Semrush, HubSpot and Conductor recur across clusters. These are the "best AEO tools" listicles and explainer articles the engines pull from — which is exactly why earned inclusion on them is the highest-leverage off-site move in the category.

3. Smaller vendors are already winning citations. aeoengine.ai, rankability, xseek, llmpulse, therankmasters, amicited and others — none of them household names — are cited repeatedly. The category is young enough that the citation bar is reachable: it's earned through presence in the surfaces above, not through brand size.

Why the engines disagree on sources

In our data, ChatGPT and Claude cited almost no overlapping domains for the same AEO questions — each reads the web through its own retrieval pipeline. ChatGPT leans on a broad set of publishers and UGC; Claude cites rarely but, when it does, narrowly. Google AI Overviews cites the most per answer (~10) and skews toward established publishers and Google's own surfaces.

The practical consequence: a citation on ChatGPT tells you almost nothing about your Claude or AI-Overviews standing. Measuring one engine and assuming the rest is the most common mistake in AEO — and the easiest to avoid once you see the divergence in the numbers.

Methodology

90+ AEO/GEO prompts spanning definitional, commercial, comparison, agency and entity intent, each run with repetitions across ChatGPT Search, Claude (web search) and Google AI Overviews on 25 June 2026 — 637 answers in total. Every answer's full text and citation list was stored with a timestamp. Citations were normalised to registrable domains. We report the domains the engines actually cited; we do not rank vendors or pass judgement on them. This is point-in-time data — AI answers change, and so will these numbers. We re-measure on a schedule.

The measurement was run with the same GetCited prompt-battery engine we use to measure citation movement for any brand. If you want the same read for your own category — which engines cite you, which cite your competitors, and which domains you'd need to be on to change that — that's exactly what a baseline scan produces.