getcited

GetCited Labs · a research imprint

We run experiments on what AI engines actually cite.

ChatGPT, Claude, Perplexity and Google AI Overviews cite different sources for the same question, and most of what gets claimed about "AI SEO" is asserted, not measured. We run controlled prompt batteries against our own measurement engine and publish what we find — the method, the raw numbers, and the caveats that keep us honest. One experiment per post.

We sell nothing here. GetCited Labs is not a lead-gen funnel for a service. We are also not a neutral party — Lead Media operates sites in some of the markets we measure, including verdikt.bet, a Canadian iGaming comparison site. Every experiment names which of our properties, if any, sit inside its data. See who runs this and our disclosure policy.

From the latest experiment

310

runs across 20 prompts and 4 AI surfaces, zero errors — exp-012, 08/07/2026.

1,584

individual citations captured and reduced to domains across those 310 runs.

8 / 211

domains cited by ChatGPT, Claude and Perplexity alike, on the prompts all three actually answered.

Every number above traces to one published experiment — we don't publish a figure we can't point you to. See the full write-up.

How we work

Fixed prompt batteries

A locked set of prompts, run N times per AI surface, with every cited URL captured — not sampled after the fact.

Like-for-like comparisons only

Engines don't answer every prompt. Cross-engine tables are computed only on the prompts every included engine actually answered.

Excluded rather than invented

If a surface's data can't be trusted — see the Google AI Overviews exclusion in exp-012 — we leave it out and say why, rather than publish a number we can't defend.