Skip to main content
πŸ”¬ SEO RESEARCH Β· AI CITATION STUDY

Which Sources Do AI Assistants Cite?

This is an ongoing study, not a finished one. We're publishing the methodology now and will publish findings when the sample is large enough to support conclusions. We're deliberately not putting an invented headline statistic on this page β€” a precise-sounding number we haven't actually measured would undermine the entire reason we publish research at all.

SECTION 01

What we're measuring

Across the US, UK, Canadian and Australian markets, for a fixed set of commercial and informational prompts in the marketing-services category: - Which domains are cited by ChatGPT, Perplexity and Google AI Overviews - Whether cited sources skew toward established publishers, vendor sites, or community content - Whether the same prompt returns consistent sources across repeat queries - Whether market (US vs UK vs CA vs AU) changes which sources are surfaced

SECTION 02

Why the market variable matters

AI answers appear to be geographically personalised to some degree, which means a business optimising for citation needs to know whether it's being cited in its own market or only elsewhere. This is under-examined publicly and is the part of the study we think adds most.

SECTION 03

What we can say now

From the prompt testing we've run so far, the consistent qualitative observation is that pages structured with a direct answer near the top, clear definitions, and comparison tables are cited more often than equivalent-quality content written in continuous prose. That aligns with what's broadly reported elsewhere, and it's the basis of the structural recommendations in our GEO guide.

We're not attaching a percentage to that yet, because we don't have one we'd stand behind.

SECTION 04

Methodology limitations we'll state upfront

  • [+]Single vertical (marketing services), which limits generalisability
  • [+]AI platforms change frequently, so findings have a short shelf life
  • [+]Prompt phrasing materially affects results, so results are sensitive to test design

Read our GEO guide β†’

Written technical findings by senior strategists. Published transparent pricing.

Read our GEO guide β†’