GEO is a measurable discipline, not a hunch. Princeton and Georgia Tech published the foundational research at ACM KDD 2024. Seer Interactive and Searchlab.nl track the industry data. Stareo is built on both.
LLM visibility lift from GEO techniques
Princeton/Georgia Tech
citation lift from expert quotations alone
arXiv:2311.09735
CTR drop on AI Overview queries
Seer Interactive
Princeton and Georgia Tech ran controlled GEO experiments across 10 domains and 9 question categories. Published at ACM KDD 2024 — the premier data mining venue.
LLM visibility lift
Techniques including statistics, quotations, and fluency improvements increased citation rates across GPT-4, Claude, and Gemini in a controlled 10-domain study.
Princeton + Georgia Tech, ACM KDD 2024→citation lift from expert quotations
Adding attributed expert quotations was the single highest-impact GEO technique tested in the study — more impactful than statistics or fluency alone.
arXiv:2311.09735, Table 3→Industry studies tracking real changes to search behaviour since AI Overviews launched.
drop in organic CTR
when a Google AI Overview answers the query instead of showing a ranked list of links.
Seer Interactive, 2025
overlap between top-10 and AI citations
Top-10 Google rankings and AI citation sets overlapped at 75% twelve months ago. Now: 17%.
Searchlab.nl, 2026
more organic clicks
for brands cited in AI Overviews compared to brands that rank in the top 10 but aren't cited.
Seer Interactive, 2026
clicks per impression
for pages cited in AI Overviews vs. pages that simply rank in the top 10.
Seer Interactive, 2026
of AI Overview clicks go to the first citation
Position in AI citations follows a power-law distribution. First is overwhelmingly dominant.
Seer Interactive, 2026
average Share of Model baseline
Brands new to GEO measurement typically find AI mentions their brand in fewer than 1 in 8 relevant queries.
Stareo customer baseline data
Every technique Stareo uses is traceable to a published finding or measured outcome.
“Expert quotations are the #1 citation signal”
Writer agent identifies top entities in your domain and generates attributed quotations from existing expert sources. Every brief includes 2+ expert quotes.
“Statistics increase citation probability”
Passage Extractor flags content with missing quantitative support. Writer drafts include at least one cited statistic per major claim.
“Fluency and readability matter to LLMs”
ComplianceOfficer includes a fluency pass using Haiku 4.5. Content that reads like AI-generated slop is rejected before it ships.
“Topical authority clusters drive AI citation volume”
The Strategist builds a topical graph. Missing subtopics become content briefs. The goal is authoritative coverage depth, not keyword density.
“Knowledge Graph signals amplify AI recognition”
KnowledgeGraphMaintainer builds Schema.org JSON-LD for every entity — Organization, Person, Product — and syncs to Wikidata where possible.
“Passage-level quality determines what gets extracted”
PassageMatcher tracks which passages AI engines quote. Writer optimises underperforming passages in counterfactual experiments.
Sources
Start a free trial and Stareo runs your first Share of Model measurement within 24 hours of connecting your site.
Start 14-day trial