Does GEO Work? What Peer-Reviewed Research Actually Shows
By the My Marketing World Editorial Team
Every month, another acronym enters the marketing vocabulary. When “Generative Engine Optimization” started circulating in 2024, the healthy instinct was skepticism. Is GEO a measurable practice with evidence, or consulting firms repackaging existing advice?
The answer sits in the middle — but the evidence is real. The first academic study on GEO, authored by researchers from Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi and published at ACM SIGKDD 2024, tested nine optimization methods across 10,000 queries. Specific content tactics increased source visibility in AI-generated answers by up to 40%. The catch: the tactics that worked were content quality improvements, not technical tricks.
What the Evidence Actually Shows
The KDD 2024 paper introduced GEO-bench, spanning 25 domains. Here is the evidence in actionable terms:
Tactic
Visibility Impact
What It Means
Add direct quotations
+41%
Attributable quotes from credible third parties
Add statistics
+31%
Specific numbers, percentages, data points
Cite sources inline
+28%
References to primary sources
Improve fluency
+28%
Clearer prose that synthesizes cleanly
Keyword stuffing
-10%
Repetitive exact-match phrases hurt visibility
A notable finding: pages ranking around position 5 saw visibility gains over 115% in generative responses. Lower-ranked content may have more to gain than top-ranked pages. Effectiveness varied by domain — statistics performed strongly in government and business contexts, fluency improvements in debate-oriented queries. Our GEO optimization for B2B growth guide maps these tactics to verticals.
Why Trust Research Matters
Visibility in AI answers only matters if users trust what they see. A December 2024 survey by Northwestern University’s Retail Analytics Council — led by Dr. Martin Block among 7,861 adults — found 46% consumer awareness of generative AI, but trust barriers persist. Nearly 38% believe AI output needs human oversight; about 30% do not trust that AI has their best interests in mind.
Source citations, statistics, and attributable quotations are not just algorithmic signals — they are signals to skeptical readers. The Princeton/Georgia Tech study found that AI systems prefer the same credibility markers that cautious consumers do.
What Google Actually Says
Google’s 2024-2025 guidance explicitly calls GEO and AEO “still SEO” — crawlability, semantic HTML, structured data, and content quality remain the primary levers. Google dismisses several marketed GEO tactics: no need for llms.txt files, no need to “chunk” content, no benefit to inauthentic mentions. Google instead emphasizes “non-commodity content” — material providing unique insight. Both the peer-reviewed research and Google’s guidance point the same direction. The future of search and AI marketing trends offers perspective on this evolution.
Where GEO Does Not Work
The KDD 2024 study was conducted on GPT-3.5-turbo with a 200-query Perplexity.ai validation subset. Whether exact lifts hold across Claude, Gemini 2.5, and GPT-4o-search is unverified. More importantly, thin or inaccurate content does not become citation-worthy through quotation marks — the 40% figure applies to content that already had substantive value. As more publishers adopt these practices, the competitive baseline shifts. Our 2026 digital marketing overview places this shift in broader context.
A Practical Starting Framework
For teams ready to apply what the research shows:
- Audit pages ranked 3-10 — These have the most upside from GEO improvements.
- Add attributable evidence — Authority quotations, statistics with citations, inline references. Match the tactic to content type.
- Restructure for extractability — Clear headings, comparison tables, concise opening sentences that stand alone as AI quotes.
- Implement schema markup — Article, FAQPage, and HowTo schema help AI systems parse structure.
- Track citations across engines — Monitor ChatGPT, Perplexity, Gemini, and Google AI Overviews for 20-30 target queries monthly.
Frequently Asked Questions
Is GEO a replacement for SEO? No. The Princeton study found pages in traditional search benefited most. Google confirms generative AI relies on the same indexing and quality systems. GEO is an extension, not a substitute.
How long before GEO changes show results? For retrieval-based engines like Perplexity and ChatGPT Search, citation changes appear within 4-8 weeks. Influencing base LLM parametric memory takes months.
Can small sites compete with large publishers? The KDD 2024 study suggests they can. Pages ranked around position 5 gained over 115% visibility, while top-ranked pages showed modest gains. AI citation appears more democratic than traditional ranking.
Do I need special files or markup? No. Google says you do not need llms.txt, special AI-readable files, or chunking. Standard structured data and semantic HTML remain the foundation. AI marketing expert strategies offer additional perspectives.
What is the highest-impact single tactic? Direct quotations from credible sources produced the largest lift (+41%). Statistics performed better in government and business contexts; fluency mattered more for debate-style content. Match the tactic to your subject. Miklós Róth’s approaches to content authority align with these principles.
Research and Practical Sources
- Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K.R., & Deshpande, A. (2024). GEO: Generative Engine Optimization. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’24), Barcelona, Spain. DOI: 10.1145/3637528.3671900. arXiv:2311.09735.
- Block, D. M. (2024). Generative Artificial Intelligence and Consumer Trends in Awareness, Usage and Trust. Northwestern University, Retail Analytics Council. Survey by Prosper Technologies, December 2024 (n=7,861).
- Google Search Central. (2025). Optimizing your website for generative AI features on Google Search.
- Official GEO paper website with code and benchmark data
- Northwestern Medill Retail Analytics Council AI consumer trends