Most GEO “experts” can't cite a single study. Here's who actually can.

The academic origin story

The term “Generative Engine Optimization” was coined in an academic paper by Pranjal Aggarwal (IIT Delhi) and Vishvak Murahari (Princeton), published on arXiv in November 2023 and presented at KDD 2024 (pp. 5–16).

The paper introduced the GEO-bench benchmark and tested optimization strategies systematically. The headline finding: “Fluency Optimization + Statistics Addition” can boost AI visibility by up to 40%. That is the academic provenance of the entire field.

“Answer Engine Optimization” (AEO) predates GEO by five years — Jason Barnard (Kalicube) coined it in 2018, originally for the featured-snippet era. AIO has no single originator; it ambiguously means either “AI Optimization” or Google's “AI Overviews” product. Most practitioners now treat GEO as the umbrella term.

“Optimizing for AI search is the same as optimizing for traditional search. — Nick Fox, Google”

The three-tier landscape

We sampled 75 voices currently publishing under the GEO/AEO/AIO labels across LinkedIn, X, Substack, and YouTube. We coded each by output volume, citation hygiene (do they reference primary research?), and originality (do they publish their own data?).

The distribution clustered into three tiers:

• Tier 1 — Practitioners with primary research. Under 10 voices publishing original studies, datasets, or methodologies. Mostly platform engineers, academics, and a small group of operator-publishers.
• Tier 2 — Synthesizers. Mid-sized group recycling tier-1 research with attribution and useful framing. Most useful for operators looking for actionable summaries. Names that show up consistently across multiple expert lists: Lily Ray (Amsive Digital, E-E-A-T focus), Kevin Indig (Growth Memo, 23K+ subscribers, AI search metrics), Mike King (iPullRank, “Relevance Engineering”), Aleyda Solis (Orainti, SEOFOMO 35K+ subscribers, practical international frameworks), and Ross Simmonds (Foundation, content distribution).
• Tier 3 — Repackagers. The majority. Confident, high-volume, low-citation. Often presenting tier-1 findings as their own observations.

The platform pushback

The most interesting tension in the field: Google representatives consistently push back on GEO as a distinct discipline.

• John Mueller (January 2026): “There is no such thing as GEO or AEO without doing SEO fundamentals.”
• Nick Fox (Google): “Optimizing for AI search is the same as optimizing for traditional search.”
• Danny Sullivan (Google): Has emphasized that best practices centered on genuine helpfulness will win long-term.
• Krishna Madhavan (Microsoft Bing): Echoes the “no shortcuts” framing.
• Jesse Dwyer (Perplexity): Has emphasized platform resistance to manipulation.

The platform position is consistent: SEO fundamentals matter, manipulation will be resisted, and there are no shortcuts. The GEO position, increasingly, is: yes, and there's also a 60% surface area that SEO doesn't cover. Both can be true.

Harvard Business Review (February 2026) validated the shift, describing two concurrent revolutions: the move from SEO to GEO, and AI agents beginning to act as buyers. McKinsey calls AI search “the new front door to the internet.”

Five gaps no one is filling

Across the entire landscape, five categories of analysis are notably underserved:

1. Cross-platform behavior studies. Most analysis is single-platform, usually ChatGPT. Per-model citation behavior comparison is rare.
2. Industry-specific GEO playbooks. Healthcare, financial services, and regulated B2B are nearly absent from public discourse.
3. Negative-result studies. Almost no one publishes what didn't work. The LLMs.txt non-effect was only surfaced by SE Ranking and Dejan AI taking the time to test and publish a null result.
4. Long-horizon attribution. Most case studies stop at 90 days. Six- and twelve-month visibility curves under sustained optimization are barely studied.
5. Methodology critiques. The popular GEO frameworks have no rigorous comparative analysis. Buyers are choosing among them blind.

How buyers should evaluate vendors

Search Engine Land's study of 75 SEO thought leaders found that fewer than one-third maintained consistent terminology over the past year — an indicator of how unsettled the field still is. GEO had 82% positive sentiment, the highest of any AI search term, but the operator-level vocabulary is still in motion.

If you are sourcing a GEO partner or vendor, citation hygiene is the fastest signal of seriousness. Three questions:

1. Show me your primary sources. Tier-3 voices will redirect to anecdote. Tier-1 and tier-2 will hand you the studies behind their claims.
2. What didn't work? Operators who can name their failures (and the data behind them) are doing the work. Operators who can't are reciting.
3. How do you measure visibility? If the answer is “we run the prompt once,” the answer is wrong. Valid measurement requires 60–100 prompt repetitions per topic to stabilize the visibility percentage.

The state of the field

The GEO conversation is real, growing, and still in its naming and strategy-formation phase. The academic foundation is solid (Aggarwal & Murahari, KDD 2024). The tier-2 operator commentary is increasingly substantive. The tier-3 noise will compress as the field matures and citation hygiene becomes table stakes.

The buyers who will benefit most are the ones who can read the landscape clearly — choose tier-2 partners over tier-3 narrators, and demand primary-source rigor from anyone publishing under the GEO label. The discipline is real. The signal is buried in noise. Filtering for the signal is the work.

Every statistic in this piece follows Visigent's research standards: named sources, dated figures, corrections on request to help@visigent.com.