A new critical survey titled "Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026)" has been published and surfaced through alphaXiv, the discussion platform built around arXiv preprints. The paper consolidates research on generative engine optimization — the practice of improving how content is cited and surfaced inside AI-driven answer systems rather than traditional ranked link lists.

The survey covers the 2023-2026 window, which is precisely the period when generative answer products moved from experiments to mainstream search surfaces. Google rolled out AI Overviews to all US users in May 2024 and expanded them to over 100 countries by October that year. OpenAI, Perplexity and Microsoft's Copilot built comparable answer engines that synthesize sources instead of presenting ten blue links. The survey's framing confirms what practitioners have suspected since early studies on the topic: visibility work now has to target two distinct systems — the classic ranking stack and the retrieval-plus-generation pipeline that decides which sources an AI answer cites.

What the survey addresses

The paper's title signals two things worth flagging for search professionals.

First, it treats generative engine optimization as a research field with a defined history. The term GEO entered the academic literature in late 2023, when a widely cited preprint from Princeton, Georgia Tech, Allen Institute for AI and IIT Delhi reported that adding statistics, quotations and citations to page content measurably increased the likelihood of being referenced by generative engines — with visibility gains reported in the 30-40% range in that original study. Since then, a steady stream of papers has tested which content attributes drive inclusion in AI answers.

Second, the word "critical" in the title indicates the authors evaluate this literature rather than merely catalog it. That distinction matters for an industry that has absorbed a large volume of GEO advice with thin evidentiary support. Tool vendors and consultancies have published plenty of correlation-style claims about AI citation patterns over the past two years; academic surveys like this one apply stricter methodological scrutiny to what is actually demonstrated versus what is speculation.

Confirmed versus speculative — where the lines sit

Search engines themselves have confirmed relatively little about how AI answer systems select sources. Google has said AI Overviews draw on its core ranking systems and that sites ranking well organically are generally the same sites cited in Overviews. Bing has described grounding answers in its index. Neither engine has published the citation-selection mechanics in detail.

Everything beyond those statements — including claims about optimal content structure, entity coverage or formatting for AI citations — falls into the tool-signal and research-experiment category. This survey sits in the latter camp: it summarizes peer-reviewed and preprint findings, not official search-engine documentation. Readers should weigh it accordingly.

Why the timing matters for publishers

The survey arrives as publishers and SEO teams are actively measuring AI-driven referral traffic, which behaves differently from organic click-through. Studies from traffic-analytics firms over the past year have reported AI-search referrals converting at higher rates in some verticals while arriving in far smaller volumes than classic organic search. Editorial, e-commerce and B2B content teams have all reported diverging patterns in how often generative engines cite them, and a structured synthesis of the 2023-2026 research gives those teams a common baseline.

For site operators, the practical value of a survey like this is consolidation. Instead of tracking scattered preprints, practitioners get a single map of what has been tested, what held up and where the literature found methodological weaknesses.

What to watch next

The full paper's reception on alphaXiv — where researchers annotate and debate arXiv papers — will indicate how the search and IR community judges its conclusions. Practitioners should monitor whether the survey identifies reproducible GEO factors that survive scrutiny, and whether future work ties those factors to measurable citation shifts in live AI Overviews, ChatGPT Search and Perplexity results across verticals.