Generative Engine Optimization Statistics: What the Data Actually Shows

A brief methodological note before the data: search behavior research is difficult to conduct at scale without bias, and studies that measure one engine at one point in time can become outdated quickly as products evolve. The studies cited here are the most rigorous available, but they should be read as directional findings, not as universal rules. Any vendor quoting you a guaranteed citation rate or a specific traffic improvement percentage is extrapolating beyond what the research supports.

This post cites the following primary sources only:

• SparkToro / Datos Zero-Click Search Study (2024) — large-scale clickstream analysis of US search behavior
• Semrush AI Overviews Study (December 2025) — analysis of AI Overview presence and click patterns across 10 million+ tracked search terms
• Pew Research Center behavioral study (July 2025) — tracking of approximately 68,879 real user searches including sessions with and without AI Overviews
• Seer Interactive client research (2025) — B2B client conversion data from AI-engine referral traffic, cited as directional rather than universal
• Aggarwal et al., "GEO: Generative Engine Optimization," KDD 2024 — peer-reviewed academic paper from Princeton University, Georgia Tech, and IIT Delhi

No invented statistics, no client case studies with fabricated numbers, and no "our data shows" claims appear anywhere in this post.

The zero-click statistic is frequently misrepresented as proof that all search traffic is collapsing. The more accurate reading: a majority of searches have always been satisfied at the SERP level for informational queries, and that proportion has grown as Google surfaces more direct answers. The AI Overview is one more direct-answer format in a long line of them, not a uniquely catastrophic change.

The intent segmentation matters significantly. Semrush's research confirms that informational queries skew heavily zero-click; transactional queries — "buy [product]", "book [service]", "[brand] pricing" — remain much more likely to generate a click. Zero-click rates are not uniform across intent types, and brands with strong commercial-intent content are less exposed to zero-click trends than publishers whose entire traffic model depends on informational page views.

What zero-click statistics mean for GEO strategy: if a large share of searches is being resolved at the AI answer level rather than by clicking through to source pages, the brand impression inside the generated answer becomes a primary marketing moment. The citation is the exposure, not the click.

The Pew study's methodology is worth understanding: it tracked real search sessions from a panel of consenting participants, not simulated searches or lab conditions. That makes it one of the most behaviorally realistic data sources available on AI Overview impact. The approximately 8% click-through rate when an AI Overview is present is not a search engine estimate — it is an observed behavior from real users.

The practical implication of both figures together: a meaningful and growing share of Google searches now involves an AI Overview, and when one is present, the probability that any individual organic listing generates a click drops substantially. For brands that depend on organic Google traffic, this is a structural shift that warrants a citation-presence strategy alongside a ranking strategy.

ChatGPT's user scale is substantial by OpenAI's public disclosures (hundreds of millions of weekly users), though we do not pin a specific figure here, as that number changes frequently and should be cited from the most current public statement rather than from a static research document. Perplexity's and Gemini's usage scale is similarly growing; directional evidence points to a significant and expanding user base for all three major AI search interfaces.

Important context for interpreting this finding:

• The 40% figure is the upper bound observed across tested techniques, not an average or guaranteed result. Individual techniques produced varying levels of improvement; the best-performing ones involved adding authoritative citations and statistical evidence to content.
• The study measured "source visibility" — how often a source was referenced or cited in generated answers — not traffic volume, conversion rate, or any commercial metric. Citation frequency is a meaningful GEO signal, but it is one step removed from business outcomes.
• The research was conducted on specific generative engines available at the time of the study. Engine architectures have evolved since publication; the directional conclusion (that structured content modifications improve citation rates) is broadly consistent with practitioner experience, even if the exact magnitude varies by engine and query type.
• This is the most rigorous peer-reviewed research specifically on GEO as a discipline and represents the strongest academic evidence that the optimization is real and measurable.

The KDD 2024 paper is available through ACM Digital Library. When citing it, attribute it to "Aggarwal et al., Princeton/Georgia Tech/IIT Delhi, KDD 2024" — not to any specific individual or organization as the sole author.

Why might AI-referred traffic convert at higher rates? The inference that makes logical sense: a user who receives a generated answer naming your brand as a solution to their problem has already completed significant pre-qualification before arriving on your site. They did not land on a page after clicking a generic keyword result — they arrived because an AI engine they trust named you specifically in response to their specific question. That is a much warmer introduction than a click from a search results page.

The appropriate caveats:

• Seer's data is from a specific set of B2B clients and may not generalize to B2C, e-commerce, or other categories where purchase decisions involve different dynamics.
• Attribution for AI-referred traffic is technically imperfect: many AI engine visits arrive as direct traffic or branded search rather than as a tracked referral from the AI engine's domain.
• Sample sizes for individual AI-engine referral pools are still relatively small for most businesses, making conversion rates volatile month to month.
• The trend direction (AI referrals converting meaningfully above organic baselines) is consistent with the mechanism, but use the specific percentages as context, not as a guarantee or as a benchmark against which to evaluate your own results.

For context on why different engines cite different domains, Seer Interactive's 2025 AI Overview study also found that each engine draws citations from a different mix of source domains — which is why per-engine optimization and measurement matters more than any single aggregate AI visibility score.

This finding is counterintuitive but important. The explanation that fits the data: AI Overviews tend to appear on queries that were already heavily zero-click because they had strong featured snippets or Knowledge Panel answers. On those queries, the AI Overview makes the SERP more useful to some users who then click for more detail, slightly increasing the click rate on those terms compared to their pre-AI-Overview baseline.

The broader picture this paints:

• Zero-click rate as an average is misleading: The headline 58.5% includes huge numbers of navigational, branded, and simple informational queries that were never going to generate clicks. Informational research queries and commercial-intent queries have very different zero-click profiles.
• AI Overviews add a new click type: Users can click directly on a source cited within an AI Overview. This click type — landing on a page cited in the Overview, not ranked in the traditional blue-link results — is not captured in pre-AI-Overview click attribution frameworks.
• The quality versus quantity trade-off is real: Fewer total clicks from AI search, but higher intent per visitor when AI search is the source, appears to be the pattern. The question is whether your business model captures value from fewer, higher-converting visits — most B2B and premium B2C models do.

Read the full statistical picture together. The headline zero-click number is alarming; the nuance on the same data set is more strategic: be in the AI answer, optimize for the citation, and treat the AI-referred visits you do receive as high-value opportunities worth disproportionate attention.

What the data does not yet support: guaranteed ROI estimates, universal citation improvement percentages, or specific traffic volume forecasts from GEO investment. The measurement infrastructure for AI search attribution is still maturing. Practitioners who make precise guarantees on GEO outcomes are extrapolating beyond the available evidence.

What the data does support, directionally:

• Structured content modifications improve citation rates (KDD 2024) — the optimization is real and measurable, even if the exact magnitude varies.
• AI Overviews are present on a meaningful and growing share of US searches (Semrush, Dec 2025) — the audience is not marginal.
• Users who see an AI Overview click traditional results at roughly half the rate of users who don't (Pew, July 2025) — citation presence in the AI layer has real visibility value even without a click.
• AI-referred visitors convert at substantially higher rates than organic (Seer Interactive) — the visitors who do click from AI answers are more qualified.
• Each engine cites different sources (Seer Interactive) — per-engine optimization and measurement is necessary, not optional.

To apply these insights to your specific brand, the starting point is a citation audit — your own prompt set run across each engine — which is what a free AI visibility audit produces. For the GEO practices that improve citation rates, see GEO best practices for 2026.