Definition: What are AI Overviews?

AI Overviews (launched in 2023/24 under the beta name Search Generative Experience, SGE) are a SERP component that uses generative AI to produce a direct answer to a user's question. They appear above the classical blue links and typically contain 3-8 sentences of synthesized text, flanked by several linked sources (inline or as link cards).

The feature launched in May 2023 as the SGE beta inside Google Search Labs, rolled out productively in the United States from May 2024 and expanded to European markets including the DACH region in early 2025. Since mid-2025, AI Overviews are the default state on informational query clusters in most markets - not the exception.

Core mechanic

AIO is not a ranking factor - it is a structural change to the SERP

The classical algorithm continues to rank organic results. In parallel, a separate system generates the AI answer. Both run side by side. Position 1 is still technically position 1 - it just no longer gets the click.

How AI Overviews work (simplified pipeline)

  1. Query classification: Google decides whether a query is AIO-eligible (informational, definitional, comparative).
  2. Retrieval: Relevant documents are pulled from the index - a hybrid of BM25 and semantic vector search.
  3. Re-ranking: Candidates are re-weighted by authority, freshness, entity match and passage quality.
  4. Generation: A Gemini-based model synthesizes a coherent answer from the top sources - with inline citation references.
  5. Display: The answer is rendered above the SERP, optionally with an expandable section.

In its guide to generative AI features, Google officially confirms two building blocks: retrieval-augmented generation over the search index and core ranking systems, and query fan-out, i.e. concurrent sub-queries issued by the model. The remaining steps are a simplified model. What matters operationally: content can drop out at every step of the pipeline. Whoever optimizes only at the retrieval layer (schema, sitemap) but fails to supply citable passages gets fetched but not cited.

The traffic effect - documented via AAR

The traffic effect of AI Overviews is measured through the AI Overview Absorption Rate (AAR). In an in-house audit cohort of 500+ enterprise domains (October 2024 - April 2026, see benchmarks), the picture is the following:

28-41%
AAR on informational queries
< 10%
AAR on brand / navigational queries
5-18%
AAR on transactional queries

Crucially, click loss occurs at stable ranking positions. Classical SEO dashboards often misread the decline as an optimization deficit - in reality it is structural. Without AAR measurement, the wrong root cause gets worked on.

Since June 2026, Google Search Console has shown impressions from AI Overviews, AI Mode and generative features in Discover in its generative AI performance report, broken down by page, country, device and date. Google launched the report on June 3, 2026 for a subset of sites and has made it available to all sites worldwide since August 31, 2026. Clicks, CTR and queries are not included yet, so AAR measurement remains necessary for the click level and root-cause analysis.

How to get cited in AI Overviews

Passage citability per QUEST

Every core passage is tested against the five criteria of the QUEST heuristic: Quotable, Unambiguous, Entity-rich, Standalone, Timestamped. Passages that satisfy 4/5 are citable. Below 3 criteria they are unusable for AIO.

Entity clarity

The brand must be anchored as a machine-readable entity: Schema.org Person/Organization markup, consistent sameAs references, ideally a Wikidata entry. Without an entity anchor, content fails systematically at the re-ranking stage.

Structured data and passage structure

AI Overviews favor content that is already structured at the passage level: short, self-contained Q&A and definition passages in the visible content. FAQPage and HowTo markup is only a semantic declaration: Google retired HowTo rich results in 2023 and FAQ rich results on May 7, 2026 (Google changelog), and according to Google no special markup is needed for generative features. Article schema with datePublished and author makes freshness and authority signals unambiguous.

Author and publisher signals

A clear author byline with Person schema, transparent publication metadata and verifiable About pages increase the chance of citation in our audits. This is the directly effective part of E-E-A-T in practice.

Typical mistakes around AI Overviews

Related terms

AI Overviews are tightly coupled with AAR, CTR drift, Passage Ranking, Zero-Click Search, Citation Rate and SGE. The overarching discipline is GEO.

FAQ on AI Overviews

What are AI Overviews?

AI Overviews are Google's generative answer section above the classical blue links. Google uses retrieval-augmented generation to synthesize a direct answer from multiple sources in its search index and links the sources as evidence.

On which queries do AI Overviews appear?

Primarily on informational queries (definitions, how-to, comparisons). On navigational brand queries and pure transactional intents they appear significantly less often.

How large is the traffic loss caused by AI Overviews?

In our audit cohort, the AAR (AI Overview Absorption Rate) on informational queries sits between 0.28 and 0.41, meaning 28 to 41 percent click loss at stable rankings. On navigational and transactional queries it is substantially lower.

How do you get cited in AI Overviews?

Through passage citability (QUEST heuristic), entity clarity, authoritative signals (E-E-A-T), unambiguous publication dates. Google requires no special markup for generative features; pure keyword optimization is ineffective.

Should you try to block AI Overviews?

According to Google, display of individual content can be controlled with nosnippet, data-nosnippet, max-snippet or noindex; Google-Extended has no effect here. In practice, blocking is rarely advisable. The better strategy is active citation optimization - not blocking.

Sources