Google AI Overviews: What They Are, How They Affect Rankings, and What to Do About It

In May 2024, Google renamed its Search Generative Experience (SGE) to AI Overviews and rolled it out to all US users — and eventually to over 100 countries. The feature places a generated AI summary at the very top of search results, above all organic listings, for a significant share of queries. For anyone who owns a website and depends on Google traffic, this is the most consequential change to the search results page since the introduction of featured snippets in 2014. The AI Overview occupies more vertical screen space than any previous SERP feature, and it answers the user's question directly — which means some users never scroll to the organic results below.

An AI Overview is a generated text summary that appears at the top of the Google search results page for certain queries. It is produced by Google's Gemini model, drawing on indexed web content. The summary includes inline citations — small numbered links that point to the source pages Google's model drew from. Those citation links are the new prime real estate in Google Search: being cited in an AI Overview gives you a clickable link above every organic listing on the page.

AI Overviews are not shown for every query. Google triggers them most consistently for informational queries with clear factual or explanatory answers — 'how does X work', 'what is the best X for Y', 'explain X' — and less consistently for navigational queries ('Facebook login'), commercial queries where a product comparison or transaction is the intent, or highly sensitive topics (medical, legal, financial) where Google has been cautious about AI-generated guidance. Queries with clear, stable factual answers are the most frequent triggers; queries with evolving or contested answers are less reliably covered.

The initial rollout of AI Overviews in May 2024 produced a wave of public criticism when users shared examples of the feature producing wrong or dangerous answers. A particularly viral example suggested putting glue on pizza to prevent cheese from sliding off; another recommended eating rocks as a source of minerals. Both were traced to Reddit posts or satirical sources that the model treated as credible. Google responded quickly, reducing the frequency of AI Overviews for queries where the model's outputs were less reliable, and strengthening the quality filters applied before an overview is shown.

By late 2024 and into 2026, the quality of AI Overviews improved substantially and their triggering became more selective. The feature is now far less likely to generate responses for queries where the model cannot identify high-quality, consensus-backed source material. The practical effect: AI Overviews appear most reliably for queries that already had strong featured snippet or knowledge panel coverage, because those are the queries where Google has the most confidence in the underlying source material.

When an AI Overview is shown, it includes numbered citations that link to the source pages the model drew from to construct the summary. These citations appear as small links within the overview text, and a panel of source cards appears to the right or below the generated text on desktop, and below it on mobile. The citations are not simply the top organic results — Google's model selects sources based on the specific claim being cited, not the overall quality of the page. A page that ranks #8 organically might be cited in the AI Overview if it contains the most precise and well-structured answer to a specific sub-question within the query.

Getting cited in an AI Overview is different from ranking in organic results, and requires a different optimisation approach. The factors that correlate with citation in AI Overviews include: directly answering the specific question in the first paragraph of the relevant section, using clear heading structure that labels the specific question being answered, having strong E-E-A-T signals (Experience, Expertise, Authoritativeness, Trustworthiness), and being on a domain with a history of being used as a trustworthy source for the topic.

The traffic impact of AI Overviews is real but uneven. For informational queries where the AI Overview comprehensively answers the question, click-through rates to organic results below the overview have fallen. Studies by Ahrefs, Semrush, and BrightEdge published in late 2024 found that pages ranking in position 1–3 for queries that triggered AI Overviews saw click-through rate reductions ranging from 15% to 40% compared to equivalent positions for queries without an AI Overview. The reduction was most pronounced for simple factual queries where the AI answer was complete, and least pronounced for queries where the user clearly needed to visit a specific site or service to complete their goal.

• Simple factual queries ('how many days in a leap year', 'what is the capital of Norway') — near-zero click value remained even before AI Overviews. Impact: minimal additional change.
• How-to and explanatory queries ('how to fix a leaky tap', 'how does compound interest work') — significant click reduction for informational-only visits. Users who need to act (not just learn) still click through.
• Comparison and review queries ('best project management software', 'seo tools comparison') — moderate impact. AI Overviews show options but rarely close the decision. Users still visit individual sites to evaluate.
• Commercial and transactional queries ('buy running shoes', 'book a hotel in Amsterdam') — low AI Overview trigger rate. Organic and paid results dominate. Minimal traffic impact.
• Local queries ('plumber near me', 'coffee shops in Edinburgh') — local pack dominates; AI Overviews rarely triggered. Minimal impact.

Google has confirmed that AI Overviews are sourced from the web index — the same pages that appear in organic results. Pages that are not indexed cannot appear in AI Overviews. Beyond indexing, the qualities that make a page a likely AI Overview citation source are consistent with the qualities that have always defined high-quality SEO content, but weighted differently.

• Direct, specific answers near the top of sections. The AI model extracts specific claims and their supporting context. Pages that bury the direct answer in the third paragraph of a section are less likely to be cited than pages that lead with the answer and support it afterward.
• Heading structure that mirrors natural questions. H2 and H3 headings phrased as questions ('How does X work?', 'What causes Y?') make it easier for the model to identify which section answers which query variant.
• Author expertise signals. Author bylines, author bios with credentials, and links to the author's other published work all contribute to E-E-A-T signals that make a page more likely to be used as a trusted source.
• Structured data. FAQ, HowTo, and Article structured data make content easier for the model to parse. Pages with well-formed JSON-LD schema are more likely to be selected as citation sources.
• Being the original source. Pages that are the first publisher of a specific fact, statistic, or analysis — not syndicated or duplicated content — are stronger citation candidates.

1. Audit which of your target queries trigger AI Overviews. Search for your top 20 keywords in an incognito window and note which show an AI Overview. These are your highest-risk and highest-opportunity queries.
2. Check who is being cited for those queries. Are your competitors cited? Is your site cited? If competitors are cited and you are not, compare their page structure, heading language, and opening paragraph directness.
3. Rewrite section openers to lead with the direct answer. If a section heading is 'Core Web Vitals' and the first sentence is 'In this section, we will explore what Core Web Vitals are and why they matter', rewrite it to 'Core Web Vitals are three specific performance metrics Google uses to evaluate page experience: LCP (loading), INP (interactivity), and CLS (visual stability).'
4. Add FAQ schema to pages covering common question-style queries. Use the Schema Tester at /tools/schema-markup-tester to validate your JSON-LD.
5. Preview your title and meta description in the SERP preview tool. AI Overviews show source titles in the citation cards. A clear, keyword-accurate title helps the AI identify your page as a relevant source. Use the Meta Preview at /tools/google-serp-preview to verify your title and description display correctly.
6. Build author entity signals. Add an author bio with credentials to your editorial pages. Link the author to their LinkedIn profile and any published work on other sites.
7. Monitor your GSC AI Overview report. Since 2024, Google Search Console shows impression and click data specifically for queries where AI Overviews appeared. Check this weekly for queries where you are getting impressions in AI Overviews but not citations (indicating your content is considered relevant but not direct enough).

Google has consistently expanded the scope of AI Overviews since launch. The feature has grown from US-only to over 100 countries. The query types that trigger it have expanded from purely informational to include some commercial and how-to queries. Google's 'AI Mode' — a dedicated AI-first search interface available to Google One subscribers and as an experimental feature — represents a more aggressive version of AI Overviews where the entire results page is structured around an AI conversation rather than a traditional blue-link listing.

The trajectory is clear: AI-generated content will occupy more of the search results page, not less. The most durable response is not to try to outmanoeuvre AI Overviews but to become a source they cite — which requires the same qualities that have always defined genuinely useful, authoritative content: accuracy, depth, clear structure, and demonstrable expertise. Sites that optimise purely for keywords and volume without substance will find AI Overviews accelerate their decline. Sites that are genuinely the best source on a topic will find that AI Overviews amplify their authority.