AI Overviews Optimization

Google says you do not need special technical optimizations to qualify for AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. or AI Mode. Evidence for this claim Google says there are no additional technical requirements or special optimizations required to appear in AI Overviews or AI Mode. Scope: Eligibility in Google Search AI features, not a promise of inclusion or performance. Confidence: high · Verified: 2026-07-14 Google: AI features and your website The familiar Search preview controls can, however, limit how content is used there. Evidence for this claim Google says nosnippet, data-nosnippet, max-snippet, and noindex controls also affect how content may appear in its AI search features. Scope: Google Search preview controls; more restrictive controls limit use and visibility. Confidence: high · Verified: 2026-07-14 Google: AI features and your website

TL;DR — AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. is the AI-written summary box Google shows at the top of search results. It’s part of Google Search, not a separate app — it pulls from Google’s normal indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed., so if you already do decent SEO, you’re most of the way there. Google literally says there’s nothing special to do to appear. The useful facts: Overviews mostly show for informational questions, ranking #1 doesn’t guarantee you get cited, and Google Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. can’t yet show you exactly how many clicks they cost you.

What AI Overviews are

When you Google a question these days, you often get a written answer at the top of the page, with a few links off to the side, before the normal blue links start. That’s an AI Overview — a summary Google’s Gemini AI writes on the spot by reading pages from Google’s search results and stitching the key facts together.

The single most important thing to understand: AI Overviews live inside Google Search. They’re not a separate product like ChatGPT or Perplexity. They use the same index, the same ranking, and the same quality checks as regular Google results. So there’s no separate “AI ranking” to game.

Do you need to do anything special?

Google’s honest answer is basically no. Its documentation says there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”. To be eligible to show up as one of the supporting links, a page just has to be indexed and eligible to appear in Google Search with a snippet. That’s it.

So “optimizing for AI Overviews” is mostly the SEO you already know: be crawlable, be indexable, write genuinely useful content, and earn a good reputation across the web. Google even says outright that from its perspective, optimizing for generative AI searchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. is “still SEO.”

A few things that are genuinely AI-Overview-specific

• They mostly show for questions, not shopping. Overviews appear far more for informational queries (“how does X work”) than for commercial or local ones.
• Ranking #1 doesn’t guarantee a citation. Google increasingly pulls its sources from related sub-searches, not just the top results for your exact query — so a page ranking below you (or not ranking at all) can get cited instead.
• They can hurt your clicks. When people get their answer in the box, fewer click through. Ahrefs measured about a 34.5% drop in click-through rate for the top result when an AI Overview shows — though there are signs that’s recovering.

Don’t fall for the “AI SEO hacks”

You’ll see people selling special tricks — llms.txt files, magic schema markup, chopping your content into tiny “chunks.” Google has publicly said none of these are needed for its AI features. Save your time for good content.

Want the deeper version — how Google sources citations, the real numbers on triggering and clicks, and exactly how to measure it in Search ConsoleGoogle's free tool for monitoring crawling, indexing, and search performance.? Switch to the Advanced tab.

Google’s documented baseline is ordinary Search eligibility, not an AI-only optimization checklist. Evidence for this claim Google says there are no additional technical requirements or special optimizations required to appear in AI Overviews or AI Mode. Scope: Eligibility in Google Search AI features, not a promise of inclusion or performance. Confidence: high · Verified: 2026-07-14 Google: AI features and your website Existing controls such as nosnippet, data-nosnippet, max-snippet, and noindex also apply to its AI features. Evidence for this claim Google says nosnippet, data-nosnippet, max-snippet, and noindex controls also affect how content may appear in its AI search features. Scope: Google Search preview controls; more restrictive controls limit use and visibility. Confidence: high · Verified: 2026-07-14 Google: AI features and your website

TL;DR — AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. is Google’s Gemini-generated answer box, drawing from the same indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. and core ranking/quality systems as classic Search — there’s no separate AI index or AI ranking factor, and eligibility is just “indexed + snippet-eligible.” What’s genuinely AI-Overview-specific: they skew hard to informational, longer-tail, low-commercial queries (~99% informational); they show outside position 1 for 8.64% of queries; sourcing has shifted from the original top 10 toward query-fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. sub-SERPs (top-10 citations fell from ~76% in mid-2025 to ~38% by early 2026); YouTube is cited disproportionately, independent of ranking; the 2025 CTR hit (~34.5%) is real but partially recovering through 2026, and being cited within the Overview roughly doubles CTR versus being an uncited source. Measurement is the big catch: GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. folds AI-feature impressions into the “Web” search type, and the new (June 2026) Generative AI performance report shows impressions only — no clicks, CTR, or queries. Ignore the llms.txtllms.txt is a proposed (not adopted) Markdown file at /llms.txt that gives AI systems a curated map of a site's most important pages. Proposed by Jeremy Howard in 2024, it's read mostly by coding agents like Claude Code — not search crawlers — and Google ignores it. / schema / chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. “hacks”; Google says on the record it doesn’t use them.

What AI Overviews actually are

An AI Overview is the summary Google’s Gemini models generate inside the regular Google SERP, drawing on Google’s normal Search index. Google describes their purpose as helping people “get to the gist of a complicated topic or question more quickly” and providing a jumping-off point to explore links.

The mechanism matters. Google says both AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. and AI Mode may use a “‘query fan-out’ technique — issuing multiple related searches across subtopics and data sources — to develop a response”. It retrieves relevant pages from the index (retrieval-augmented generationRAG is the retrieve-then-generate pattern behind AI search: the system retrieves relevant passages from an external index at query time, injects them into the model's context, and generates an answer grounded in those sources — without changing the model's weights., which Google calls groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it.), extracts facts, and writes them into prose with citations. Google describes the feature as rooted in core Search ranking and quality systems — I can’t stress that enough. For the mechanics of RAG and groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. as concepts, those are covered in the grounding and RAG deep dives; I won’t re-explain them here.

The consequence is the accuracy spine of this whole topic: because it runs on core Search, traditional SEO eligibility gates AI Overview eligibility. Google is blunt that “to be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet… There are no additional technical requirements.” In its May 2026 optimization guide Google went further: the SEO best practices still apply “because our generative AI features on Google Search are rooted in our core Search ranking and quality systems.” Search Engine Journal’s framing of that guide was that Google “directly contradicts tactics promoted by the growing AEO/GEO service industry”.

What triggers an AI Overview vs. a normal SERP

Overviews don’t show everywhere. Google says they’re “only shown when our systems determine that it is additive to classic Search, and as such, often don’t trigger.” The pattern in the data is strongly skewed, not random. From our 300K-keyword analysis (Louise Linehan, comparing 150K AIO SERPs against 150K non-AIO SERPs):

• ~99.2% of AIO-triggering keywords were informational in intent. Overviews had only about a 10% chance of showing for commercial and transactional queries.
• AIO keywords are longer-tail: ~4-word median phrase length vs. ~2 words for non-AIO SERPs.
• They’re lower-difficulty: median Keyword Difficulty ~12 vs. ~33, and content needed just ~13 referring domains to rank in an AIO SERP on average (vs. ~41 for non-AIO).
• They co-occur with other SERP featuresSERP features are any element on a search results page beyond the classic ten blue links — featured snippets, People Also Ask, knowledge panels, sitelinks, image and video packs, AI Overviews, and structured-data-driven rich results. Google documents that losing rich-result eligibility doesn't affect ranking; their SEO relevance is CTR (they redistribute clicks). Some are unlocked by markup; most are purely algorithmic.: 99.9% of AIO SERPs had at least one other feature (three on average), triggering 849% more Featured Snippets and 258% more Discussions than non-AIO queries.

My own 55.8M-Overview study across 590M searches lines up: AIOs showed for 9.46% of all keywords on desktop (16% in the US), skewed toward informational, longer, higher-volume queries, and away from branded, local, and short queries. And they mostly showed on non-monetized searches — about 71.67% of AIO searches had no CPC data at all. My honest read there was blunt: companies investing in content are the ones most likely to show for AIOs, and some may even benefit, but the vast majority, I think, will lose traffic.

AI Overviews aren’t always position 1

A common assumption is that the Overview always sits at the very top. It doesn’t. In my July 2025 study — pulling 10M SERPs with AI Overviews, 1M per country across 10 countries, run by our data scientist Xibeijia Guan — AI Overviews showed outside position 1 for 8.64% of queries, and the lowest we saw was position 6. There were country differences too (Japan and Spain had more Overviews outside position 1). It makes me wonder if they have some kind of scoring for how good the generated Overview is, and place it lower when confidence is lower.

How Google sources and grounds the answer — and why it’s shifting

This is the most important AI-Overview-specific trend, and most competing guides miss it because it only shows up if you re-run the data over time.

My original study found that about 76% of AI Overview citations pulled from the top 10 organic results. But when we re-ran it at 2x the sample — 863K keyword SERPs and 4M Overview URLs in March 2026 — the number had dropped to 37.9%. The rest split almost evenly between positions 11–100 (31.2%) and beyond the top 100 blocks (31.0%). Google’s own account is that fan-out is doing more of the work: Overviews are relying less on the direct search results and more on the sources showing up in fan-out query SERPs. Practical translation: ranking #1 for the exact query is no longer sufficient, and it’s becoming less sufficient over time as fan-out matures.

Two more sourcing facts worth internalizing:

• YouTube is cited disproportionately, independent of ranking. Among Overview- cited pages that didn’t rank in Google’s top 100 for the keyword, 18.2% were YouTube URLs — and YouTube is the single most-cited domain in AI Overviews, up 34% over six months. Video content gets pulled in on its own merits.
• UGC and reference domains dominate the top-cited list. In my 55.8M study, the top 50 domains accounted for 28.90% of all mentions, led by Reddit, Wikipedia, Quora, and YouTube. That’s the internet-talks-about-you signal showing up in the Overview.

What content formats and structures get pulled in

Google’s guidance here is deliberately minimal, and I’d trust it over the folklore. It recommends organizing content the way you would for humans: “People generally appreciate it when web pages are organized by paragraphs and sections, along with headings that provide a clear structure to navigate content.” There’s no ideal length and no chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. requirement (more on that in the myths section).

The one real content lever Google names is the commodity-vs-non-commodity distinction. It contrasts “commodity content (for example, something like ‘7 Tips for First-Time Homebuyers’)” — common-knowledge restatement anyone could write — against non-commodity content that offers a unique expert or first-hand take. This maps directly onto the information-gain idea covered elsewhere in this cluster: the way to stand out in a world where the AI can already summarize the obvious is to say something the obvious sources don’t. Bing frames its own version slightly more prescriptively — its AI Performance announcement says “clear headings, tables, and FAQ sections help surface key information and make content easier for AI systems to reference accurately” — but that’s Bing, not Google.

Click-through impact: real, contested, and moving

This is where the honesty has to be sharpest. The anchor number is our 34.5% CTR reduction study (Ryan Law and Xibeijia Guan, April 2025): across 300,000 keywords, the presence of an AI Overview correlated with a 34.5% lower average CTR for the top-ranking page versus similar informational keywords without one. Position-1 CTR for AIO keywords dropped from 0.073 to 0.026 comparing pre- and post-rollout.

But it’s not a settled, one-directional story:

• CTR has been partially recovering. Per an independent Seer Interactive study (53 brands, 5.47M queries, 2.43B impressions) reported by Search Engine Land, AI Overview CTR climbed to 2.4% in February 2026 from a December 2025 low of 1.3%. (These are relayed figures from a secondary source — treat as directional.)
• Being cited matters a lot. That same reporting found an AI Overview with a citation to the site averaged ~2.1% CTR versus ~0.9% without a citation — though both trail the ~3.3% CTR when no Overview showed at all.
• Google’s own claim runs the other way. Sundar Pichai has claimed that content and links within AI Overviews get higher click-through rates than outside them, and Google’s docs say clicks from Overview SERPs are “higher quality (meaning, users are more likely to spend more time on the site).” Treat that as a claim in tension with the aggregate independent data, not a settled fact.

There’s also a measurement wrinkle behind all of this — the great decoupling of clicks and impressions in GSC (impressions up, clicks down), which my colleague Ryan Law documented and which, as I’ve pointed out, sites with very few AI Overviews don’t share. That’s your bridge to the measurement problem.

How to actually track AI Overviews (and what GSC still can’t tell you)

Here’s the correction most people need. Google Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. does not have a “Search type = AI Overviews” filter. In the standard Performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data., AI- feature impressions are “reported on in the Performance report, within the ‘Web’ search type” — folded in with everything else, no separate value to filter on. (For the mechanics of that report generally, see the GSC Performance report deep dive.)

As of June 2026, Google ships a separate report: the Generative AI performance report. Google describes it as giving “dedicated views of your impressions within generative AI features on Search, such as AI Overviews and AI Mode.” Read the fine print carefully, because it’s easy to over-promise on it:

• It shows impressions only — how many times links to your site were shown in a generative AI feature. No clicks, no CTR, and no query dimension.
• Dimensions are Pages, Countries, Dates, and Devices — nothing else.
• Per the Help Center, it draws “data from the Web search type in the Performance report”, and it’s rolling out to a subset of properties for testing before wider release.

So the honest state of play: GSC still cannot isolate AI Overview clicks or CTR. Ryan Law flagged exactly this back in 2025 — despite Google’s optimistic click-quality claims, there was still no way to disambiguate AI Overview clicks and impressions from the rest of your Search ConsoleGoogle's free tool for monitoring crawling, indexing, and search performance. data — and the new report still doesn’t solve it, because it’s impressions-only.

The practical, two-layered workflow that follows:

1. Use the Generative AI performance report for visibility (impressions by page/country/device) where you have access.
2. Estimate click impact indirectly: filter Ahrefs’ Site Explorer Organic Keywords report by the AI Overview SERP feature crossed with Position 1–10 to find pages pulling double-duty, use Brand Radar-style citation tracking, and do before/after comparisons in the regular Performance report around when an Overview started showing for a query.

Worth noting Bing went further than Google here: its AI Performance report (public preview, Feb 2026) exposes actual grounding queries — the fan-out-style phrases that triggered a citation — which Google doesn’t surface anywhere.

What doesn’t move the needle

Google’s May 2026 guide reads like a direct rebuttal of the “AIO hacks” being sold:

• llms.txtllms.txt is a proposed (not adopted) Markdown file at /llms.txt that gives AI systems a curated map of a site's most important pages. Proposed by Jeremy Howard in 2024, it's read mostly by coding agents like Claude Code — not search crawlers — and Google ignores it. and special AI files. Google: “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search” — Search itself doesn’t use them. (The full null-result treatment lives in the llms.txt deep dive.)
• Chunking. Google: “There’s no requirement to break your content into tiny pieces for AI to better understand it.”
• Over-focusing on structured dataStructured data is a standardized way of labeling page content (using the schema.org vocabulary in JSON-LD, Microdata, or RDFa) so search engines can understand its meaning. It's not a direct ranking factor — its value is rich results and entity understanding.. Google: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” Keep schema for rich-result eligibility, but it’s not an AIO lever. (See the schema-markup-for-ai deep dive for the controlled null-result study.)

The unsourced third-party claims — “pages over 20,000 characters get ~10x more citations,” “40–61% of Overviews use lists” — I couldn’t trace to any primary study during this research. I’d treat those as SEO-blog folklore, not fact.

Where this fits in the bigger picture

Zoom out and this connects to the cluster’s headline finding: what correlates most with showing up in AI answers isn’t your on-page format, it’s what the rest of the web says about you. As covered in the AI search optimizationAI search optimization is the practice of making your brand and content visible, citable, and accurately represented across AI-powered search — Google AI Overviews, ChatGPT, Perplexity, Copilot. It's built on traditional SEO plus a heavier emphasis on off-site brand mentions and content AI systems can cite. hub, branded web mentions correlate with AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. far better (0.664) than Domain Rating (0.326). AI Overviews are the Google-specific instance of that broader reality.

If your question is the multi-platform version of this — how to show up across ChatGPT, Perplexity, and Copilot as well as Google — that’s the territory of generative engine optimization and answer engine optimization, the two siblings that own the cross-platform framing. This article is deliberately just about the box in Google’s own SERP. And the retrieved-vs-mentioned-vs-cited distinction is the right vocabulary for reasoning about measurement across all of them.

AI summary

A condensed take on the Advanced version:

• AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. = a feature of Google Search, not a separate platform. Gemini generates the box from Google’s normal indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. via query fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. + groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. (RAGRAG is the retrieve-then-generate pattern behind AI search: the system retrieves relevant passages from an external index at query time, injects them into the model's context, and generates an answer grounded in those sources — without changing the model's weights.). No separate AI index, no separate AI ranking factor.
• Eligibility = plain SEO. A page must be indexed and snippet-eligible; Google says there are no additional requirements and that optimizing for it is “still SEO.”
• Triggering is skewed: ~99% informational intent, longer-tail (~4-word median), lower KD (~12 vs ~33), mostly non-commercial (~72% no CPC data); ~10% chance on commercial/transactional queries.
• Not always position 1 — shown elsewhere for 8.64% of queries (as low as position 6).
• Sourcing is shifting away from the top 10. Top-10 citation share fell from ~76% (mid-2025) to ~38% (early 2026) as fan-out matured; ranking #1 no longer guarantees a citation. YouTube is cited disproportionately, independent of ranking.
• CTR impact is real but moving. ~34.5% CTR drop (2025); partial recovery through 2026; being cited within the Overview roughly doubles CTR vs. an uncited source. Google claims cited links click better — treat as contested.
• Measurement gap: GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. folds AI-feature impressions into the “Web” search type; the new (June 2026) Generative AI performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data. shows impressions only — no clicks, CTR, or queries. GSC still can’t isolate AI Overview clicks. Estimate clicks via Ahrefs SERP-feature filtering + before/after.
• Myths (Google-denied): llms.txtllms.txt is a proposed (not adopted) Markdown file at /llms.txt that gives AI systems a curated map of a site's most important pages. Proposed by Jeremy Howard in 2024, it's read mostly by coding agents like Claude Code — not search crawlers — and Google ignores it., special schema, and chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. don’t help.

Official documentation

Primary-source documentation from Google and Bing.

Google

• AI features and your website — how AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. and AI Mode work, eligibility, query fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics., content controls, and where the traffic shows up in Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. (the “Web” search type).
• Google’s Guide to Optimizing for Generative AI Features on Google Search (May 2026) — the mythbusting + best-practices guide: “still SEO,” RAGRAG is the retrieve-then-generate pattern behind AI search: the system retrieves relevant passages from an external index at query time, injects them into the model's context, and generates an answer grounded in those sources — without changing the model's weights./groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. definition, commodity vs. non-commodity content, and the llms.txtllms.txt is a proposed (not adopted) Markdown file at /llms.txt that gives AI systems a curated map of a site's most important pages. Proposed by Jeremy Howard in 2024, it's read mostly by coding agents like Claude Code — not search crawlers — and Google ignores it. / schema / chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. rebuttals.
• Introducing Search Generative AI performance reports in Search Console (June 3, 2026) — the announcement of the separate, dedicated impressions view for generative AI features.
• Generative AI performance report (Search) — Search Console Help — the dimensions (Pages/Countries/Dates/Devices), the impressions-only scope, and the “Web search type” data source.

Bing / Microsoft

• Introducing AI Performance in Bing Webmaster Tools (Public Preview) (Feb 10, 2026) — Bing’s comparable report: total citations, average cited pages, groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries (which Google doesn’t expose), and Bing’s content-format recommendations.

Quotes from the source

On-the-record statements from Google and Bing. Each link is a deep link that jumps to the quoted passage on the source page.

Google — it’s still SEO, no special optimizations

• “There are no additional requirements to appear in AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. or AI Mode, nor other special optimizations necessary.” — Google Search Central docs. Jump to quote
• “To be eligible to be shown as a supporting link in AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. or AI Mode, a page must be indexedStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. and eligible to be shown in Google Search with a snippet, fulfilling the Search technical requirements. There are no additional technical requirements.” Jump to quote
• “From Google Search’s perspective, optimizing for generative AI searchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. is optimizing for the search experience, and thus still SEO.” Jump to quote
• “In short, yes! The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems.” Jump to quote

Google — how they work and when they show

• “AI Overviews help people get to the gist of a complicated topic or question more quickly, and provide a jumping off point to explore links to learn more.” Jump to quote
• “Both AI Overviews and AI Mode may use a ‘query fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics.’ technique — issuing multiple related searches across subtopics and data sources — to develop a response.” Jump to quote
• “AI Overviews are only shown when our systems determine that it is additive to classic Search, and as such, often don’t trigger.” Jump to quote

Google — content, and the myths

• “People generally appreciate it when web pages are organized by paragraphs and sections, along with headings that provide a clear structure to navigate content.” Jump to quote
• “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.” Jump to quote
• “There’s no requirement to break your content into tiny pieces for AI to better understand it. Google systems are able to understand the nuance of multiple topics on a page and show the relevant piece to users.” Jump to quote
• “Structured dataStructured data is a standardized way of labeling page content (using the schema.org vocabulary in JSON-LD, Microdata, or RDFa) so search engines can understand its meaning. It's not a direct ranking factor — its value is rich results and entity understanding. isn’t required for generative AI searchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity., and there’s no special schema.orgSchema markup is code that uses the schema.org vocabulary to label what your content means so search engines can understand it and show rich results. It's most often written in JSON-LD, and it's not a direct ranking factor. markup you need to add.” Jump to quote

Google — measurement

• “they’re reported on in the Performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data., within the ‘Web’ search type.” — on where AI-feature traffic shows up in Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.. Jump to quote
• “We’ve seen that when people click from search results pages with AI Overviews, these clicks are higher quality (meaning, users are more likely to spend more time on the site).” Jump to quote
• “The generative AI performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data. includes data from the Web search type in the Performance report.” — Search ConsoleGoogle's free tool for monitoring crawling, indexing, and search performance. Help. Jump to quote

Bing / Microsoft

• “We are happy to introduce AI Performance in Bing Webmaster ToolsMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility., a new set of insights that shows how publisher content appears across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations.” Read the announcement
• “Clear headings, tables, and FAQ sections help surface key information and make content easier for AI systems to reference accurately.” Read the announcement

”Am I in AI Overviews, and what should I do?”

A quick walk-through for deciding where to spend effort. It assumes you’ve already got the basics (indexable, snippet-eligible) — because without those, none of the rest applies.

Mistakes that waste your time (or money)

Concrete things people do for “AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. optimization” that Google has publicly contradicted or that the data doesn’t support.

Adding an llms.txtllms.txt is a proposed (not adopted) Markdown file at /llms.txt that gives AI systems a curated map of a site's most important pages. Proposed by Jeremy Howard in 2024, it's read mostly by coding agents like Claude Code — not search crawlers — and Google ignores it. file to “help AI find you.” Why it’s wrong: Google says plainly it doesn’t use machine-readable files, AI text files, markup, or Markdown to appear in Search, including its generative AI features. Do instead: nothing here — spend the time on content and reputation. (Full null-result treatment in the llmsA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4)..txt deep dive.)

Bolting on special schema markupSchema markup is code that uses the schema.org vocabulary to label what your content means so search engines can understand it and show rich results. It's most often written in JSON-LD, and it's not a direct ranking factor. to force a citation. Why it’s wrong: Google says structured dataStructured data is a standardized way of labeling page content (using the schema.org vocabulary in JSON-LD, Microdata, or RDFa) so search engines can understand its meaning. It's not a direct ranking factor — its value is rich results and entity understanding. isn’t required for generative AI searchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. and there’s no special schema.orgSchema markup is code that uses the schema.org vocabulary to label what your content means so search engines can understand it and show rich results. It's most often written in JSON-LD, and it's not a direct ranking factor. markup for it; a controlled 1,885-page study foundA 302 (\"Found\") is a temporary redirect: it forwards users to a new URL while telling search engines the original URL should stay in the index. It's a weak canonicalization signal, not the zero-equity dead end of SEO folklore. no citation lift. Do instead: keep schema for genuine rich-result eligibility, but don’t treat it as an AIO lever. (See the schema-markup-for-ai deep dive.)

ChunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. content into tiny passages “so the AI can parse it.” Why it’s wrong: Google explicitly denies any chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. requirement and says its systems understand multiple topics on a page. Do instead: organize for humans — clear headings, paragraphs, sections — and cover the topic fully on one strong page.

Assuming ranking #1 guarantees you’ll be cited. Why it’s wrong: only ~38% of AI Overview citations now come from the top 10 (down from ~76% in mid-2025) as query fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. pulls from sub-query SERPs. Do instead: cover the surrounding subtopics the fan-out asks about, earn branded mentions, and don’t ignore video.

Trying to filter Search ConsoleGoogle's free tool for monitoring crawling, indexing, and search performance. by “Search type = AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index..” Why it’s wrong: there’s no such filter — AI-feature impressions live inside the “Web” search type, and the dedicated Generative AI performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data. is a separate report showing impressions only. Do instead: use the new report for impressions and estimate clicks with Ahrefs SERP-feature filtering and before/after Performance-report comparisons.

Repeating “pages over 20,000 characters get ~10x citations” / “40–61% use lists.” Why it’s wrong: I couldn’t trace either figure to a primary study; they read as unsourced SEO-blog folklore, and they don’t square with Google’s “no ideal length” guidance. Do instead: cite verifiable studies, and lead with non-commodity, first-hand content over length or list quotas.

Treating the 34.5% CTR drop as a fixed, permanent verdict. Why it’s wrong: it’s a 2025 aggregate; CTR has partially recovered through 2026, and being cited within the Overview roughly doubles CTR versus being an uncited source. Do instead: measure your own before/after, and aim to be cited, not just present in the sources.

AI Overviews readiness checklist

A pass to make sure you’re eligible and measuring the right things — not chasing hacks.

• Target pages are indexedStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. and snippet-eligible (no accidental noindex, nosnippet, or robots blocks) — this is the whole eligibility gate.
• Content is non-commodity: it says something first-hand or expert that the obvious sources don’t (information gainInformation gain is how much new information a page adds beyond what a searcher has already seen in prior results on the same topic — novelty relative to the existing corpus, not general content quality. The term comes from a granted Google patent that Google has never confirmed using in live ranking.), not just a restatement of common knowledge.
• Content is organized for humans — clear headings, paragraphs, sections. No chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. gymnastics.
• You’ve mapped the subtopics a fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. would likely ask, and your content (or your site) covers them — not just the head query.
• You’re building branded mentions across the web (the strongest correlate with AI citationAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking.), not just chasing links or Domain Rating.
• You’ve considered video/YouTube for topics where it’s cited disproportionately and independent of ranking.
• You are not relying on llms.txt, special AI schema, or content chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking. — Google says it doesn’t use them.
• You’ve checked whether your property has the Generative AI performance report and are reading impressions by Page/Country/Date/Device.
• You understand GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. can’t isolate AI OverviewAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. clicks/CTR, and you’re estimating click impact via Ahrefs SERP-feature filtering + before/after Performance-report comparisons.
• For Bing/Copilot visibility, you’ve looked at Bing Webmaster ToolsMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility.’ AI Performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data. (which also exposes groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries).

The mental models

1. It’s a feature, not a platform — so it’s still SEO. AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. run on Google’s normal indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. and core ranking/quality systems. There’s no separate AI index or AI ranking factor. If you catch yourself looking for an “AIO-only” lever, stop — the lever is eligibility (indexed + snippet-eligible) plus good SEO.

2. Eligibility gates everything. A page can’t be cited if it isn’t indexed and snippet-eligible. Before any “optimization,” confirm the page can actually appear in Search at all. Most “why am I not in AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index.” problems are ordinary technical-SEO problems.

3. Fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. breaks the “rank #1, get cited” assumption. Google increasingly sources from related sub-query SERPs, so citation share from the top 10 has fallen (~76% → ~38%). Think in topics and subtopics, not a single head keyword. Cover the neighborhood the fan-out will explore.

4. Presence ≠ citation ≠ click. Three different things: showing in the sources, being visibly cited, and getting the click. Being cited roughly doubles CTR versus being an uncited source. Aim for the citation, then the click — don’t stop at “we’re in the Overview.”

5. Impressions are measurable; clicks are inferred. GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. gives you AI-feature impressions (folded into “Web,” or broken out in the new report). It does not give you AI Overview clicks or CTR. Model click impact from before/after and third-party SERP-feature data — don’t expect GSC to hand it to you.

6. Off-site reputation beats on-page format. The cluster’s headline finding holds here: branded web mentions (0.664) correlate with AI citationAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. far better than Domain Rating (0.326). What the internet says about you matters more than tweaking your markup.

AI Overviews cheat sheet

Fact vs. myth

Triggering profile (where AIOs show)

Measurement fast facts

• Standard GSC Performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data.: AI-feature impressions folded into “Web”.
• New (June 2026) Generative AI performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data.: impressions only — Pages/Countries/Dates/Devices; no clicks, CTR, or queries; rolling out to a subset.
• GSC cannot isolate AI Overview clicks/CTR. Estimate via Ahrefs Site Explorer (AI Overview SERP featureSERP features are any element on a search results page beyond the classic ten blue links — featured snippets, People Also Ask, knowledge panels, sitelinks, image and video packs, AI Overviews, and structured-data-driven rich results. Google documents that losing rich-result eligibility doesn't affect ranking; their SEO relevance is CTR (they redistribute clicks). Some are unlocked by markup; most are purely algorithmic. + Position 1–10) and before/after comparisons.
• Bing goes further: its AI Performance report exposes groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries.

Numbers worth remembering

• Citations from top 10: ~76% (mid-2025) → ~38% (early 2026).
• Outside position 1: 8.64% of queries (as low as position 6).
• 2025 CTR impact: ~34.5% lower for the top result; partially recovering in 2026.
• Off-site correlation: brand mentions 0.664 vs. Domain Rating 0.326.

Citation presence across a fixed query set

Metric: Queries in a stable monitoring set where the site is cited in an AI Overview. What it tells you: Whether the site’s sources appear in observed answer panels, not whether every user saw the same panel. How to pull it: Capture the query, market, device, date, panel, and cited URLs under consistent conditions. Benchmark / realistic range: Establish a baseline for the same query set; panel availability and citations vary, so there is no universal target. Cadence: Sample on a consistent schedule and annotate major product or query-set changes.

Search performance for targeted pages

Metric: Impressions, clicks, click-through rate, and conversions for pages and queries included in the optimization program. What it tells you: Whether total Google search visibility and outcomes changed, including traffic that reporting does not separate as AI OverviewAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. traffic. How to pull it: Join Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. page and query data with first-party analytics and conversion events. Benchmark / realistic range: Compare with each page’s pre-change baseline and a suitable control where possible; do not assign all organic movement to AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index.. Cadence: Weekly for monitoring and monthly for decision-making.

Factual accuracy of observed mentions

Metric: Captured brand or product statements classified as accurate, incomplete, or incorrect. What it tells you: Whether visibility represents the entity and its claims faithfully. How to pull it: Review saved answer captures against current primary-source facts using a documented rubric. Benchmark / realistic range: The goal is accuracy for material facts; prioritize by business risk rather than inventing an industry percentage. Cadence: Monthly and after important product, policy, or brand changes.

Test yourself: AI Overviews Optimization

Five quick questions on how AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. work, source, and get measured. Pick an answer for each, then check.

Resources worth your time

My related writing

• 8.64% of AI Overviews Appear Outside Position #1 (And as Low as Position #6) — my study of 10M SERPs across 10 countries on where the Overview actually sits.
• Insights From 55.8M AI Overviews Across 590M Searches — my largest-scale AIOAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. study: prevalence, triggering conditions, and top-cited domains.
• 76% of AI Overview Citations Pull From the Top 10 — my original citation-vs-ranking study.
• Update: 38% of AI Overview Citations Pull From The Top 10 — the 2x-sample follow-up showing the top-10 share falling as fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. matured, plus the YouTube citation finding.

My speaking

• GEO? AEO? LLMO? What’s With All This AI SEO Stuff? (Speaker Deck, Ahrefs Evolve 2025) — my “optimize your site” vs. “optimize how the internet talks about you” framing that runs through this whole cluster.

Official

• Google — AI features and your website and Google’s Guide to Optimizing for Generative AI Features.
• Google — Generative AI performance reports announcement and the Search Console Help page.

From around the industry

• AI Overviews Reduce Clicks by 34.5% (Ryan Law & Xibeijia Guan, Ahrefs) — the anchor CTR-impact study, and the observation that GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. still can’t disambiguate AI Overview clicks.
• The Great Decoupling (Ryan Law, Ahrefs) — clicks-down/impressions-up in GSC, and why low-AIO sites don’t share the pattern.
• I Analyzed 300K Keywords: What I Learned About AI Overviews (Louise Linehan, Ahrefs) — the foundational triggering-conditions data (intent, length, difficulty, SERP-feature co-occurrence).
• Google’s New AI Search Guide Calls AEO And GEO “Still SEO” (Search Engine Journal) — coverage of the May 2026 guide and its mythbusting.
• Google AI Overviews CTR shows early signs of recovery: Study (Search Engine Land, reporting Seer Interactive) — the 2026 CTR-recovery data and cited-vs-uncited gap.
• AI Overviews optimization guide: How to rank in generated results (Search Engine Land) — a thorough competing guide on mechanics and ranking factors.
• Introducing AI Performance in Bing Webmaster Tools (Bing) — the Microsoft-side measurement tool that also exposes groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries.

Stats worth citing

• Triggering is ~99% informational. ~99.2% of AI-Overview-triggering keywords were informational; Overviews had ~10% odds on commercial/transactional queries. Source
• AIOAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. prevalence: ~9.46% of desktop keywords (16% in the US), from my 55.8M- Overview / 590M-search study. Source
• Not always position 1: 8.64% of queries show the Overview outside position 1 (as low as position 6), from 10M SERPs across 10 countries. Source
• Top-10 citation share fell ~76% → ~38% between mid-2025 and early 2026 as query fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics. matured; 18.2% of top-100-non-ranking cited pages were YouTube URLs. Source
• ~34.5% lower CTR for the top result when an AI Overview is present (2025, 300K keywords) — partially recovering through 2026. Source
• Brand mentions (0.664) beat Domain Rating (0.326) as a correlate of AI citation — the cluster’s headline off-site finding. Source

AI Overviews Optimization

AI Overviews optimization is the practice of getting your pages cited and represented in Google's AI Overviews — the Gemini-generated summary box in Google's own search results. Google says there are no special optimizations for it: the same index, ranking, and quality systems decide eligibility, so it's still SEO.

Related: Generative Engine Optimization (GEO), Answer Engine Optimization (AEO)

AI Overviews Optimization

AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. is the AI-generated summary box Google shows above or within its regular search results, produced by Gemini models built into Search. AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. optimization is the practice of getting your pages cited and shown as supporting links inside that box.

The key thing to understand: AI Overviews are a feature of Google Search, not a separate platform like ChatGPT or Perplexity. They draw from Google’s regular Search indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. and are gated by the same core ranking and quality systems. Google’s own guidance is explicit — there are no additional requirements or special optimizations to appear, and a page only needs to be indexed and eligible to be shown with a snippet. In Google’s framing, optimizing for its generative AI features is still SEO.

That said, AI Overviews behave in specific, measurable ways: they trigger mostly on informational, longer-tail, low-commercial-intent queries; they aren’t always in position 1 (they show elsewhere for about 8.64% of queries); and Google increasingly sources citations through “query fan-outQuery fan-out is the technique where an AI search system breaks a single user question into multiple related sub-queries, runs those searches concurrently, and synthesizes the retrieved results into one answer. Google confirms AI Overviews and AI Mode 'may use a query fan-out technique' issuing multiple related searches across subtopics.” sub-searches rather than the original query’s top 10 results — so a #1 ranking no longer guarantees a citation.

For the multi-platform version of this work across ChatGPT, Perplexity, and Copilot, see generative-engine-optimization and answer-engine-optimization. This term covers Google’s AI Overviews specifically.

Related: Generative Engine Optimization (GEO), Answer Engine Optimization (AEO)