Get a free AI visibility report
.png)
Cognizo pulled its own Google AI Overviews citation data across the AEO and AI visibility software category. The question: which domains actually get cited, and how much credit goes to the brand itself versus everyone else.
Most teams treat ranking well in Google and getting cited in Google AI Overviews as the same job. The domain-level data says otherwise. Cognizo pulled citation records for every AI Overviews answer generated across AEO and AI-visibility prompts over a three-month window. The pattern that emerged had less to do with any single brand's SEO. It had more to do with which kinds of pages Google's summarization layer trusts by default.
This is not a universal, cross-industry citation index. It is a category-specific look at what gets cited for AI visibility, AEO, or AI search tool questions. The data comes from Cognizo's own tracking, not a modeled estimate. The scope is narrower than a study spanning millions of unrelated queries. Every figure below reflects a real citation event.
Cognizo captures the actual rendered AI Overview a real searcher sees. That is UI-based tracking, not just an API sample of the underlying model. The distinction matters for AI Overviews citations specifically. An AI Overview can differ from a raw model call to the same system, so a UI capture reflects what buyers actually read on the results page.
For this data study, citations were filtered to the Google AI Overviews provider. The window ran June 1 through September 3, 2026, restricted to the United States. Any prompt built around Cognizo's own brand or a named competitor was excluded. What remained was a set of category-level questions. Think AEO platforms, AI visibility tools, AI rank trackers, and how to optimize for AI search, the kind of question a buyer asks before they know which vendor they want. The sample covers several thousand individual citation events. It runs across dozens of tracked prompts, on a recurring basis rather than a single snapshot.
Across every domain that earned a Google AI Overviews citation in this window, domains owned outright by a tracked software brand accounted for roughly 12% of total citation volume. The remaining 88% went to pages nobody in the category controls. That includes social platforms, review sites, press coverage, comparison pages, and forum threads.
This finding should reset how most teams plan an AI Overviews citations strategy. Optimizing a brand's own site is necessary, but it is not where most of the citation volume lives. Cognizo's citation share metric splits along exactly this line. Owned citations point to a brand's own domain. Earned citations point to a third party. In this category, earned dominates by a wide margin. Tracking brand mentions across the third-party pages that actually earn citations is the more productive move. Spend the next content or PR cycle there, rather than only auditing owned properties.
More than 23,000 distinct domains earned at least one Google AI Overviews citation in this dataset. Even the single most-cited domain, YouTube, captured only about 13% of total citation share. Reddit and LinkedIn came next, at roughly 4% and 3%. The top three combined still stay under a quarter of all AI Overviews citations tracked.
That fragmentation cuts against a common assumption in AEO planning. Most teams assume a handful of dominant sources control the answer. In practice, Google AI Overviews draws from a wide and shifting set of pages. Source mention rate is how often a given domain gets cited across a prompt set. It needs to be tracked across dozens of domains, not just the five or six a team assumes matter most.
Grouping every cited domain by page type shows where the volume concentrates. The categories below are mutually exclusive, so they sum to roughly 100%.
Corporate and vendor pages, company homepages, product pages, and blog content published by software companies, account for roughly 36% of citations. That includes a brand's own marketing pages, plus roundup or comparison content that other vendors publish about the category.
Software product and documentation pages add close to another quarter, at roughly 24%. Social and community platforms, primarily YouTube, Reddit, LinkedIn, and Quora, account for about 22% on their own. Corporate, software, and social pages combined make up close to four out of every five citations tracked.
Editorial and press coverage takes roughly 5%. Search engines and Google's own documentation pages take about 2%. Review platforms such as G2 sit under 1%. Institutional sources, directories, and niche categories like finance, legal, and health content each account for a fraction of a percent. This category rarely intersects with those topics.
This pattern is not unique to Cognizo's tracked category. Search Engine Land reviewed citation behavior across ChatGPT, Google AI Mode, Gemini, Perplexity, and AI Overviews. Reddit topped every one of those engines, with YouTube, LinkedIn, Wikipedia, and Forbes also placing in the top five. Review platforms such as G2 appeared frequently in recommendation-style queries (Search Engine Land).
The read here is not that a Reddit thread is inherently more authoritative than a vendor's own content. It is that AI Overviews treats community discussion and professional networks as a proxy for real-world usage and sentiment. That is the same signal behind measuring AI share of voice across a prompt set. A brand absent from those conversations misses a meaningful share of the citation pool. That holds no matter how strong its own site is.
One caveat belongs in any AI Overviews-specific data study. The citation logic that governs AI Overviews does not automatically apply to Google's other AI surfaces. Search Engine Land found Reddit accounting for 44% of all social media citations inside Google AI Overviews in one month. Inside Google Gemini, the same month, that share was just 5%, despite both surfaces being built by the same company (Search Engine Land). A citation strategy built around this data should not be assumed to transfer elsewhere. AI Mode, Gemini, and other non-Google engines need their own separate tracking.
Three practical takeaways follow from this data.
First, treat earned citations as the primary lever, not a supplement to on-site SEO. Owned domains account for roughly 12% of citation volume in this category. Most of the opportunity sits on properties a brand does not control. Think G2 reviews, Reddit threads, comparison articles, and YouTube content that mentions the brand accurately.
Second, track source mention rate at the domain level, not just brand-level visibility. Knowing a brand is mentioned in a third of tracked prompts is not enough. What matters more is knowing which specific domains are doing the citing. That is where PR, affiliate, and content placement effort should go.
Third, build this into a continuous workflow, not a one-time audit. The domain landscape shifts week to week. New comparison pages get published, and older ones lose relevance. A citation source worth targeting in June may not be worth targeting in September. Cognizo's Autopilot tier runs this loop on an ongoing basis. It identifies which third-party domains are gaining citation share in a category. It then prioritizes content optimization and outreach around them, so the shift shows up in reporting automatically. Customers on any paid tier can also pull this breakdown through Cognizo's API or MCP integration. That access starts at the Platform tier. Teams can feed it into their own reporting stack instead of exporting a dashboard by hand.
The clearest pattern is concentration at the type level, not the domain level. No single domain dominates. The top-cited domain in this dataset held about 13% share, and more than 23,000 domains earned at least one citation. But domain types cluster tightly. Corporate pages, software product pages, and social platforms together account for close to four out of five citations. A brand's own domain accounts for only a small fraction of that total. AI Overviews citations are earned across a wide, shifting set of third-party pages, not concentrated in a handful of authoritative sources.
Cognizo's tracked window ran June 1 to September 3, 2026, in the US, across AEO and AI-visibility category prompts. Within it, domains owned by a tracked brand accounted for about 12% of total citation volume. YouTube was the single most-cited domain, at roughly 13% share. Reddit followed at about 4%, and LinkedIn at about 3%. Corporate pages made up roughly 36% of citations, software and documentation pages about 24%, and social platforms about 22%. More than 23,000 distinct domains earned at least one citation across the window.
Cognizo captures the actual rendered Google AI Overview a real searcher sees. That is UI-based tracking, not just an API sample of the underlying model. Citations were filtered to the Google AI Overviews provider and restricted to the United States. They were limited to category-level prompts about AEO, AI visibility, and AI search tools, excluding any prompt built around Cognizo's own brand or a named competitor. Every domain-level count in this article reflects a real citation event captured during that window, not a modeled estimate.
Rarely, relative to everything else it cites. In this dataset, a brand's own domain, homepage, product pages, or blog, accounted for roughly 12% of total citation volume. That doesn't mean owned content is worthless. A well-structured product page is often what earns a citation on a comparison site in the first place. But treating on-site optimization as the whole strategy misses the 88% of citation volume that lands somewhere else.
More than most teams expect. Cognizo tracked over 23,000 distinct domains earning at least one AI Overviews citation within a single software category. That was over a three-month window. The top domain still captured only about 13% of total share, and the top three combined stayed under a quarter. That long tail means a domain-by-domain tracking approach covers more of the real opportunity. Focusing narrowly on the five or six sources a team assumes matter most misses the rest.
An owned citation links directly to a brand's own domain: its homepage, a product page, or a blog post. An earned citation links to a third party that discusses the brand without controlling it. Think a review site, a comparison article, a forum thread, or a press mention. In the dataset behind this study, earned citations outnumbered owned citations roughly seven to one. That is consistent with how AI Overviews sources answers more broadly. It pulls from a wide set of independent pages, not brand-controlled ones.
Not consistently. Search Engine Land's cross-engine research found Reddit, YouTube, and LinkedIn recurring across ChatGPT, AI Mode, Gemini, Perplexity, and AI Overviews. But the weighting differs sharply by engine. The same research found Reddit's share of social citations running far higher in AI Overviews than in Gemini. Both are Google products. A domain that earns AI Overviews citations is a reasonable candidate for other engines too. Still, verify it separately rather than assume.
Continuously, not on a fixed monthly or quarterly cycle. The domain landscape in this dataset shifted meaningfully within the three-month window alone. Some domains gained citation share while others that mattered earlier lost it. A snapshot taken once a quarter misses most of that movement. Cognizo's platform tracks this continuously. Shifts in source mention rate surface as they happen, not after a reporting cycle closes. That is what lets teams redirect content and PR effort toward a domain before a competitor claims the citation.
.svg)
.svg)
