Google AI Mode & AI Overviews Cite Different URLs, According to Ahrefs Report

According to Ahrefs report, Google AI Mode and AI Overviews frequently cite different URLs, signaling shifts in how Google surfaces sources.

New analysis from Ahrefs suggests that Google’s AI Mode and AI Overviews usually give semantically similar answers while frequently pulling in different supporting URLs. For SEOs and brands, that means visibility in one AI surface does not guarantee citations or mentions in the other, even for the same query.​

How Ahrefs Compared AI Mode And AI Overviews

In the report, published on the Ahrefs blog, the Ahrefs team used September 2025 U.S. data from its Brand Radar tool to compare responses from AI Mode and AI Overviews to the same queries.

They examined 730,000 query pairs for content similarity, and a subset of 540,000 pairs for citation and URL overlap, giving a large sample across a wide range of intents and topics.​

Low URL Overlap, High Semantic Similarity

Ahrefs found that AI Mode and AI Overviews cited exactly the same URLs just 13% of the time; when restricting the comparison to the top three citations in each response, overlap rose only slightly to 16%.

Yet the underlying meaning was closely aligned: the average semantic similarity score between responses was 0.86, and 89% of pairs scored above 0.8 on a 0–1 scale, where 1.0 represents identical meaning.​

Linguistically, the outputs differed as well. The two systems shared just 16% of their unique words on average, and the very first sentence matched exactly in only 2.5% of cases.

As Ahrefs’ Despina Gavoyannis puts it:

“Put simply: 9 out of 10 times, AI Mode and AI Overview agreed on what to say. They just said it differently and cited different sources.”

Different Source And Format Preferences

The report highlights distinct source patterns for each experience. AI Mode leaned more heavily on encyclopedic and Q&A-style sites, citing Wikipedia in 28.9% of responses versus 18.1% for AI Overviews, and referencing Quora roughly 3.5 times as often.

It also pulled from health sites at about double the rate of AI Overviews, and its outputs were, on average, four times longer and richer in entities.​

AI Overviews, meanwhile, showed a stronger bias toward video and core site pages. YouTube was the most frequently cited domain in AI Overviews, and this format cited videos and core pages (like homepages) nearly twice as often as AI Mode, though both still preferred article-style pages overall.

Reddit appeared at similar rates in both, suggesting some consistency in how community content is used as a supporting source.​

Entity Mentions And Citation Gaps

Because AI Mode responses are longer, they tended to include more entities and brand mentions. On average, AI Mode referenced 3.3 entities per answer, compared with 1.3 for AI Overviews, and in about 61% of cases it contained all the entities mentioned in the AI Overview plus additional ones.

Still, many responses from both systems omitted named entities entirely, particularly for generic informational queries where specific brands or people are not central to the answer.​

On the citation side, AI Mode was more likely to include sources at all. Only 3% of AI Mode responses lacked citations, versus 11% of AI Overviews, with gaps often tied to cases like simple calculations, sensitive topics, or unsupported languages.

That difference further complicates how “visibility” is measured, since some AI Overview answers may not credit any external URL even when content clearly informs the response.​

How Google’s Own Docs Explain The Differences

Google’s documentation explains that both AI Overviews and AI Mode can use a “query fan-out” technique, where the system issues multiple related searches across subtopics and data sources while generating a response.

AI Mode’s help page notes that it divides a user’s question into subtopics, runs those in parallel against multiple sources, then aggregates the results into a single answer with supporting links.​

Google also states that AI Mode and AI Overviews may rely on different models and ranking techniques, which naturally leads to variation in phrasing and link selection even when the “core” answer is similar.

The Ahrefs data supports this: semantic intent is usually aligned, but the underlying link graph and content snippets differ often enough to matter for brands and publishers.​

Strategic Implications For SEOs And Brands

For measurement and strategy, the main implication is that AI search visibility is now multi-surface by default. A citation or mention in AI Overviews does not guarantee a corresponding placement in AI Mode for the same query and AI Mode’s longer answers create more room for additional brands, entities and competitors.

In related research, Ahrefs’ separate work on AI Overview volatility, which found that 45.5% of citations can change when AI Overviews update, adds another layer of instability to track.​

Practically, this suggests:

• Tracking AI Overviews and AI Mode visibility separately, rather than treating them as a single channel.
• Focusing on semantic authority being the best answer for a topic cluster rather than optimizing for specific wording that one system might use.​
• Paying attention to formats and platforms each surface favours (for example, video and core pages for AI Overviews, deeper articles and Q&A sources for AI Mode) when planning content and distribution.​

Bottom Line

Even if Google continues to refine how these systems converge, the current low citation overlap indicates that AI Mode and AI Overviews are distinct experiences and brands that care about AI search presence should treat them that way.