Google AI Mode vs AI Overviews: What Differs

What is the difference between Google AI Mode and AI Overviews?

The difference is where each one meets the reader. Google AI Overviews are AI-generated summaries that sit above the classic blue links on a traditional search, arriving whether or not anyone asked for them. Google AI Mode is a conversational surface you enter deliberately, built for follow-up questions and comparisons. Neither is a separate product from Google Search.

Google describes the split by the job each does. AI Overviews exist to get someone to the gist of a complicated question quickly, and Google says they appear only where its systems judge them additive to classic Search, so they often do not trigger. AI Mode is aimed at questions needing exploration, reasoning or comparison that would previously have taken several searches.

Both surfaces sit on the same foundation, which is the part most comparisons skip. Google states that its generative AI features on Search are rooted in its core Search ranking and quality systems, and that they highlight content from the Search index using retrieval-augmented generation, the technique also called grounding.

The one difference Google documents in the retrieval itself is deliberately vague, and it is the licence for everything below. Google states that AI Mode and AI Overviews may use different models and techniques, so the set of responses and links they show will vary. That is a much narrower claim than the field usually makes on Google's behalf.

Do Google AI Mode and AI Overviews cite the same sources?

Mostly they do not. Ahrefs compared both surfaces on the same queries (opens in a new tab) and found they cited the same URL only 13.7 per cent of the time. Looking at just the top three citations from each lifted it to 16.3 per cent, which is still low enough that the two rarely agree on a source.

The study is worth stating properly, because most pages quoting it do not. Ahrefs analysed 540,000 query pairs for the citation and URL analysis and 730,000 for content similarity, drawn from its Brand Radar dataset. The data is September 2025 US data, and the study was published on 15 December 2025. Publication date and collection date are three months apart, and the collection date is the one that matters.

What makes the finding interesting is that the disagreement is about sources rather than about substance. The same study measured 86 per cent average semantic similarity between the two answers, with 89.7 per cent of response pairs scoring above 0.8 on cosine similarity. Ahrefs summarised it as the two surfaces agreeing on what to say nine times out of ten while citing different places for it.

Word-level overlap was low too, at a Jaccard similarity of 0.16. The two produced an identical opening sentence 2.51 per cent of the time and an identical response 0.51 per cent of the time. So AI Mode is not an expanded AI Overview, and Ahrefs went in expecting that it might be.

How different is 13.7 per cent overlap, really?

A 13.7 per cent overlap between Google AI Mode and AI Overviews is a large difference, and the way that number is usually reported makes it impossible to judge. The common rendering is that 87 per cent of the time the two surfaces use completely different sources, measured against an implied 100 per cent. That baseline is wrong, because an AI Overview does not agree with itself between runs either.

Ahrefs measured that separately, in a study nobody pairs with the overlap one (opens in a new tab). Analysing more than 43,000 keywords that each had at least 16 recorded AI Overviews over a month, it found that consecutive AI Overviews for the same query share only 54.5 per cent of their cited URLs. Nearly half the sources change from one observation to the next, and the content changes roughly every 2.15 days.

Setting those side by side gives the figure a scale. Two different Google surfaces agree on 13.7 per cent of cited URLs. One Google surface agrees with its own previous answer on 54.5 per cent. The gap between the surfaces is roughly four times the gap a single surface opens with itself, which is what makes the case that they are genuinely separate retrieval systems rather than one system observed twice.

Treat that as a reference point rather than a controlled comparison, because the two studies were not built to be compared. They use different samples, different windows and different measurements: cross-surface overlap on single generations against same-surface overlap on consecutive generations. Ahrefs makes the underlying point itself, noting that citation pools may overlap more than a single-snapshot comparison suggests.

The practical consequence is about how you read your own data. A brand that appears in an AI Overview one week and not the next has probably observed normal churn rather than a ranking change, and the same caution applies with more force when comparing across the two surfaces. One check of one query is close to worthless as evidence on either.

Which sources does each Google AI surface prefer?

The two surfaces lean towards different kinds of site, and the directions are consistent rather than dramatic. In Ahrefs' data, AI Overviews showed a stronger preference for video and for community platforms like Reddit, with YouTube holding the top position in its citations, cited more often than encyclopedic sources.

AI Mode leaned the other way, towards reference and detail. Wikipedia appeared in 10 per cent more AI Mode citations than AI Overview citations, Quora appeared 3.5 times more often, and health sites and Facebook were each cited about twice as often. Reddit was the exception, cited a similar amount by both. AI Overviews cited videos and core pages, meaning homepages and main category pages, close to twice as often as AI Mode.

Our own AI Overview panel points the same way on video, on a much smaller sample. Across the 100 most recent AI Overview citations RankX AI recorded for five tracked keywords in the 30 days to 8 September 2026, youtube.com was the most cited domain at 28, more than five times the next domain at 5.

That is corroboration of a direction and not a measurement of the web, and the limits matter. Five keywords in a single niche is a narrow panel, and one of them is ambiguous with a Power BI function, which pulls unrelated documentation sites into the results. It agrees with Ahrefs on the point where agreement is cheap to check.

The useful conclusion is about format rather than about surface. If a topic supports a genuinely good video, that video has a better chance in AI Overviews than the equivalent page does, and neither finding suggests writing two versions of the same article.

Do you need a separate SEO strategy for Google AI Mode?

No, on everything Google documents. This is where the evidence parts company with most of the advice written about the pair, including advice built on the same citation study quoted above. Different sources do not imply different work when both AI systems are sitting on the same eligibility rules and the same ranking substrate.

Google states one eligibility bar and applies it to both. 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, and Google says plainly that there are no additional technical requirements. Its July 2026 guide adds one more condition covering both surfaces together: the site must be included in Search generative AI features in Search Console.

The controls are shared too. Google's position is that AI is built into Search, which is why robots.txt directives for Googlebot are the control for how sites are crawled for Search, with nosnippet, data-nosnippet, max-snippet and noindex limiting what can be shown. There is no directive that admits you to one surface and excludes you from the other.

Google also addresses the vocabulary directly, which is unusual. On the terms answer engine optimisation and generative engine optimisation, it says that from Google Search's perspective optimising for generative AI search is optimising for the search experience, and is therefore still SEO. Take that as a statement about Google's surfaces only. Other assistants publish their own guidance and behave differently, which is the subject of the pillar on ranking in AI search.

None of this makes the two surfaces interchangeable to a reader or to a marketer. It means the lever is the same one: be snippet-eligible on the sub-questions a topic genuinely contains, and be worth quoting once you are. What differs is the odds, and the shape of the answer you land in.

Does query fan-out make Google AI Mode different?

No, because both surfaces use it. This is the most common factual error in comparisons of the two, and it is easy to check. Google's documentation states that both AI Overviews and AI Mode may use a query fan-out technique, issuing multiple related searches across subtopics and data sources.

Google defines the mechanism in its optimisation guide as a set of concurrent related queries the model generates to fetch additional relevant search results. Its worked example is the query "how to fix a lawn that's full of weeds" fanning out into searches such as best herbicides for lawns, removing weeds without chemicals, and preventing weeds in lawn.

Fan-out does explain why so few cited pages rank for the original query, on either surface. The pillar covers that evidence for AI Mode specifically, including Moz's finding on AI Mode citations and organic results, and the same pattern shows up in AI Overviews.

Where this turns into a trap is the advice that usually follows. A common recommendation is to write dedicated content targeting predicted fan-out queries. Google names that as a spam policy violation. Its optimisation guide (opens in a new tab) warns against creating separate content for every variation of how people might search, giving "other queries that people have asked, or fan-out queries" as its own example. Doing so primarily to manipulate rankings or generative AI responses violates its scaled content abuse policy.

Google adds that it is ineffective as well as against policy. A high quantity of pages does not make a site higher quality or more relevant, and Google says its systems have improved at understanding a page's relevance even without an exact query match. Covering the sub-questions a topic genuinely contains, on one good page, is the version of this that is not spam.

If you appear in an AI Overview, will you appear in AI Mode?

Usually, and with more competition around you. Ahrefs found that 61 per cent of the time, AI Mode included every entity the AI Overview had mentioned and then added more on top. Appearing in one is a reasonable predictor of appearing in the other, even though the cited URLs differ.

The reason is length. AI Mode responses are roughly four times longer and mention 3.3 brand or person entities on average, against 1.3 for AI Overviews. A longer answer has room for more names, so the same brand tends to survive the transition while sharing the space with sources that did not make the AI Overview.

That gap matters when reading an AI visibility report. An AI mention in AI Mode is worth less per mention than one in an AI Overview, because it is one of several rather than one of one, and a blended score across both surfaces will hide the difference. The distinction between a mention and an AI citation is separate again and covered in how AI assistants choose what to cite.

Both surfaces also answer without naming anyone a great deal of the time. In Ahrefs' dataset 59.41 per cent of AI Overview responses and 34.66 per cent of AI Mode responses contained no brand or entity at all, so the most likely outcome on either surface is that nobody gets named. AI Mode failed to cite any source 3 per cent of the time, against 11 per cent for AI Overviews.

Can you measure Google AI Mode and AI Overviews separately?

Not from each other, and not in Search Console. Google's Search generative AI performance reports, announced on 3 June 2026 and rolled out to all websites worldwide on 31 August 2026, report impressions for generative AI features on Search as a single group that covers AI Overviews and AI Mode together.

The report gives impressions broken down by page, country, device and date, and that data is also folded into the overall Performance report under the Web search type. There is no filter that isolates one surface from the other, and clicks, click-through rate and average position still carry no AI breakdown at all. The detail on what that report will and will not answer belongs to tracking your presence in Google AI Overviews.

The consequence for anyone comparing the two is that Google's own reporting cannot settle the comparison. Separating them means running a fixed panel of queries against each surface and recording what came back, which is also the only way to see the run-to-run churn described above rather than mistaking it for a trend.

The click side of the comparison is served worse still, and this is the gap worth knowing about. What an AI Overview citation earns has been measured repeatedly and at scale, and what a Google AI Overview citation is actually worth collects those figures. For AI Mode there is no equivalent measurement. The click-through rate quoted for it across dozens of pages traces back only to other pages quoting it, with no published sample, window or method behind it anywhere in the chain.

So the honest position on AI Mode click behaviour is that nobody has published one worth quoting. Anybody comparing the commercial value of the two surfaces is comparing a well-measured number against an absent one, and the comparisons that circulate simply do not say so.

What has changed since the AI Mode citation data was collected?

Enough that the figures above should be read as a dated measurement rather than as current behaviour. The overlap study used September 2025 data, twelve months before this article, and Google has changed both surfaces materially since. No comparable study has been published on the current versions.

At Google I/O on 20 May 2026 (opens in a new tab), Google brought AI Overviews and AI Mode into one AI Search experience, so that a question can move from a results page with an AI Overview to a follow-up in AI Mode. Google said the change was live that day across desktop and mobile worldwide. Two systems the study treated as separate destinations are now two stages of one path.

Google also made Gemini 3.5 Flash the default model for AI Mode globally on the same day, and said AI Mode had passed one billion monthly users with queries more than doubling every quarter since launch. Swapping the AI models underneath a surface is exactly the kind of change that would move a citation overlap figure in either direction.

What has not changed is the documentation the strategic conclusion rests on. Google's AI features guidance still states that the two may use different models and techniques, still sets one eligibility bar for both, and its July 2026 optimisation guide still names fan-out targeting as scaled content abuse. The measurement has aged; the instruction has not.

How RankX AI tracks Google AI Overviews

RankX AI made the citation panel this article draws on, and the disclosure belongs on a page arguing that a fixed panel is the only honest method. AI Overviews tracking produced the YouTube figure above, because it stores the cited domain behind every check rather than only the verdict on our own site.

The AI Overview Checker answers a single keyword free with no account. What none of it reaches is AI Mode, which exposes no per-surface reporting to anybody, and that is the honest limit on every comparison here including ours. The method on the AI Overviews side is set out in tracking your presence in Google AI Overviews.