Obility Editorial · · 4 min read
AI Overviews and AI Mode answer the same question from different sources
A marketing team that finds its brand named in a Google AI Overview often assumes the work is done for Google. It is an easy assumption. Both features sit inside Google Search, and Google tells site owners there are no extra requirements to appear in either. Google’s own guide to AI features then adds a line that matters: AI Mode and AI Overviews may use different models and techniques, so the responses and links they show will vary. The same guide says AI Overviews appear only when Google’s systems judge them additive to classic results, and often do not trigger at all.
Ahrefs measured how far apart the two are. In a study published in December 2025, it captured an AI Overview and an AI Mode answer for the same query across hundreds of thousands of US queries, using September 2025 data. The two cited the same URLs only 13.7 percent of the time, rising to 16.3 percent when only the top three citations were compared. They shared 16 percent of their unique words. Yet their meaning scored 86 percent on semantic similarity. They mostly agree on what to say and disagree on where they found it.
The 13.7 percent needs a yardstick, and Ahrefs supplies one
A low overlap between two surfaces means little until you know how much one surface overlaps with itself. Ahrefs published that number a month earlier. Tracking more than 43,000 keywords over a month, it found that 45.5 percent of the sources cited in an AI Overview were new from one observation to the next, and that 46 percent of named entities changed as well.
So AI Overviews agree with themselves on roughly half their citations, while AI Mode agrees with AI Overviews on about one in seven. The two figures were not computed in exactly the same way, so treat the comparison as rough. It still points one way: the gap between the surfaces is real, and it is wider than the churn inside either one. One snapshot of each, compared side by side, mixes the two effects together.
The 61 percent figure is not the chance your brand carries over
The number most often repeated from the study is that a brand named in an AI Overview has a 61 percent chance of also appearing in AI Mode. Ahrefs’ own summary phrases it that way. The body of the post describes something else: 61 percent of the time, the AI Mode answer contained every entity the AI Overview had mentioned, then added more. That is a measure of whole lists, counted per query. It is not a per-brand probability. A single brand could carry over most of the time while the full list rarely does, because one missing name breaks the match.
The denominator is also unstated. The same study found that 59.41 percent of AI Overviews named no brand or entity at all, and an AI Mode answer trivially contains every entity of an overview that has none. The post does not say whether those pairs were left out. If they were counted, nearly all of the 61 percent is empty overviews. Either way, the figure cannot tell you your carry-over rate on your questions. What the study does show clearly is crowding: AI Mode answers named 3.3 entities on average against 1.3 for AI Overviews, so when you do carry over, you share a longer answer with more competitors.
Comparison questions lean toward AI Mode, and the study’s queries did not
Ahrefs drew its queries from its Brand Radar dataset, and its examples of brand-free answers include zodiac sign lookups and meditation tips. About a third of all answers in the dataset named no brand or person. That mix reflects search at large, not a buyer comparing vendors. Google describes AI Mode as most helpful for questions that need further exploration, reasoning or complex comparisons, which is where shortlist questions live.
That has an awkward consequence for teams that only check overviews. A question that triggers no AI Overview can still have an AI Mode answer that names three of your competitors, and a check built around overviews will record that question as having no AI answer at all.
Track the two surfaces as separate engines
Treat AI Overviews and AI Mode as two engines with two scores. For each buying question, record whether an overview appeared, whether your brand is named in it, and whether your brand is named in the AI Mode answer. Repeat the runs over a few weeks and count presence per surface, since a single capture of either one is unstable on citations. Compare brand presence rather than cited URLs. The study shows the two surfaces reach similar conclusions through different pages, so a URL dropping out of one tells you less than a brand dropping out of both.
Obility’s Answer Engine Insights keeps the engine, prompt wording, language, geography and collection time attached to each observation, so an AI Overview result and an AI Mode result for the same question stay comparable without being merged into one Google number. AI Mode is among the engines listed on our site, and the engines and markets covered for your team are confirmed at the demo.
One boundary is worth stating. If your brand appears in AI Overviews and is missing from AI Mode, nothing in this research suggests one content change fixes both. Look at what AI Mode cites for that question first. Ahrefs found it leans harder on reference sources, with Wikipedia in 28.9 percent of AI Mode citations against 18.1 percent for AI Overviews, and the gap usually sits in those sources rather than on your own page.