[DON’T HAVE TIME? READ THIS]
Are AI Mode and AI Overviews the same thing with different word counts?
No. I analyzed 730,000 query pairs and here’s what happened:
- 87% different sources – Only 13.7% citation overlap between systems
- 84% different wording – Just 16% word-level overlap; they write independently
- 86% same meaning – They agree on conclusions, not sources
- AI Mode shows 2.5x more brands – 3.3 entities vs 1.3 in AI Overviews
- 61% visibility carryover – Get cited in AI Overview? You’ll probably appear in AI Mode (with extra competitors)
Bottom line: These are separate systems using different sources to reach similar conclusions. Optimize for both separately.
I analyzed a dataset from Ahrefs covering 730,000 response pairs from September 2025. These systems cite different sources, use different words, but somehow agree on answers 86% of the time.
Here’s what the data showed.
Different Sources, Same Conclusions
Only 13.7% of citations overlapped between AI Mode and AI Overviews. When narrowed to just top 3 citations, overlap increased slightly to 16.3%.
Translation: 87% of the time, they pull from completely different websites for identical queries.
Domain preferences showed clear patterns:
AI Overviews prefers:
- YouTube (cited more than Wikipedia)
- Reddit for community content
- Videos and core pages (homepages, category pages) – 2x citation rate vs AI Mode
AI Mode prefers:
- Wikipedia (28.9% of citations vs 18.1% in AI Overviews)
- Health websites (nearly 2x more citations)
- Quora (3.5x more citations than AI Overviews)
- Facebook (2x more citations)
AI Mode leans encyclopedic and medical. AI Overviews leans visual and community-driven.
Both favor article-format content overwhelmingly, but the secondary preferences diverge completely.
They Don’t Copy Each Other’s Writing
Jaccard similarity scored 0.16.
Only 16% of unique words matched between responses. They started with identical first sentences 2.51% of the time. Fully identical responses appeared in 0.51% of cases.
This proves AI Mode isn’t just adding detail to AI Overview’s answer. Both systems research queries independently and generate new responses from scratch.
Even when they semantically agree (which is 86% of the time), they express it completely differently.
But They Agree on the Answer
Semantic similarity averaged 86% using cosine similarity measurement.
On a 0-to-1 scale (where 1 = identical meaning), nearly 90% of response pairs scored above 0.8.
9 out of 10 times, both systems agreed on what to say. Different words, different sources, same core answer.
Why? Google’s documentation confirms both use “query fan-out” – running multiple related sub-queries to find supporting content while generating responses.
Since they use different models and techniques, they cite different sources while reaching similar conclusions.
It’s like asking two SEO consultants the same question. Different case studies, different examples, but the strategy aligns.
AI Mode Cites Way More Brands
AI Mode responses run 4x longer and mention 3.3 brands/people on average. AI Overviews mention 1.3.
Test case: “cloud storage alternatives”
AI Overview: 7 brands mentioned once each, mostly at the end
AI Mode: 23 brand mentions, repeated throughout, starting from sentence one
Here’s the important part: 61% of the time, AI Mode includes every entity AI Overview mentioned, then adds competitors on top.
Example: AI Overview cites Mayo Clinic. AI Mode cites Mayo Clinic + Cleveland Clinic + WebMD.
For SEO: If you’re in AI Overview, there’s a 61% chance you appear in AI Mode too – but you’re sharing space with brands that didn’t make the shorter version.
When Both Skip Citations Entirely
59.4% of AI Overview responses contain zero brands or entities. AI Mode: 34.7%.
About one-third (32.8%) of all responses mentioned no brands at all. Usually informational queries where no brand is expected (“meditation before bed,” “Nov 21 zodiac”).
Citation gaps:
- AI Mode: 3% have no sources cited
- AI Overviews: 11% have no sources cited
This happened for edge cases: simple math, sensitive content, help redirects, unsupported languages.
Why AI Mode cites more consistently? Longer responses need grounding. Plus users expect transparency in chat experiences more than quick search results.
What This Actually Means
Track both separately. 13.7% source overlap means success in one doesn’t guarantee the other. Use tools like Ahrefs’ Brand Radar or similar tracking platforms.
Build topical authority. 86% semantic similarity shows both look for the same themes, just expressed differently. Cover topics comprehensively, not just specific phrases.
Format strategically. AI Mode’s 97% citation rate favors long, well-sourced, encyclopedic content. AI Overviews favors video and Reddit. Same content won’t work equally in both.
Expect competition in AI Mode. 61% carryover rate means you’ll likely appear if you’re in AI Overview, but with additional competitors.
Make encyclopedic content. Wikipedia appears in 28.9% of AI Mode citations vs 18.1% in AI Overviews. Your content needs reference-level depth.
Treat these as separate channels with overlapping goals but different execution. Don’t assume visibility in one means visibility in the other.
