Blog / Studies & Surveys / Gemini 3 wiped out 46% of cited domains and left 1 in 10 AI Overviews without any sources

Gemini 3 wiped out 46% of cited domains and left 1 in 10 AI Overviews without any sources

Written by

SEO and Content Marketing Expert at SE Ranking. Yevheniia blends strategic thinking with hands-on research to create insightful content on SEO, marketing, and AI in search

Reviewed by

Svitlana Tomko

SEO Research Analyst specializing in data-driven SEO analysis, experiments, and industry studies.

Feb 11, 2026

10 min read

After Google announced Gemini 3 as the default model for AI Overviews on January 27, 2026, the SEO community started noticing something strange: sources were disappearing from AI Overviews.

Lily Ray, for example, shared screenshots showing layouts where sources were missing or heavily reduced, replaced by several internal links and large YouTube elements. Google responded, attributing the changes to a bug they were working to fix.

Hey Google…

Whatever happened to including citations in AI Overviews? Where did the sources go? Almost all links here go to new Google searches/YouTube?

Are you seriously testing this? It's beyond unethical & unfair to site owners.@rajanpatel @rmstein pic.twitter.com/TmMrYVyA5l
— Lily Ray 😏 (@lilyraynyc) February 4, 2026

But the discussion raised an important question: Was this just a temporary glitch, or was AI Overviews behavior really changing?

We decided to let the data answer and compared two snapshots of AI Overviews from January 2026 (before and after the Gemini 3 rollout).

The results turned out to be more interesting than expected: The new model pulled back on sources, affecting all 20 niches we tracked. Throughout 2025, the trend was consistent growth, but Gemini 3 reversed that trajectory within a week.

Disclaimer: In this research study, we only analyzed sources that appear in the right-side bar, not in-line links.

Key takeaways

AI Overviews appear less frequently post-Gemini 3.

The appearance rate dropped from 60.85% to 55.21% after Gemini 3, reversing a consistent growth trend that had been building throughout all of 2025.
More than one in ten AI Overviews now show no sources at all.

The share of sourceless AIOs jumped from 0.11% to 10.63%. It’s critical for highly sensitive niches that need content validation and proper sourcing. Also, AIOs without links to click push search even further toward a zero-click experience.
Nearly half of all previously cited domains disappeared from AIO sources.

46.3% of domains cited before Gemini 3 no longer appear in AI Overviews after the rollout, while only the top-tier domains remained untouched.
The top-cited domains barely moved.

The top 10 most-cited domains remained almost identical after the Gemini 3 rollout. YouTube still leads (9.40% of all citations), followed by Reddit, Facebook, Quora, and Indeed. In some niches, YouTube has become the single most-cited source, including Healthcare.
The overall pool of cited domains shrank.

After the Gemini 3 rollout, the total number of unique domains appearing in AI Overviews decreased by 14% compared to the pre-rollout period.

AI Overviews pulled back after Gemini 3

Throughout 2025, AI Overviews had been expanding rapidly. In May 2025, our research showed AI Overviews appearing for approximately 28% of queries. By December 2025, that number reached 59.75% (more than a 2x increase in just 7 months). The trend continued into early January 2026, with the appearance rate climbing to 60.85% in the pre-Gemini 3 period.

Then Gemini 3 became the default model, and that trajectory reversed

After the rollout, the AIO appearance rate fell to 55.21%. This is a noticeable pullback from the steady growth we’d been tracking and not a niche-specific phenomenon. Every single category in our dataset saw a decrease. For example:

Business: from 78.66% to 74.42%
Finance: from 78.32% to 71.34%
Sports and Exercise: from 77.38% to 64.85% (one of the biggest drops)

Business

Career and Jobs

Finance

Relationships

Sports and Exercise

Education

Pets

Insurance

Technology

Legal

Healthcare

Travel

Entertainment and Hobbies

Food and Beverage

Ecommerce and Retail

News and Politics

Self-Care and Wellness

Cars

Fashion and Beauty

Real Estate

Niche

January 2026 (pre-Gemini 3)

January 2026 (post-Gemini 3)

Business

78.66%

74.42%

Career and Jobs

78.32%

72.14%

Finance

77.38%

71.34%

Relationships

75.70%

68.26%

Sports and Exercise

75.58%

64.85%

Education

71.24%

64.74%

Pets

69.98%

64.32%

Insurance

68.74%

63.18%

Technology

66.84%

59.88%

Legal

64.06%

57.42%

Healthcare

61.46%

55.70%

Travel

60.54%

53.69%

Entertainment and Hobbies

55.46%

51.02%

Food and Beverage

54.90%

49.86%

Food and Beverage

54.90%

49.86%

Ecommerce and Retail

54.16%

49.78%

News and Politics

52.82%

46.00%

Self-Care and Wellness

49.56%

44.40%

Cars

45.98%

43.06%

Fashion and Beauty

36.60%

33.76%

Real Estate

19.08%

16.34%

A note on the Real Estate niche: We recorded significantly fewer AI Overviews in this niche than in any other category in our dataset. This is likely explained by the niche queries, which are mostly local, transactional phrases like “houses for sale in Brooklyn” or “apartments for rent in Pensacola, FL.” For these types of queries, Google typically surfaces paid ads and the Places feature rather than AI Overviews.

We didn’t observe any patterns specific to long-tail keywords (queries of 5+ words). The same trends we identified across the dataset were present for such queries as well. This suggests that the Gemini 3 rollout affected AI Overview behavior broadly, regardless of query length or complexity.

More AI Overviews appear without sources

While fewer queries triggered AI Overviews, more answers started appearing without sources.

Before Gemini 3, only 0.11% of AI Overview responses appeared without a sources block. The model showed sources 99.9% of the time.

After Gemini 3, that number jumped to 10.63% AIOs without sources.

So, not only are fewer queries triggering AI Overviews, but among those that do, more than one in ten AI answers now appear with no sources at all.

More sources on average, fewer sources in reality

One of the first things we looked at was the average number of sources cited in AI Overviews. We took into account only those responses that included sources (49,339 post-Gemini 3 and 60,498 pre-Gemini 3).

At first glance, the results seemed reassuring.

The average actually increased slightly, from 11.55 to 12.1 sources per answer after Gemini 3 became the default model.

The average number of sources in AI Overviews

On the surface, this could suggest that AI Overviews are citing more sources than before. But averages can be misleading.

When we looked deeper, we noticed that what really changed was the distribution.

Gemini 3 created a split: Some answers now cite far fewer sources than before, while others cite significantly more. This pushes the average up even as many individual answers show fewer citations.

Here’s what the numbers reveal:

1 – 5 sources

6 – 10 sources

11 – 15 sources

16 – 20 sources

21 – 25 sources

26 – 30 sources

30+ sources

Distribution range

January 2026 (pre-Gemini 3)

January 2026 (post-Gemini 3)

Change

1 – 5 sources

6.07%

12.63%

+108%

6 – 10 sources

46.15%

38.63%

-16%

11 – 15 sources

30.75%

19.49%

-37%

16 – 20 sources

8.46%

15.32%

+81%

21 – 25 sources

5.60%

10.48%

+87%

26 – 30 sources

2.29%

3.01%

+31.44%

30+ sources

0.68%

0.45%

-33.82%

Before Gemini 3, only 6.07% of AI Overviews cited between 1 and 5 sources. After the rollout, that number more than doubled to 12.63% (a 108% relative increase). In other words, answers with very few sources became twice as common.

At the same time, mid-range citations dropped sharply:

Answers citing 6-10 sources fell from 46.15% to 38.63%.
Answers with 11-15 sources dropped even more, from 30.75% to 19.49%.

At the high end, answers with 16-20 sources increased by 81%, and those with 21-25 sources rose by 87%.

So, we ended up with a bimodal distribution where many AIOs cite very few sources, while a smaller group of AIOs cites many sources, pulling the average upward and masking what’s actually happening for most queries. This means Gemini 3 didn’t increase citations, but it changed how citations are distributed across answers.

Top-cited domains barely changed

Gemini 3 didn’t shake things up much here, and the most-cited domains after the rollout look almost identical to those before it:

YouTube: 54,584 citations (9.40%)
Reddit: 25,496 (4.39%)
Facebook: 10,203 (1.76%)
Quora: 8,513 (1.47%)
Indeed: 8,466 (1.46%)

Wikipedia, US News, NIH, Amazon, and Bankrate round out the top ten.

Compared to the pre-Gemini 3 data, the ranking barely changed: YouTube is still the leader, followed by Reddit. The top five domains were identical in composition, with only minor shifts in shares.

Looking deeper into niches reveals a more dynamic picture in some topics. For example:

Travel: Tripadvisor dominated before Gemini 3, holding the top position with 5.53% of citations. But after the rollout, it fell to fourth place (4.01%). At the same time, US News rose to second place (4.57%), and Expedia moved into third place (4.18%).
Finance: Bankrate held a strong position with 5.54% citations before Gemini 3. After the rollout, its citations dropped to 5.18%, losing the first place to YouTube with now has 6.41% of citations in the niche.
Healthcare: NIH (National Institutes of Health) ranked first, accounting for 6% of citations before Gemini 3. After the rollout, its share declined slightly to 5.4%. And YouTube, again, became the top-cited source for healthcare queries in AIOs.

Gemini 3 hit sites with small citation counts the most

When a new model launches, we’d expect some domain shuffling. Some sites gain citations, others lose them. That’s normal. What’s not normal is the scale of change we observed.

Out of the 89,262 unique domains cited before Gemini 3, 41,336 domains (46.3%) completely disappeared from AIO citations after the rollout. That’s nearly half of all previously cited domains no longer appearing in AI Overviews.

At the same time, 30,267 new domains (33.9%) appeared after Gemini 3 launched.

Together, that’s over 70,000 domains affected by the model change.

And here’s the kicker: only one of these domains was in the top 500 by citation count.

The domains at the top (those with large numbers of citations) remained virtually untouched. Among the top 500 most-cited domains, 99.98% kept their citations after the rollout (essentially every single one). The changes hit sites with small citation counts the hardest. Domains with just 3 citations held at 81.39%, and domains with only 1-2 citations saw stability drop to just 49.96%.

Citations became more concentrated

After the Gemini 3 rollout, we ended up with 14% fewer unique domains in the dataset overall. This drop is likely influenced partly by the increase in answers with no sources at all, but the trend toward consolidation is clear.

To measure how concentrated citations have become, we used the Herfindahl-Hirschman Index (HHI), a standard metric for measuring market concentration. The higher the HHI, the more concentrated the distribution, meaning fewer domains are capturing a larger share of citations.

Before Gemini 3, the overall HHI stood at 1.0. After the rollout, it increased to 1.22. This pattern played out across nearly every niche in our dataset.

Ecommerce and Retail saw the biggest concentration increase, from 4.47 to 5.25 (driven largely by YouTube, which now accounts for 20.33% of all citations in this niche)
Career and Jobs went from 4.14 to 4.25
Fashion and Beauty increased from 2.52 to 2.8

Ecommerce and Retail

Career and Jobs

Entertainment and Hobbies

Fashion and Beauty

Sports and Exercise

Pets

Self-Care and Wellness

Relationships

Cars

Food and Beverage

Business

Travel

Healthcare

Real Estate

Education

Finance

News and Politics

Technology

Legal

Insurance

Niche

January 2026 (pre-Gemini 3)

January 2026 (post-Gemini 3)

Ecommerce and Retail

4.47

5.25

Career and Jobs

4.14

4.25

Entertainment and Hobbies

2.79

3.21

Fashion and Beauty

2.52

2.8

Sports and Exercise

2.4

2.35

Pets

2.35

2.34

Self-Care and Wellness

1.97

2.19

Relationships

1.86

Cars

1.6

1.97

Food and Beverage

1.53

1.89

Business

1.41

1.82

Travel

1.4

1.21

Healthcare

1.37

1.2

Real Estate

1.12

1.2

Education

1.01

1.17

Finance

0.98

1.06

News and Politics

0.83

0.97

Technology

0.81

0.97

Legal

0.67

0.68

Insurance

0.52

0.5

The takeaway? After Gemini 3 became the default model powering AI Overviews, citations shifted, but most importantly, they concentrated. A smaller number of domains now capture a larger share of all citations, while tens of thousands of smaller sites lost visibility entirely.

Traditional media holds steady in AI Overviews

We tracked 33 major media outlets, including The New York Times, BBC, CNN, The Wall Street Journal, and The Guardian.

Before Gemini 3, these media outlets collectively accounted for 2.17% of all citations. After the rollout, that share decreased only slightly to 2.14%.

Individual media outlets showed only minor fluctuations:

US News: from 0.77% to 0.75%
Yahoo: from 0.31% to 0.34%
The New York Times: from 0.13% to 0.15%
CBS News: from 0.09% to 0.07%

These movements are small enough that they could reflect normal variance rather than systematic changes in how Gemini 3 evaluates news sources.

Despite the overall modest presence of media sites, one domain stands out: usnews.com. It is the only media website in the global Top 10 by number of citations, ranking seventh overall. And it also shows up in top 10 lists across multiple categories, not only News and Politics:

Education: 4.52% of all citations
Travel: 4.57%
Insurance: 1.42%
Cars: 1.44%

This cross-niche presence is unusual. Most domains in our top 10 lists are niche-specific, except for platforms like YouTube, Reddit, and Facebook that dominate broadly.

This pattern becomes even more interesting when viewed in a longer timeframe. In our May 2025 AI Overviews research, not a single media outlet appeared in the global Top 10 by citation volume. Now, US News has claimed the seventh spot. This suggests that changes in how AI Overviews source and cite content extend beyond the Gemini 3 rollout.

Research methodology

For this study, we analyzed 100,000 keywords divided evenly across 20 niches (5,000 keywords per niche). All data was collected from the US market.

We captured data at four points in time:

November 10, 2025
December 13, 2025
January 5, 2026 (before the Gemini 3 announcement)
January 29, 2026 (after the Gemini 3 rollout)

We analyzed sources that appear in the right-side bar of AI Overviews only. Inline links were not included in this analysis.

Our findings reflect patterns observed within our defined dataset, timeframe, and methodology. AI Overviews are volatile and evolving, and source selection may vary over time, across regions, or under different circumstances.

We admit that there might be alternative interpretations of the data and encourage you to consider our findings as one data point in a broader conversation.

Final thoughts

For the industry, the most important signal is the direction everything is moving in: fewer triggers, fewer sources, fewer cited domains, more concentration at the top. Big names like YouTube or Reddit weren’t affected. The impact fell almost entirely on the smaller publishers and niche websites that built their presence in AI Overviews over the past year.

Google has consistently framed its AI search improvements as a way to help users find better answers faster. That may well be true. But from a publisher’s perspective, the Gemini 3 data suggests that those “best-in-class” answers mean pulling content from fewer but more familiar sources. Or pulling content from smaller sites without any visible attribution, leaving publishers with no credit and no traffic.

We’ll keep an eye on the situation and share more insights in our upcoming studies. Stay tuned.

144 views

Written by

Yevheniia Khromova

Yevheniia is an SEO and Content Marketing Expert at SE Ranking. With a profound expertise in digital marketing and content strategy, she creates impactful SEO-friendly content that resonates with diverse audiences. Yevheniia believes simplicity is key as it helps reduce the clutter in our heads and focus on what matters. This is the principle she uses in her writing. She follows the latest SEO, content, and marketing trends and incorportaes them into her workflow. She also enjoys cooking, learning new spices and flavors, traveling, and playing board games.