Google AI Overview disrupts search marketing
| |

10 Essential Insights From Our Google AI Overviews Study

Introduction: Why AI Overviews Matter for Digital Marketers

Google’s AI Overviews (AIO)—the generative summaries sometimes appearing at the top of search results—are creating new opportunities and challenges for SEO. They provide quick, synthesised answers, often citing multiple organic pages. For marketers, AIOs can potentially reduce clicks on traditional organic links while also spotlighting the websites Google trusts most.

As we cover in our analysis of the top issues facing digital marketing in 2025, we believe AIOs will cause substantial disruption to search marketing in the months and years to come.

To understand how AIOs work “in the wild,” our team analysed 4,000 search queries (2,000 each from the US and the UK). These queries covered awareness, consideration, and conversion keywords as well as question vs. non-question formats and short-tail, medium-tail, and long-tail variants. The analysis was carried out on 23rd January 2025.

Below, we share our 10 biggest takeaways.

Note: For those who want to dive into statistical significance tests (e.g., Kruskal-Wallis, Z-statistics, McNemar’s test), check out the Detailed Methodology & Stats at the end.

1. AIOs Appear for Almost Half of Searches

Across the full range of search terms, 47% triggered an AI Overview while 53% did not. In other words, you’ve got nearly a 50-50 chance of seeing an AI Overview for a given query.

As Google refines its generative approach, this rate may well increase—making it crucial to understand how (and why) AIOs surface.

2. Minimal Geographic Differences: US vs. UK

Breaking down the same dataset by location, we see a 46.63% incidence for the UK and 47.76% for the US. Given how closely these figures align, it appears that Google’s AIO rollout is similarly advanced in both markets.

If you’re ranking for queries in either region, anticipate a comparable chance of generating an AI Overview.

3. Question-Based Queries Are Far More Likely to Yield AIOs

One of the standout findings involves query intent:

Question-format keywords: ~85% show an AIO

Non-question keywords: ~43% show an AIO

If your audience tends to type “How do I…?” or “What is…?”, these types of queries are extremely likely to trigger an AIO.

4. Funnel Stage Influences AIO Appearance

We classified each query as awareness, consideration, or conversion. Here’s how often each stage displayed an AI Overview:

  • Awareness: ~56% AIO incidence
  • Consideration: ~23%
  • Conversion: ~42%

The disparity suggests that broader, top-of-funnel queries (awareness) are more likely to get summarized. Consideration-stage queries often have more nuanced, comparison-oriented information that might not lend itself as readily to a single overview. That said, nearly a quarter of consideration queries still generated an AIO.

5. Long-Tail Keywords Trigger AIOs Most Frequently

Next, we grouped queries into short-tail, medium-tail, and long-tail:

  • Long-tail: ~65% include an AIO
  • Medium-tail: ~49%
  • Short-tail: ~34%

Specific, more detailed queries tend to invite more thorough AI Overviews—likely because Google sees a stronger need to synthesize multiple sources for these narrower topics.

6. The Nuanced Story of AIO Length

We also looked at the word count of AI Overviews based on funnel stage, keyword length, and whether it was a question or not.

Although these differences are statistically significant, the real-world impact may be relatively small. For example, consideration queries produce slightly longer overviews, but this could also be due to the complexity of the topic rather than the funnel stage itself.

In other words, AIO length is not driven exclusively by these factors—they’re just part of the picture. Content complexity and Google’s available data likely play a bigger role in shaping final word counts.

7. Nearly All AIOs Cite the Top 10 Organic Results

If you’re wondering whether organic rankings still matter when an AI is rewriting the SERP, the answer is a resounding yes. We found:

  • 97.3% of AIOs cited at least one result from the top 10 organic listings.
  • Only 2.7% didn’t cite any top 10 result.

Put simply, almost every AI Overview references Page 1. If you’re outside the first page, the chance of being included in the AIO drops significantly. Secondly, top ranked pages essentially shape the narrative of AI Overviews.

8. The Two-Tier Citation Pattern: Positions 1–7 vs. 8–10

Digging deeper, we discovered a sharp divide in citation rates:

  • Positions 1–7: ~83–85% citation rate
  • Positions 8–10: ~56% citation rate

Our statistical tests show no meaningful difference between the top 3 positions and positions 4-7.

However, there’s a highly significant drop from positions 1–7 to positions 8–10. If you want to influence the AI Overview, aim to be within the top 7. After that point, citation rates plummet.

9. Number of Citations: More About Availability than Keyword Type

We also measured how many sources are typically cited in an AIO.

While the averages for awareness, consideration, conversion, and different keyword lengths varied (e.g., ~5–6 citations), the data suggests these slight differences don’t strongly correlate with the funnel stage or keyword type.

Instead, the number of citations appears to be driven by how much topically relevant content Google deems credible and already ranks highly. In other words, Google’s AI pulls from however many high-ranking pages it believes best serve the user query—so the richer the existing pool of quality and relevant content, the more citations you’ll see.

10 . Why Top Organic Results Are the Bedrock of AIO

One of our most compelling discoveries is also the most logical: AI Overviews essentially stand on the shoulders of Google’s ranking algorithms. Consider the following:

  1. World-Class Relevance: Google has spent decades perfecting how to rank content. It’s far more efficient for them to build AI Overviews using the top-ranked pages than to re-invent the relevance engine from scratch.
  2. Consistency vs. Contradiction: The AI Overview typically aligns with the consensus of top results. If an AIO were to contradict the highest-ranked sources, it would raise questions about whether the AI is incorrect or Google’s rankings are flawed. Either scenario undermines user trust.

For brands, this means SEO fundamentals still matter—perhaps more than ever. If you can get into (and stay in) the top 7 organic positions, you stand a much higher chance of:

  • Being front and centre when an AI Overview appears.
  • Shaping the narrative around a topic

Conclusions & Next Steps

The bottom line is that AI Overviews rely on the top organic results for both content and credibility. While AIOs may feel like a shortcut for users, relevance and authority remain rooted in Google’s ranking algorithms. With nearly half of all queries we examined triggering an AIO, your best strategy is to:

  • Continue optimizing for Page 1, especially the top 7 positions.
  • Craft comprehensive, relevant content that serves user intent (particularly for question-based queries)
  • Monitor funnel-stage performance and adapt content strategy for awareness vs. consideration vs. conversion.

Generative AI may shift how users consume information on the SERP, but it doesn’t replace the fundamental importance of ranking highly for competitive keywords.

Boost Your Marketing with AI Overviews


Detailed Methodology & Stats

  1. Data Collection & Sample
    • 4,000 queries total: 2,000 each in the US and UK. Selected as indicative of client behaviour modelled over buying funnel.
    • Queries were classified by funnel stage, keyword length, and question vs. non-question.
  2. AIO Incidence & Word Count
    • We recorded whether an AIO appeared and measured its word count if it did.
    • Sub-analyses split results by region, funnel stage, question format, and keyword length.
  3. Citations
    • For each AIO, we noted whether it cited organic positions 1–10, and how many total sources it included.
  4. Statistical Tests
    • We employed Kruskal-Wallis for comparing median word counts across multiple groups.
    • We used Z-statistics and McNemar’s tests to compare proportions (e.g., top 3 vs. positions 4–7, 1–7 vs. 8–10).
    • Significant differences suggest we can be confident the patterns (e.g., citation rates) are not due to random chance.

Similar Posts