AI Overviews Research: How Google’s AI answers vary across five states in the US

AI Overviews are evolving, and so is how they present and cite information. Since our last AIO recap in 2024, Google has been actively experimenting by tweaking formats, adjusting layouts, adding internal links within summaries, and more. With AIOs appearing more frequently in SERPs, we took the opportunity to investigate deeper.

This round of research focuses on AI Overview citation patterns across five states in the US: Colorado (Denver), New York (New York), California (Los Angeles), Texas (Houston), and Washington, DC. We explored how AI-generated summaries vary by region and what it will mean for website visibility and SEO strategies moving forward.

Central to this research was our AI Results Tracker, which gathered regional insights. You can also use this tool to keep track of AIO-related content (and in real time). Beyond tracking sources featured in AI-generated answers, this tool can monitor keywords for their AIO presence, competitive dynamics, and regional differences across countries like the US, UK, Canada, Australia, Japan, South Korea, and New Zealand (more to come soon). You can also access AIO SERP feature detection directly in our Keyword and Competitive Research Tools. The SE Ranking platform is also free for 14 days.

Now, let’s break down what we discovered through our research and how it can impact your SEO strategy.

General insights into AI Overviews across the US

How do AI Overview appearance rates differ across each state?

Overall data by state 

We analyzed 100,013 keywords and found that about 30% (from 27,753 to 28,664 keywords) of queries triggered an AI-generated response across five US locations, with only slight variations:

Paul Andre de Vera

SEO Consultant | Host and producer of the SEO Video Show

I’ve seen a clear increase, especially in the San Francisco Bay Area where new search behaviors tend to catch on faster. Across both enterprise and local business sites I manage, traffic from AI-driven engines and AI Overviews has grown over 600% year-over-year, now representing around 3% of total organic traffic for some properties.

Chris Raulf

International SEO and AI Expert

Over the past few months here in Colorado, I’ve definitely noticed an increase in how frequently Google’s AI-generated search result overviews are showing up. From a Micro-SEO perspective, this shift isn’t just interesting, it’s critical to understand. It means that we have to be even smarter, more agile, and more intentional about how we optimize content. It’s no longer just about ranking well traditionally; now, it’s just as important to position content to become part of Google’s AI-generated overviews.

Janna Pugh

SEO Specialist at Five9 | Global Search & Agency Awards Judge

I have definitely noticed a staggering increase in AIOs personally and professionally, primarily within the last 6-8 weeks. There is a massive amount of nuance when it comes to the queries that Google is serving AI Overview results for. It is my opinion that, while this is a way for Google to help modernize SERPs to match a simple LLM layout, they still have a long way to go before AIOs are accurate enough to replace a traditional search journey.

Barry Schwartz also confirms this variation in AI Overview appearance rates:

Barry Schwartz

President and Owner of RustyBrick, Inc.

My primary Google account is a workspace account; so for some reason, AI Overviews do not trigger all that often for me. But when I test AI Overviews on my normal Google account, I see them more often.

Data by location and niche

When analyzing by niche, AI Overviews appeared most frequently in the Relationships category (60.54%–62.38%). This was followed by Business (56.38%–57.52%) and Education (48.70%–51.86%).

On the other end of the spectrum, Fashion and Beauty resulted in the least number of AIOs—just 1.34% to 1.46%. Other low-ranking categories included E-commerce and Retail (2.14%–2.48%) and News and Politics (3.76%–4.50%).

How many sources do AI Overviews cite in each state? 

We analyzed the average and maximum number of sources (also referred to as references) included in AI Overview responses and found consistent results across all five US locations:

The average reference count across regions varies by just 0.13 links. Los Angeles recorded the highest average at 13.41 references, while New York had the lowest at 13.28. 

With maximum counts, Houston and Los Angeles lead with 95 links in a single response. Denver, New York, and Washington recorded slightly lower peaks at 77, 87, and 82 links, respectively.

The maximum reference count across regions varies by 18 links. 

Note, the minimum values for each state are 1 source, so we didn’t give it much attention.

What is the overall distribution of link counts in AI Overview responses?

We analyzed the overall distribution of the number of links included in AI Overview responses and found that the most common answers contained:

What is the link count distribution in AI Overview responses by state?

We observed that the most common responses across all locations studied were those with 8-9 references. This trend was consistent across all regions.

However, regional differences did emerge:

• In California and New York, responses with 8 references are the most common (6.57% and 6.56%, respectively).

• In Texas and Colorado, responses with 9 references are the most common (6.60% and 6.59%, respectively).

• Washington has a higher frequency of answers with 6 (5.64%) and 12 (5.84%) references compared to all other regions. 

• Responses with over 50 references are rare across all regions, and differences between regions aren’t significant due to their limited frequency.

These observations highlight that while regional differences exist, the general trend in the number of references remains consistent across U.S. states, with a slight regional variation in specific reference counts.

Are there significant differences in AI Overview response lengths between states?

We found no significant differences in text length by state.

Note: The average character and word counts presented below, except for the highlighted differences, have been rounded for clarity.

Character count

The difference between the longest (New York) and shortest (Washington, DC) responses is only 12.6 characters, less than 1% of the total length.

Word count

The difference in word counts is similarly minimal, only 2.38 words (less than 1%) between the highest and lowest averages.

Do AI Overviews appear next to SERP features in each state?

For this analysis, we observed five SERP features that appeared most frequently in Q4 2024. Based on the data combined from all five states (141,507 total AIO appearances), AI Overviews almost always appear with at least one of the five analyzed SERP features, 99.25% of the time (140,444 of cases). Only 0.75% of AIOs appear without any of these SERP features. 

Variation by state is minimal for both AI Overviews with and without other SERP features. 

• No SERP features: From 0.70% (Los Angeles) to 0.81% (Denver)
• At least one SERP feature: From 99.19% (Denver) to 99.30% (Los Angeles)

We dived even deeper into the SERP features analyzed to observe the following frequencies.

• People Also Ask is by far the most frequent companion to AI Overviews, appearing in 98.54% of cases. This pairing makes sense as both elements are designed to answer user questions and complement each other.

• Video snippets are the second most common feature, appearing next to 45.17% of AI Overviews. This indicates that Google often enhances text-based AI answers with visual content to meet user intent.

• Reviews appear in 25.30% of cases, typically with queries related to products, services, or businesses where user feedback is valuable. 

• Images appear in 11.67% of cases, mostly accompanying visual-oriented queries where supplementary images help users better understand the AI-generated content.

• Featured snippets appear side-by-side with AI Overviews in only 8.63% of cases, indicating that Google often chooses between these two answer-focused elements rather than displaying both. This is also nearly 4 times less frequent than we observed earlier for US SERPs.

Location (at least in the US) has minimal impact on these patterns. The variation across the five states observed is extremely small. For instance, People Also Ask ranges from 98.46% (Washington, DC) to 98.69% (Los Angeles). All other features show similarly tiny variations across locations.

Are there any sources in AIO responses that link to Google’s search results?

Recently, the SEO community noticed that Google was using “internal links” within the text of AIO answers, redirecting users to organic results. To understand this phenomenon better, we analyzed how often these links appear in AIO responses.

Based on the data combined from all five states (141,507 total AI Overview appearances), our data analysis shows that 43.42% (61,437 times) of AI Overview responses contain links to Google organic results, while 56.58% of responses do not. 

The variation across states is minimal, with responses containing links to Google ranging from 43.11% (Washington) to 43.75% (Houston).

Interestingly, the number of links to Google’s organic results is not limited to one or two links. On average, AIO responses contain 4-6 links to Google’s organic search results, which is a lot, showing Google’s AI-generated answers link back to its resources frequently.

Chris Raulf

International SEO and AI Expert

Google citing its own organic results 43% of the time in AIOs certainly raises an eyebrow or two. It could simply reflect Google’s effort for internal consistency, user experience, or a practical linking strategy. On the other hand, it could hint at Google’s leveraging AIOs for increased interaction with its own services. Regardless of intent—and it’s not something we can definitively answer without deeper investigation—it’s something marketers and the broader digital community should keep an eye on. Why? Because if Google’s intent skews toward its internal content and reduces visibility for independent publishers, we’ll need to become even smarter in positioning our content strategically within Google’s ecosystem. Specifically, spending more time on winning SERP features, structured data markup, and producing precisely optimized localized content becomes essential—again aligning beautifully with Micro-SEO.

Keyword metrics and AIO presence across each observed state

Does keyword search volume impact AIO appearance differently by state?

The data clearly shows that there is an inverse trend between keyword search volume and the percentage of AIOs. As search volume increases, the AIO appearance frequency drops a lot in all regions studied.

Here’s the breakdown by search volume:

Does keyword CPC impact AIO appearance rates differently in each state?

Unlike search volume, the relationship between CPC and AIO appearance is not linear. Overall, in this US-focused dataset, mid-range CPCs ($2–$5) have the highest percentage (32.02%) of AIO appearances. The lowest percentages occur in the $0.5-$1 range (20.07%) and 10+ (17.31%) range. 

Here’s a breakdown of AIO appearances by CPC:

Does keyword difficulty impact AIO appearance differently by state?

We also analyzed how keyword difficulty (KD) impacts the likelihood of AI Overview appearances across US regions. The relationship is non-linear, similar to CPC trends.

Overall, the highest AIO appearance rates occur for mid-difficulty keywords (21–40). The lowest rates are observed for the most difficult keywords (81–100).

Overall breakdown:

Does the number of words in a search query affect the appearance of AI Overviews?

Our analysis reveals a clear correlation between query length and the likelihood of triggering AI Overviews in search results. In general, the longer the query, the more likely it is to generate AIOs.

The numbers below represent the overall AI Overview appearance rates across all states. State-by-state analysis shows minimal variation in this pattern. The difference between states is typically less than 3% for any given query length. All states show the same clear upward trend as the query length increases.

Here’s a breakdown of our findings:

This trend reveals several key patterns:

AI Overview sources and key citation patterns by state

Does the average length of AI Overviews influence how many sources it cites?

Yes, and quite significantly. Our data shows a strong positive correlation: the longer the answer, the more sources it includes. For example:

• Answers under 600 characters include an average of 5.31 sources.

• Answers in the 6600–7200 character range jump to 28 sources. That’s more than a 5x increase.

The most noticeable spike happens in the medium-length range: answers between 3000 and 3600 characters already cite around 19.76 sources, nearly 4x more (from 5.31 to 19.76) than the shortest ones (up to 600 characters).

What are the top-linked websites in AI Overviews for each state?

Google.com is the absolute leader in all five states studied, with mentions ranging from 82,685 (Houston) to 79,139 (New York). Regional variation has little to no influence on which domains get cited.

The other top 9 linked domains (nearly identical across locations) are:

This points to a standardized source selection within Google’s AI Overviews. Geographic location doesn’t significantly influence which domains are referenced.

When looking at the percentage of responses that mention these domains, Google.com dominates at ~44% of all responses across all locations.

The next tier of domains each appear in about 13% of responses across all states observed:

• Reddit.com (13.20%-13.38% of responses)
• Quora.com (13.28%-13.75% of responses)
• Youtube.com (13.11%-13.50% of responses)
• En.wikipedia.org (12.83%-13.32% of responses)

How do cited domains perform by traffic, keywords, referring domains, and backlinks?

We used SE Ranking’s Competitor Analysis and Keyword Research tools to examine the domains featured as sources in AI answers. 

Traffic data

The overall traffic distribution shows that 25.30% of domains (21,151) receive 1,001-10,000 visits.

When looking at state-by-state data, we saw only slight differences in traffic patterns across all five locations.

Keywords

As for the overall keyword distribution, the largest segment (35.80% or 30,684 domains) ranks for 1,001-10,000 keywords. 

There was little variation in keyword visibility across the five states.

Referring domains 

The referring domain data provided by our Backlink Checker shows that nearly half of all domains (29,319 or 34.50%) have between 101-500 referring domains. 

The number of referring domains varied little from state to state.

Backlinks 

The backlink distribution shows that Google’s AIOs tend to cite domains with large backlink profiles.

Despite geographic spread, backlink counts were uniform across cities, with only slight variations in each range.

How many unique and exclusive domains are cited across each state?

This domain diversity analysis shows relatively small differences across the five locations studied. Total source citations (link counts) range from 368,209 in New York to 382,812 in Houston, showing only a 4% variation.

To clarify: unique domains are the total number of distinct domains cited in a region, while exclusive domains appear only in that region and nowhere else.

The share of exclusive domains ranges from 21.31% in Washington, DC to 22.72% in Houston.

Houston, despite having the lowest number of unique domains overall, has the largest proportion of exclusive domains. These small differences suggest that while Google’s AI Overview source pool is largely standardized, some localization effects may be at play. We’ve seen that Houston provides a uniquely distinct set of sources in its search results.

Do AI Overviews cite exclusive domains (specific to each state)?

Each location in our study features a distinct set of exclusive domains that don’t appear in other states. These often represent local government sites, regional businesses, or local organizations.

Steven Schneider observes about this local vs. global relevance:

Steven Schneider

CEO at TrioSEO

I don’t think local vs. national or global matters as much, unless you’re searching for a local service provider. If I’m doing general research, I’d prefer a more authoritative source/citation than one who’s local.

Krista Seiden brings more focus to the value of localized insights:

Krista Seiden

Founder at KS Digital, Ex-Google, Marketing Exec

It’s important to reflect local results, so I do think AIOs should tailor their source citations to better reflect local expertise.

Below are the top exclusive domains by state.

Denver, Colorado

The top 3 exclusive domains in Denver, Colorado are:

Houston, Texas

The top 3 exclusive domains in Houston, Texas are:

Los Angeles, California

The top 3 exclusive domains in Los Angeles, California are:

New York City, New York

The top 3 exclusive domains in New York City, New York are:

Washington, DC

The top 3 exclusive domains in Washington, DC are:

Chris Raulf

International SEO and AI Expert

AIOs might predominantly rely on the same “big players” like Google, YouTube, Reddit, Quora, and Indeed, but they’re starting to sprinkle in distinctly local sources unique to particular cities and states—like metrodenver.org or denbar.org here in Colorado. As someone obsessed with localized SEO strategies, this is an important nuance. Google’s localized sourcing within AIOs, even if just emerging, suggests there’s a real opportunity for savvy marketers to boost their visibility by emphasizing geographic contextual relevance, structured micro content targeted clearly at common local queries, and strong domain authority building in their immediate region.

Kayle Larkin

Head of Marketing Elevar

Relevance isn’t just topical – it’s local. Local sources are closest to the ground and often offer the most accurate, timely info. AIOs should surface those insights, not just recycle national voices.

How many unique domains do AI Overviews cite per query by state?

The average number of unique domains per query is consistent across all five locations, ranging from 9.81 in New York and Washington, DC to 9.90 in Los Angeles.

The difference between locations is minimal, with no more than 0.09 domains per query. This indicates a high consistency in unique sources cited by Google’s AI Overview system (regardless of geographic location).

Are there observable differences in source diversity between each state?

When we analyzed how AI Overviews cite domains in each state, we found that for nearly half of all queries (47.05% or 14,470), the sources (domains) were completely identical in all locations. So the same domains were cited in every location. 

But the other half shows varying degrees of regional differentiation:

For almost half of all queries, AI Overviews select the same sets of sources regardless of where the search occurs. More than 53% of all queries have a domain match rate of at least 50%. This aligns with our earlier findings that showed global domains dominating in AIOs. For many queries, Google’s AI tends to rely on the same set of authoritative, globally recognized sources regardless of where the search is made.

The other half tells a different story. In many cases, the sources vary depending on the user’s location, likely reflecting a degree of local adaptation. The most interesting group is the 6.34% of queries where the sources don’t overlap between states. These likely involve highly local topics, like city services, small businesses, or government programs, where only locally relevant sources make sense.

If we look at the niches, this pattern becomes even clearer.

Paul Andre de Vera

SEO Consultant | Host and producer of the SEO Video Show

One of the key limitations I see in today’s AI Overviews is the underrepresentation of local expertise, even when the source is a local business. This is a missed opportunity for both the user and the business, especially in service-based industries where local recommendations play a critical role in user trust and decision-making. AIOs need to evolve to surface these local businesses directly in the answer, not just bury them in the source panel or leave them invisible.

Research methodology

For our study, we analyzed a total of 100,013 keywords across five states in the US: Colorado, Texas, California, New York, and Washington, DC. The keyword set covered various search intents, volume, and word counts to reflect a diverse spectrum of user queries. 

We applied a consistent methodology across all five locations to ensure comparable results. Our goal was to explore the behavior of Google’s AI Overviews across each of the five states, focusing on citation patterns, source diversity, and regional variations. We aimed to provide SEO specialists and digital marketers with actionable insights into the regional impact of AI-generated answers on SEO strategies.  

We examined 20 niches, analyzing around 5,000 search queries in each. Below are the niches included in our study:

We performed the data collection as a one-time check on April 15 with the following parameters: 

SE Ranking tools used for analysis: 

Wrapping up

The results of this research have made one thing clear: regional variation across the US in AI Overviews is minimal. But we do see some location-based differences for more local queries, where exclusive domains appear at the state level.

Chris Raulf

International SEO and AI Expert

These findings reinforce the importance of playing a smart, strategic local SEO game to appear prominently within AI snapshots. While Google’s rollout is clearly standardized at broader levels, we now have documented proof that localized nuances are starting to matter. This means smart marketers who lean into hyper-localization and carefully studied content can exploit this trend positively. To me, this trend makes Micro-SEO not just relevant today, but crucial in shaping SEO strategies for the future.

Paul Andre de Vera

SEO Consultant | Host and producer of the SEO Video Show

Your study highlights the highly standardized, templated nature of AIO responses across different states, despite varying local needs and contexts. In my work, it’s become clear that brands need to adapt beyond SEO fundamentals: ensuring content is fully crawlable in raw HTML, implementing structured data at scale, and actively maintaining their profiles in authoritative knowledge bases like Wikidata and Crunchbase and getting their brand mention on as many outlets as possible like press releases and citations. We’re at a pivotal moment where the SEO playbook is shifting from the art of SEO to the science of LLM optimization—and those who proactively test and adapt will be best positioned to thrive.

Kayle Larkin

Head of Marketing Elevar

SE Ranking’s research confirms AIOs are highly standardized. That’s a signal! If you’re cited once, you can scale. But the real value lies in spotting those edge cases where your local content does break through. That’s your wedge.

We also uncovered a few surprising shifts in Google’s behavior across the US region since our previous reports from February-March 2025 and November 2024. The most intriguing are Reddit and Quora returning as cited sources, featured snippets appearing far less frequently alongside AIOs, and Google increasingly linking to its own search results within AI Overviews. 

We hope these latest insights help you make informed SEO decisions in this fast-moving space. Keep an eye on our blog as we continue to explore AIOs from multiple angles to bring you the latest findings!