Let’s Put the “Oh!” in Google’s AI Overviews

Let’s state the obvious: search isn’t the same today as it was weeks ago. The emergence of AI search platforms and Google’s AI Overviews permeating informational, navigation, and even transactional searches disrupt organic traffic acquisition. There are, however, still benefits to earning a spot as a citation in Google’s AI Overviews, such as building brand awareness, ideas for improving content, understanding consumers’ follow-up needs, and generating topical relevance for your website.

This is purely positing, but I would bet that the systems that select what sources to use as citations from a corpus of documents will eventually impact organic ranking because of signals like consumer information gain and alignment with the general consensus of response – as of now, it’s evident that ranking in the top 5 organic positions does not have carry much weight, but we’ll get into that in a bit.

AI Overviews in Automotive: The Setup and Basic Findings

Using a client’s Google Search Console, I pulled more than 300 search queries and bucketed them into three basic intents: transactional, commercial, and informational.

Across several weeks, I conducted those searches at two-week intervals with a clean slate to identify any significant volatility in changes of what Google selected for the top three citations. I also wanted to identify shared traits, averages, and other distinguishing characteristics. I also needed to layer on engaging with the citations and/or top organic rankings to see if that behavior had any influence.

So, first, let’s answer this question: How often did AI Overviews populate the search queries?
29% of informational queries
29% of commercial queries
3% of transactional queries

Did the AI Overview include a dealership as a top 3 citation?
88% of informational queries
100% of commercial queries
100% of transactional queries

Was the dealership local to me?
0% of informational queries
33% of commercial queries
100% of transactional queries

This indicates, to me, that there are ample content opportunities not just for dealerships to earn spots as a citation across all search intents – and it’s very likely that content programs implemented by agencies and internal marketing teams are foregoing informational content either because (a) the doesn’t “drive local traffic,” (b) they do not know how to build a strategic cluster of topically related content, or (c) the agency has sold an empty bill of goods on “content at scale” leveraging “internal AI.”

Interesting AI Overview Change Based on Behavior

I would be remiss if I didn’t share this one finding on the impact of user behavior.

The search query “cars for lease” would trigger an AI Overview when you have a clean slate – cleared history, cache, et cetera; however, once I engaged with a citation or an organic link and conducted the exact search again a few days later, it would no longer populate an AI Overview but a standard search results page with a local pack and blue links.

This indicates that Google’s systems recognize the need to change the results page based on user engagement/behavior.

AI Overview Citation Averages and Shared Traits
Commercial (mid-funnel) Findings

66% of the top 3 citations had an exact or near-match to the keyword phrase in the title tag, the H1, and the content itself – yes, it was the same percentage across those three features.

For the content aspect of tracking keyword usage, the exact or near-exact use took the primary, core keywords and only counted them if they were in the same sentence and relative proximity to each other: very rarely did that exceed five instances.

The AI Overview had an average of 6 sections of response that contained the link icon to specific citations.

The average word count across the top 3 citations was 1,458 words.

An average of 4 images was used across the top 3 citations in the cited content.

Informational (high-funnel) Findings

50% of the top 3 citations had an exact or near-match to the keyword phrase in the title tag, the H1, and the content itself – yes, again, it was the same percentage across those three features.

Again, for the content aspect of tracking keyword usage, the exact or near-exact use took the primary, core keywords and only counted them if they were in the same sentence and relative proximity to each other: only once did it exceed five instances.

The AI Overview had an average of 7 sections of response that contained the link icon to specific citations.

The average word count across the top 3 citations was 604 words.

An average of 3 images was used across the top 3 citations in the cited content.

So Does This Indicate Anything About AI Overviews?

I believe this shows that there’s a significant relation to a consumer’s “information gain.” This is a tangential connection here, but I’m going to make it anyway. In Google’s Information Gain patent, there’s a portion about how it ranks a related topic’s next set of search results that reads as follows:

“An information gain score for a given document is indicative of additional information that is included in the given document beyond information contained in other documents that were already presented to the user.”

“An information gain score for a given document is indicative of additional information that is included in the given document beyond information contained in other documents that were already presented to the user.”

I believe that a version of this is used to generate the full AI Overview that provides additional context beyond the most basic, essential answer. It is also why there is not actual correlation between the top 3 citations and the number of times they’re used in all of the sections of the Overview with linked citations.

The entire AI Overview is the answer and expected next set of search results.

In looking at the automotive query, the top 3 citations were used in as many as 8 out of 9 in-text sections and as few as 2 out of 10. It all depends on the context within the content and how much content is directly related information gain.

I also believe that this can hopefully be the final nail in the coffin about “keyword density” in content. There were some instances where none of the top 3 citations had an exact or near-match use of the keyword phrase in the title tag, H1, or content but rather used semantically related terms, incorporating them naturally into the content.

It appears that the AI Overviews can parse out the keywords and related terms that comprise the entirety of the query phrase, so you don’t have to obsess over writing the query phrase verbatim in your content – so don’t neglect the keywords but focus more on fully and naturally addressing the consumer’s intent (and potential follow-up interests or questions).

I aslo hope this can put to rest the notion that still persists in SEO about word counts. I did see that commercial terms in automotive were more precisely written nd often required more substance (hence the average of 1,458 words versus informational queries at 604 words).

Informational queries hadd a much greater variance of word count, ranging from 104 words up to 1,083 words – the consumer’s intent and complexity of the query dictated what was necessary to provide a sufficient answer.

Other Google AI Overview Findings
General Consensus
There is a position in SEO that general consensus – the idea that documents in a corpus affirm, negate, or are neutral toward an answer – impacts ranking. This had traditionally been applied to the organic blue links and not in the context of Google’s AI Overviews. I did, however, see this in action with the selection of citations for the search query “safest ford suv.”

The content that ranked first organically, which was an actual local dealership – not just a small business far, far away – was not used as a citation because the context of the content did not align with the general consensus of the answer.

The general consensus is that the Ford Explorer is the safest Ford SUV; whereas the content ranked first organically had content that spoke to all of the the safety feature and ratings of all Ford models.

Now, to be fair, this is also an issue because it isn’t fully aligned with the search intent as “safest Ford SUV” is explicit that the person wants to learn about which of Ford’s SUVs is the safest of all, not just how safe are Ford models.

Impact of Search Intent
It should come as no surprise that search intent really dictates whether or not an AI Overview is necessary – not just based on search behaviors and engagements but also initial search intent. And when the intent has an overt connection to imagery and media, then the results align with that. Take the case of “ford edge interior” prioritizing images and then, not seen in the screenshot below, pages of content that are more media rich.

People don’t just want you to write about cargo space or dimensions. They want to see the interior of the vehicle.

There was another interesting tidbit pulled from “near me” searches, which wasn’t part of the keyword set I originally pulled in; however, after seeing some social media posts around this type of search and the results return, I did find something of note.

For “[brand] dealership near me”, an AIO populates — BUT only when there is one option for that brand in the local market. The AIO provides more information about other relatively close dealerships from which to choose; whereas, there is no AIO when the local market has multiple options for the brand.

In this particular instance, the dealerships in the citations are correctly collated by distance to where I am at.

In addition to this, I also believe there may be a cutoff on mileage at which Google determines to not populate an AI Overview.

This assumption is derived from the following:
There is one Audi dealership right near where I live. The next closest Audi dealership is in Mobile, Alabama, which is 68 miles away, coming out to be around 1 hour and 5 minutes of travel time. For “audi dealership near me” searches, no AI Overview populates.

There is one Honda dealership right near where I live. There are two more listed in that AI Overview that does populate: Fort Walton, which is over an hour of travel time but only 40 miles away; Daphne, Alabama, which is 47.5 miles and 46 minutes away – and the AIO prioritized them based on mileage (even though it technically takes more time to Fort Walton than Daphne).

This is important for dealerships who may be in a much lighter market and want to target a neighboring, larger metropolitan area, especially if there are few same brands in that other metro because there may be opportunities via AIOs to get visibility where it was harder organically.
Impact of Organic Ranking
It’s essentially the wild, wild west with correlations between ranking in the top 5 positions and showing as one of the top 3 citations – or a citation at all. What can be discerned is that at least one of the citations did organically rank in the top five …
88% of the time for informational searches with an average of 1 out of 3 citations.
100% of the time for commercial searches with an average of 2 out of 3 citations.

Because of the such small return of AIOs, I don’t want to try to average out transactional searches.

In the cases of informational and commercial searches, even though the percentages are high, it was a mish-mash of which of the three citations were also ranking and in which position they ranked organically. There was no real connection or correlation.
Schema Markup

Only 46% of all citations had some form of schema, and the vast majority of it was either Article or FAQ schema – as a preview of the next piece I’ll be writing about AI Overviews, the legal industry, so far, is coming in hot with 83% of all citations having a form of schema with most being Article or FAQ followed by Author/Person.

With that being written, that’s not to say that had another page of content implemented schema that they would not have then been used as a citation; nor is it to say that removing schema from that 46% would drastically change the citations. I’m just seeing that, at least for the automotive industry, schema markup can be a great strategic tactic layered onto other efforts in buidling content but isn’t necesssary to be used as an AI Overview citation.

AI Overviews’ Impact on Clickthrough Rate
In looking at the automotive queries that I reviewed, I wanted to ensure a few things to keep things fair in assessing the impact on clickthrough rate:
If there was a drop in CTR, there was not a worsening average position because that would play a role and obfuscate the data.
The query had to already be driving enoguh clicks to at least look at a 3-month vs. preceding 3-month period – and some queries were based on brand new content driving that visibility

For the queries that met those two criteria and triggered an AIO, there was an average drop of 7..2% CTR.

In one case, despite the website moving up 9 organic positions, the CTR dropped from 27.3% to 5.3%.

So What Now?
Just because “there’s a chance” to get a click, the obvious point of trying to understand Google’s AIOs, especially as it penetrates more transactional searches, is to build brand awareness and establish the site as a hub of valuable, accurate information around topics most closely associated with your core business.

Here’s where I think all agencies and internal teams need to focus.

Search Intent
I think in the rush to produce more and more and more, a lot of practitioners are losing track of the precise intent of the consumer’s search. It’s clear that not everything generates an AI Overview, so understand what does and what doesn’t – and what is prioritized in the search results for that intent.

Information Gain
It’s time to start leveraging all sorts of tools, both free and paid: SEMRush, Ahrefs, Answer the Public, People Also Ask, Google Search Console, and the actual additional citations in the AI Overview. Understanding the next steps – the next question, next need, next interest – that are closely related to the original core intent can help you do one of two things: (1) build out the content even further to make it as encompassing but still highly relevant to the search intent to be a citation in more of the in-text sources of teh overview; (2) understand the next piece of content you may need to build if it’s only on the periphery of the intent and possible leading to a whole new set of questions and information gain.

Context, Structure, and Clarity
Stop writing to fill the page with words and get the context correct so you satisfy the searcher. Structure the content and really consider how it can naturally flow among all of the semantically and topically related queries that comprise a whole piece of content.Then, as you write or edit the content, ensure clarity ifn your language, making it read as naturally and conversationally as possible.

And stop worrying about working in the exact keywords in every element.

Your context, structure, and clarity need to be focused on people.

Metrics
It’s 2025. Yes, keyword ranking should be considered as one of a multitude of signals in performance health; however, this should not be a KPI often discussed. That’s because 15% of searches every day are brand new, and the way many ways taht people search for any product or service varies so greatly.

To steal a line from Robin Werling here at SearchLab, “I go to the salon. My dad goes to the barber. My kids go get a haircut.” – or something close to that. Sorry, Robin, if I totally got it wrong!

The point is that there are so many variants that what you want is more visibility and more clicks.

You’ll also want to track direct and branded traffic with the caveat that you have to bear in mind that marketing isn’t just SEO. There are a confluence of factors that can influence that and that brand awareness in AI Overviews is definitively not the only potential factor; however, as you spend time to work on acquiring a spot as a citation in Overviews, you can track over longer periods of time that growth with branded and direct traffic – is there a potential correlation and analyze that against other marketing and branding efforts.

You’ll want to look at organic as a share of conversions. After all, the ultimate goal is to drive more leads for the dealership. So, has the strategy been working so that your organic traffic is the strongest and largest driver of conversions? All of the other efforts should be strategized to ultimately provide equity and, thus, visibility to all of those low-funnel, purchase intent pages on your website.

With privacy controls, obfuscation or blocking of data, including location data, disruptors like AI Overviews and AI platforms, the simple tracking of just traffic won’t cut it – we all need to dig deeper into the metrics that really matter.

And keep in mind that this all will continue to change over time.

Questions?

If you have questions or would like SearchLab Digital to assess the health of your website’s foundational optimizations and content strategy, as well as strategy for generative search results, please don’t hesitate to reach out to me at [email protected].