Chasing the AI Search Evolution
- Matt Lang
- May 24
- 9 min read
Setting the Stage for AI’s Search Takeover
AI-enabled information models, interfaces, chatbots and search experiences have exploded in recent years. With the proliferation of AI-optimized knowledge retrieval has come a realignment of the way people search and interact with information. This has created an AI layer – a new intermediary experience synthesizing answers to what people seek online. This new paradigm syphons attention and engagement from previously established entities and surfaces a new set of considerations for brands and marketers.
To put into context how rapidly this shift has happened, it’s worth reminding ourselves that ChatGPT was first made available for public use by OpenAI in November 2022. Even more recently, Google officially launched their AI Overviews experience only in May 2024. In less than three years we’ve seen a dramatic adjustment in how search functions and where people get information online.
In early 2024, Gartner predicted that traditional search engine volume would drop by 25% by 2026. While we’re not quite there yet, there are some strong signals suggesting it could be coming - most notably Google’s search volume dropped below 90% for the first time in a decade, almost certainly due to AI experiences eating into their usage share. So it seems the behavior shift is starting, but the landscape remains fragmented.
With a wide field of leading LLM players (at least eight significant general purpose ones by my count: ChatGPT, Perplexity, Gemini, Claude, Meta AI, Deepseek, Copilot, Grok) and AI experiences being distributed across search, social media platforms, and standalone LLM chat apps there is currently a gluttony of options vying for usage. Making things more competitive is the inter-category feature race developing with many players prominently launching new capabilities to capture whitespace (see: Google’s AI Mode, OpenAI’s SearchGPT functionality, or Perplexity’s shopping assistant or advertising beta for just a few examples).
AI Experiences Are Changing Search
With all these new experiences available, constant media speculation and proclamations of a new internet it’s important to dig into the true impact injecting all these AI powered experiences into the internet has had. So, here’s the story so far:
While the presence of AI Overviews in search varies drastically across queries, according to BrightEdge’s research some categories are seeing 50-80% of replies ‘answered’ by AI summaries. One hypothesis is that the categories seeing higher incidences of AI overviews are ones that lend themselves to more in-depth and informational responses (e.g., B2B Technology, Healthcare) vs. purchases or local queries (e.g., Ecommerce, Restaurants).
These AI Overviews, along with other AI experiences disrupting search, appear to be having a strong effect. Bain has found that there’s a recent surge in what they call ‘zero click’ search behavior where people search and get information without ever clicking a link provided. Further, they say “about 80% of search users rely on AI summaries at least 40% of the time.”
The apparent confluence of AI enhanced search’s arrival and peoples’ behavior shifting has already led to impacts in marketing programs and businesses overall.
Exhibit A is Seer Interactive’s recent interrogation into AI Search’s effect on CTRs for both paid and organic links. Their research found that “For queries where AIOs appear, organic CTR plummeted from 1.41% to 0.64% year-over-year. This highlights how AIOs can dominate user attention and limit opportunities for organic visibility. Paid CTR dropped across the board: Regardless of AIO presence, paid CTR saw significant declines year-over-year.”
Businesses are starting to notice differences in traffic, in some cases significant, leading to organizations like Chegg even pursuing a lawsuit against Google on the grounds that their AI initiatives in search have damaged their revenue.
Beyond the more quantifiable impact of search engine AI overviews and experiences on link traffic and paid search efforts, the broader LLM models are also growing rapidly and likely stealing share from traditional search. Activate Consulting’s recent study found that the ‘use of generative AI as a first stop for online search’ is growing at a 26% CAGR. Related to this, in potential good news for brands, BrightEdge research from Spring 2024 saw that “referrals from Perplexity to brand sites are growing at nearly 40% month over month” suggesting that LLM users are still finding their way to websites just not using traditional search to get there.
Meanwhile, retailers including Walmart, Amazon and Instacart are implementing GenAI Search quickly as well, hoping to advance consumer e-commerce experiences. Now, we are starting to see that basically any platform with a search bar is being enhanced with LLM and AI technology to help surface information and products.
Behaviors Starting to Embrace AI
When it comes to exactly how people are engaging with the GenAI, their search replacement behaviors give an idea. Tools are being used for research, content summarization, quick information retrieval requests, and some say they use it as a shopping support channel. Bain’s research says that “Even among those who say they are skeptical of generative AI, about half say that most of their queries are answered on the search page without a click.”
Returning to Activate Consulting’s report, we can get a better idea of search-based use cases and tasks that are likely to be, or already being, replaced by GenAI tools. Not surprisingly, information gathering and processing along with tasks like translation and summaries are being usurped by GenAI along with content generation tasks (like personalized planning) that traditional search is a poor fit to execute.
Shopping and facilitation of purchases however, remains an area that may take AI longer to become an accepted vehicle for consumers needs. As separate research from Bain last summer discovered, many consumers aren’t yet seeing the value for shopping with AI assistance and are hesitant to share financial information data that could improve the experience.
Thinking ‘AI-First’ for Search
Given where things stand today with behavior and impact on digital experiences, how should brands be thinking about making their content more ‘AI-First’? To start that discussion, we need to separate LLMs from AI Overviews and understand how each functions. LLMs are user driven, context and conversation-rich experiences vs. AI Overviews which intercept typical search behavior and seek to provide a concise, one-stop answer. Crystal Carter, Head of SEO Communications at Wix has summarized the distinctions nicely in her chart below
Beyond the usage and functionality differences, the replies coming from each type of AI experience differ significantly. An analysis from LLM visibility company Profound suggests there is low overlap, about 8-12%, in sources and answers between Google and ChatGPT. Additionally, the LLMs appear to put forth more ‘deep dive’ contextual content and sources vs. purchase links and direct brand sources more often cited by Google’s AI Overviews. To underscore the gap in consistent sources between the two, Profound says: “For brands looking to market in ChatGPT, being mentioned in Google/having a top-ranked product means nothing to ChatGPT.”
Understanding AI Overviews
BrightEdge has done industry-leading work analyzing what goes into an AI Overview and reverse-engineering what queries are most likely to be consolidated into one. Their research finds that queries returning Featured Snippets or that are question-based are more likely to be handled via AI Overview than other types giving content strategists and marketers a good starting point to consider.
Further deconstructing AI Overviews, revealed that, consistent with optimizing for other zero-click features such as Featured Snippets, AI Summaries like to see the following elements in website content:
Lists and Bullet Point Organization
Short and Clear Descriptions
Emphasis of Key Product Features or Article Points
Use of Markup for Important Perks (i.e. free delivery and stock status)
Leveraging Product Viewer Capabilities
Another under-discussed area to consider with Google’s AI Overviews is the seemingly growing influence of YouTube videos and content on their answers. Videos, particularly those focused on product walkthroughs and features, are increasingly appearing within AI Overviews. Transcriptions and other other copy linked to videos hosted on YouTube is also being referenced in AI Overviews. Perhaps most notably on this topic, testing is underway to potentially integrate AI Overviews directly into YouTube search which could create a new set of implications for video optimization and discovery.
There’s a lot to watch here related to the convergence of Alphabet’s many businesses around AI assisted search and overviews. Most recently, Google announced at their 2025 Marketing Live event that they are testing new ad products within both AI Mode and AI Overviews. Adding a monetization layer to the product will certainly create some interesting conversations in the paid search space.
As reactive marketers absorb the above information, there may be an understandable urgency to shift SEO strategies and start to position around AI and capture these AI Overviews. It’s not a bad instinct, but it’s important the strategy is put in place correctly.
In an SEMrush study of 500 keywords tested in October 2024, they found there was URL stability within the AI Overviews on just 4 of every 10 days and that, on average, Google changed the URLs of an AI Overview 12 times within the month.
“The data points to there being some serious concerns around URL placement within AIOs. With 91% of URLs falling out and just 43% returning, the AIOs we analyzed seem far from consistent. That could call the value of focusing on AIOs as part of your strategy into question.” – SEMrush
Given the current volatility of AI Overviews in Search, brands need to consider how valuable it is to try and ‘win’ certain queries. Instead, it may be more beneficial to think broadly about topic areas and tackle AI Optimization for several different related queries to build credibility and, hopefully, surface more consistently.
Understanding LLMs
Different LLM models provide different answers with different citations and varying levels of brand emphasis. It’s important to track if your pages are cited at all – this is essentially equivalent to ‘indexing’ in traditional search engines. Part of this comes down to differences in core search engines they leverage for responses. Depending on the model, optimizing for traditional SEO rank is likely to have an impact on being cited.
And let’s not forget that many AI companies have inked partnerships with specific publishers or data sources which may weigh heavier in some citation results. A solid overview of the current state of play is below from BuzzStream:
Another consideration is ensuring LLM web crawlers are being utilized to help flag site content to LLMs. Note: these instructions are used differently depending on the model either as general context or answer considerations.
If keeping track of crawlers, publisher relationships and search engine bases is making your head spin, you’re not alone. The good news is that most of best practice SEO for traditional web search is still registering as a positive for LLMs. Authority, credibility and content focused on user needs remain paramount. It is perhaps more important now that brands ensure their content comprehensively covers the journey customers take to find your them and their products.
Taking a look at brand visibility specifically, XFunnel AI did an analysis of 22,000 queries and found that brands and companies have strong citation presence through the funnel, increasing as users go further toward mid and lower funnel queries. Brands with a strong top of funnel presence also tended to benefit down-funnel. However, the top of funnel is less crowded with far less companies mentioned per response than mid-lower funnel queries (about 2-3 vs. 5-8)
Where We Are With “AI Optimization”
While I hesitate to use the term “AI Optimization” since we are so early in the game and much of “best practice” today overlaps with years of established SEO guidance, the prominence of the new AI experiences has created an eager audience wanting to learn how to improve their standing. Based on the many surveys, experiments, and analyses shared above, the following is what brands should be thinking about today to position themselves well going forward:
Monitor, Monitor, Monitor. Look to understand both what queries are getting replaced with AIOs and which pages of your site are being cited by LLM applications. Pay attention to sites and publishing partners that are being sourced more often.
Don’t Chase Individual AI Overview Keywords, but Strategize Around Topics. AI Overviews may be volatile and ‘winning’ a citation for a few days will only take you so far, but remaining credible and authoritative across a topic will continue to help content surface when relevant.
Update for AI. Where it makes sense, consider optimizing web content to be in “AI-friendly” formats such as lists, multimedia inclusions, short descriptions, and relevant markup (which are all modern SEO practices). Consider extending aspects of this approach to PR releases and other owned brand content that is likely to get cited.
Experiment with Crawlers and LLMs. Add web crawler instructions to your content and priority pages leveraging llms.txt files to see how various LLMs react.
Ensure Owned Product Content is Optimized. Making sure product information pages, and related vendor pages, are directly answering key questions across the purchase journey.
Comments