AI-driven search has sparked panic amongst marketers, with a desperate dash to unveil the “secrets” of generative engine optimization (GEO). But anybody with real SEO experience knows there is never a single definitive “trick” to nail SEO—and the same is true when it comes to LLM SEO.
This article offers a level-headed guide to GEO and AI-driven search, helping you think clearly about a subject that will define the next decade of online marketing.
Using the latest data and our own ongoing experimentation, we explore what currently works and the strategies you need to future-proof your online visibility.
AI-driven search refers to tools that use large language models (LLMs) to synthesize comprehensive responses to search queries. This takes two basic forms:
AIOs are comprehensive summaries that sit above the fold on Google and Bing searches.
While slightly less than 13% of Google searches currently produce an AIO, most experts expect that number to rise rapidly over the coming months and years.
Importantly, 13% of all Google searches translates into an enormous amount of eyeballs.
Research suggests over 1.5 billion people globally have been served an AIO—equating to over a quarter of all internet users.1
While AIOs include “blue links”—various sources for the summary—these are less likely to drive traffic to your website for a few reasons:
AI platforms like ChatGPT and Perplexity that function as chatbots: users input a prompt and receive a synthetic textual output.
These tools can function as a personal writing assistant; zero-budget therapist; or sounding board for decisions and creative ideation. However, their most popular use is as a search engine.
Adobe reports that 77% of ChatGPT users treat it like a search engine, with 30% trusting it more than traditional platforms like Google.2
This has led to widespread panic amongst marketers that traditional search will become obsolete and existing SEO value will be lost.
But is that realistic—or simply industry talk?
Organic search rankings and website traffic across multiple industries—including financial services—are considered under threat from LLMs.3 The fear that AIOs and Chatbots will steal organic traffic and reduce online visibility for advisory firms has been widely stoked.
Early reports suggested 25% of organic search traffic would be lost to LLMs by 20264; today, some argue up to 87% of online search may use LLMs by 20295—marking an almost complete inversion of the existing search landscape.
This has led plenty of agencies to rapidly reposition themselves as an “LLM SEO company,” trying to cash in on the industry panic.
But these efforts are not just cynical; they overlook the fact that existing data on LLM SEO is highly conflicted.
Just take “Zero-click” search: many warned that AIOs would be the end of organic traffic. This was supported by reports that specific publishers had seen organic search click-through rates (CTRs) fall off.
This was true, but later analyses revealed that zero-click behavior has not trended in a straight line. While zero-click behavior increased after the introduction of AIOs, it actually peaked in January 2025—before dropping again over subsequent months.6
Since then, reports have continued to circulate that specific businesses have seen traffic dip as much as 89%.7
But these stories are not universal; the impact of LLMs on search traffic and marketing effectiveness will be different for every business—and largely determined by how they respond over the coming months and years.
The key to making sense of LLM search—and what you should do about it—is to look at how it has influenced search behavior.
The shift from traditional search to LLM-driven search has sparked a striking evolution in search behavior. Three factors stand out:
The average query on a traditional search engine ranges from 2 to 6 words long, with most sitting at 3 to 4 words. Most users find that Google and Bing do not reward much more detail beyond this, unless they are searching for something highly specific.
However, LLMs actively thrive on longer, more detailed searches. While Google delivers the same basic service for every query—a list of potentially relevant websites—Chatbots can adapt the kind of response they give based on the search intent.
As a result, longer queries give more context and help Chatbots calibrate the length, level of detail, topic, and even format of their response.
Users quickly noticed this, and the data reveals that fact: search terms with 7-8 words in them have nearly doubled across all search engines since the launch of ChatGPT.8 We predict this trend will continue, and the length and type of queries will evolve along with the technology.
For example, the rise of speech-enabled Chatbots means more users now speak their prompts. This will not only naturally lead to longer searches, but more intent-rich searches, as people “brain dump” rather than searching for specific terms or asking short, clear questions.
Traditional search journeys are discrete events.
An individual might search for information about retirement savings; then look up their 401(k)s options; then look for advice about selecting a 401(k). These are all separate searches that Google simply responds to as if they are unrelated.
The same journey using ChatGPT or Perplexity would function like an ongoing conversation. For example, the interaction might include:
Importantly, this is true even if the user doesn’t make all of their searches in a single session. Most Chatbot users allow the platform to “remember” information about them and slowly learn about their interests and preferences.
The takeaway from a marketing perspective is simple: ranking in LLMs creates far more opportunities to be included in highly contextually relevant search journeys. While Google acts as a kind of directory, Chatbots are closer to “word of mouth” marketing—notoriously one of the most powerful and trusted forms of promotion.
As discussed above, the rise of LLMs has caused an increase in “zero-click” search, where users get everything they want from a search response without the need to click on a website.
While not universal (or as clear-cut as many commentators claim), this is a challenge for marketers. The problem is particularly pertinent on Google, where despite concerns about accuracy, just 8% of users click through to check the source material for search summaries.9
So search traffic has taken a hit since Chatbots and AIOs arrived. Bain Co. reports roughly 80% of searchers now rely on AIOs around 40% of the time.10
Another issue is that Chatbots do not always even contain sources—let alone driving the average user to click on them.
Some commentators claim this is a disaster for marketers, with reports that it has led to 15-25% less organic traffic. 11
But it’s important to remember that organic traffic does not equate to commercial results. LLM search may drive less traffic, but the traffic it does generate is far higher-intent.
This means SEO and GEO differ in two ways:
Changing Search Behavior: An Expert Overview |
||
|
Behavior Category |
Traditional Search |
LLM-Driven Search |
| Query Length | 2–6 words on average; shorter, keyword-based queries perform best | 7–12+ words; thrives on detailed, context-rich prompts and full sentences |
| Search Type | Transactional and task-based; each search is independent | Conversational and continuous; builds on previous context across multiple prompts |
| Response Type | Static list of ranked websites | Dynamic, adaptive answers tailored to intent, tone, and format |
| User Interaction | One-way: user searches, engine responds | Two-way: chatbot asks follow-ups, refines answers, and remembers preferences |
| Marketing Implication | Focus on keyword rankings and traffic generation | Focus on context inclusion, brand authority, and conversational visibility |
Generative engine optimization (GEO) is the emerging set of practices designed to help companies appear in AIOs and Chatbot responses. It differs from traditional SEO in a few important ways:
SEO focuses on whole pages, whereas LLMs leverage individual chunks of information.
GEO can apply to pages that have little chance of getting to page one of Google, as Chatbots and AIOs can easily pull small chunks of high-quality information from pages that otherwise do not meet search intent.
This simultaneously opens the playing field and creates more complexity for marketers. You cannot simply focus on a handful of central pages; every page is a potential GEO play and should be optimized to be valuable to LLMs.
The longer, more detailed searches LLMs encourage lead to a richer keyword universe.
GEO efforts should focus less on short, chunky keywords and more on answering a wide range of nuanced questions related to your services and industry.
For example, where SEO strategies might underplay industry jargon in favor of the terms users search for within Google, LLMs will see complex terminology as a signal of authority and depth of content.
While relatively few Google users would voluntarily search for the definition of “contango” (a market condition in the futures market, if you’re asking), plenty will ask ChatGPT to clarify what it means during a relevant search.
The buyer journey is harder to track and less linear when using AIOs and Chatbots.
But there is still clear value to GEO from a marketing perspective. In fact, recent research suggests the volume of revenue generated by LLMs will increase by 75% over the coming years.12
GEO drives significant commercial benefits through:
However, these differences should not cloud the basic fact that GEO and SEO are very similar beasts. Rather than a transformational break, GEO represents a new phase of SEO—one that will carry forward many features already well-familiar to SEO experts.
The reality is that GEO and SEO share several foundational requirements:
This is demonstrated by the fact that roughly half of all pages that appear on page one of Google also regularly appear in Chatbot responses.14 And that means we can learn a lot about the future of GEO from the history of SEO.
Because SEO is such a well-established part of modern digital marketing, it’s easy to forget it was once an exciting, emerging practice—and many brands took a long time to accept its importance.
Consider this (heavily condensed) timeline:
|
Year / Period |
Milestone / Change |
Significance / Impact |
|
Early 1990s |
First web search engines (e.g. Archie, WebCrawler) and directories |
Data was manually indexed; content discovery was primitive |
|
Mid 1990s |
SEO begins (even before the term existed) |
Webmasters start manipulating meta tags, keyword stuffing, and directory submissions |
|
1998 |
Launch of Google and introduction of PageRank |
Links become a key ranking signal—SEO shifts beyond just on-page keywords |
|
Early 2000s |
“Link Era,” proliferation of link schemes & directory spam |
SEOs exploit link networks, reciprocal linking, and exact-match domains |
|
2003 – Florida Update |
Google’s first major update to penalize over-optimization |
Marks shift toward algorithmic policing of spammy SEO techniques |
|
2006 |
Launch of Google Analytics & Webmaster Tools |
Provides SEO practitioners with real data on indexing, traffic, crawl errors |
|
2011 |
Schema.org / structured data introduced |
Helps search engines interpret and display rich results |
|
2013 |
Google Hummingbird update |
Moves Google toward natural language, conversation, and context over isolated keywords |
|
2014 |
Google Pigeon update (local search emphasis) |
Signals that location and proximity become influential in SERPs |
|
~2015–2020 |
Rise of content clusters, mobile-first, E-E-A-T & quality signals |
SEO begins to emphasize holistic content strategy, user experience, authority |
|
2021 (Oct) |
Launch of IndexNow protocol |
Enables websites to proactively notify search engines when content changes, speeding indexing cycles |
|
2023 – 2025 |
Growth of generative / AI-driven search & “Generative Engine Optimization (GEO)” concept |
SEO must adapt to being surfaced in AI responses, not just traditional SERPs |
GEO will likely evolve just as much—if not more—than SEO. Just as Google’s regular algorithm updates threw the entire industry into chaos multiple times, we should expect future LLM models and emerging competitors to shake up GEO and force marketers to adapt.
Research from Princeton finds that the right set of GEO practices can increase visibility in AIOs and Chatbots by up to 40%.15
Based on extensive research and existing SEO expertise, our team recommends several clear steps to improve your rankings in generative engines:
AI engines prioritize content they can verify and cite with confidence. Your goal is to become a "machine-trusted source."
How you organize information matters as much as what you say. AI models parse content differently than human readers—optimize for both.
AI systems evaluate trust before featuring content. Strengthen your E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signals across your site.
Local and industry-specific visibility depends on consistent, well-maintained profiles across the web.
Off-site signals remain essential for both traditional SEO and GEO. Quality matters more than ever.
ProperExpression is monitoring the evolving GEO landscape—and building repeatable processes to help wealth managers adapt. With both SEO and GEO, we can help your firm future-proof its digital presence and win more clients.
Want to book 15-minutes to discuss your biggest GEO easy-wins?
1. https://ahrefs.com/blog/insights-from-56-million-ai-overviews/
3. https://www.semrush.com/blog/semrush-ai-overviews-study/
5. https://explodingtopics.com/blog/llm-search
6. https://www.semrush.com/blog/semrush-ai-overviews-study/#
8. https://searchengineland.com/keyword-query-length-insights-445376
9. https://explodingtopics.com/blog/ai-trust-gap-research
12. https://explodingtopics.com/blog/llm-search
13. https://ppc.land/google-ai-search-features-drive-click-rates-to-historic-lows/
14. https://www.semrush.com/blog/is-seo-dead/#