The Evolution of Search: Why AI-Driven Optimization Is Still Just SEO

Critical Market Signals:

• Query Fan-Out Technology: Google’s AI systems now execute multiple parallel searches beyond the user’s original query, fundamentally changing how content gets discovered and cited in AI Overviews.
• Multimodal Search Dominance: Users increasingly search with video and images while expecting textual responses — creating new opportunities for content-rich publishers who integrate multiple formats.
• Quality Conversion Shift: Traffic from AI-generated answer formats demonstrates higher engagement rates and time-on-site metrics compared to traditional blue-link clicks, signaling a fundamental change in user intent quality.

The search industry faces a nomenclature crisis. GEO (Generative Engine Optimization), AEO (Answer Engine Optimization), LMEO (Language Model Engine Optimization) — practitioners scramble to brand the “next evolution” of search optimization. Yet according to Danny Sullivan, Google’s Search Liaison, this proliferation of acronyms obscures a fundamental truth: the core principles haven’t changed. What has changed is the sophistication of systems interpreting those principles.

Sullivan’s analysis, delivered in collaboration with Google Search Advocate John Mueller, dismantles the premise that AI-driven search formats demand revolutionary new tactics. Instead, they argue that Google’s ranking systems — whether serving traditional blue links or AI Overviews — optimize toward a single north star: rewarding content created for human satisfaction, not algorithmic manipulation. The implications extend beyond semantic debates. Organizations investing in format-specific optimization strategies may be optimizing for transient system behaviors rather than the underlying reward mechanisms that persist across Google’s evolving interface.

The Umbrella Framework: Why SEO Encompasses All Format Optimization

Sullivan reframes the relationship between traditional SEO and emerging optimization practices through a hierarchical model. SEO remains the umbrella discipline — defined as “the practice of improving content for search engines” — while format-specific approaches like AEO or GEO function as subsets addressing particular interface types. This mirrors established patterns: local SEO focuses on geographic listings without constituting an entirely separate discipline; voice search optimization addresses specific query structures without abandoning core SEO principles.

The architectural logic becomes clear when examining Google’s system design. Sullivan explains that AI Overviews, conversational search modes, and traditional result pages all draw from the same underlying ranking infrastructure. The systems evaluate content quality, relevance, and user satisfaction through consistent signals — whether that content surfaces in a featured snippet, an AI-generated summary, or position three in organic results. Format presentation changes; evaluation criteria remain stable.

This has immediate strategic implications. Organizations maintaining separate “SEO teams” and “AEO teams” introduce unnecessary complexity. The skills required — understanding user intent, creating authoritative content, ensuring technical accessibility — apply universally across formats. Sullivan notes that practitioners who “focus on your content and not really worry about this” often achieve better long-term outcomes than those chasing format-specific tricks.

Strategic Bottom Line: Consolidate optimization efforts under a unified SEO framework that addresses format variations as tactical considerations, not strategic pivots.

Query Fan-Out: The Mechanism Behind AI Overview Selection

Google’s query fan-out technology represents the most significant architectural shift in how content gets evaluated for AI-generated answers. When a user submits a query, the system doesn’t simply match that exact phrase against indexed content. Instead, it automatically generates and executes multiple related queries — the incremental refinement searches users would traditionally perform manually over several minutes.

Mueller illustrates the mechanism: “It does a whole bunch of searches for you. So it’s kind of in a way what you were describing before like all of those small incremental searches that you could have done it does for you and based on the results that it finds it puts together an AI answer.” This explains why content ranking well for the user’s literal query may not appear in the AI Overview — the system evaluated dozens of related queries the user never typed.

The practical consequence: traditional keyword optimization targeting a single phrase becomes insufficient. Content must demonstrate topical authority across the semantic cluster surrounding that phrase. Sullivan provides context: “You’ve getting a lot more context. So you are ending up with the thing that is probably closer to what you wanted in the first place.” Users arriving from AI Overviews exhibit higher engagement because the system pre-qualified their intent through multi-query validation.

This creates a measurement challenge. Sullivan acknowledges: “People be like, I did this search and I wasn’t in I’m in the blue links, but I’m not in the AIO. And it’s like, yeah, dude, the search is what the person searched for, but we went beyond that with the AIO’s when all those, as you said, those sort of incremental things.” Publishers tracking rankings for specific keywords miss the broader set of queries their content must satisfy to earn AI Overview citations.

Strategic Bottom Line: Shift from keyword-centric optimization to comprehensive topical coverage that addresses the full semantic cluster around core topics, ensuring content satisfies both the explicit query and the implicit information needs the system anticipates.

The Commodity Content Threshold: Where AI Answers Displace Publishers

Sullivan identifies a critical inflection point in content strategy: the commodity information threshold. Factual, non-original information — the kind that requires no unique perspective or firsthand experience — increasingly gets answered directly by AI systems without generating clicks to source websites. His canonical example: annual “What time is the Super Bowl?” content.

The pattern Sullivan describes became standard practice during the SEO content boom: “Multiple places would then all write there, what time does the Super Bowl start in 2011 post? And then they would write these giant long things. And this in the history of Super Bowls before the the dinosaurs roamed on Earth, people wondered about time. What is the concept of time? How does it take us across the universe? Well, and then like and the Super Bowl will be at 3:30 p.m. Eastern.”

Google’s systems evolved to extract the factual answer — 3:30 p.m. Eastern — and present it directly, eliminating the need for users to visit any of those elaborately-constructed pages. Sullivan notes: “Most people, I think the vast majority of people say that’s a good thing. Thank you for telling me the time of the Super Bowl. It it wasn’t super original information.”

The strategic vulnerability extends beyond Super Bowl times. Sullivan warns about sites generating “a huge amount of traffic for the answer to various popular online word-solving games. Just every day I’m going to give you the answer to it.” These publishers face existential risk: “That is great until you know the system shift or whatever and it’s common enough or we’re pulling it from a feed or whatever and now it’s like here’s the answer. But that wasn’t really your strength as a publication or as a site or whatever. That wasn’t your original voice.”

Strategic Bottom Line: Audit content portfolios to identify commodity information assets that provide no unique perspective or firsthand insight — these represent declining value in an AI-answer environment and should be deprioritized in favor of original analysis and experiential content.

Original Voice as Competitive Moat: The Authentication Imperative

Sullivan positions original voice and authentic perspective as the primary defensible assets in AI-mediated search. “Your original voice is that thing that only you can provide. It’s your particular take,” he emphasizes. This isn’t aspirational advice — it’s a structural observation about what AI systems cannot commoditize: firsthand experience and unique analytical frameworks.

The authentication imperative extends beyond written content. Sullivan observes increasing user demand for experiential content — videos, podcasts, and social posts delivering firsthand perspectives. Google’s search systems have responded by integrating “more social more experiential content. Not to take away from the expert takes, it’s just that people want that sometimes like you’re just wanting to know someone’s firsthand experience alongside some some expert take on it as well.”

Sullivan draws a direct parallel to social media success patterns: “A lot of people who are on social media, the ones I especially like, they just have something great that they want to share or they they have a passion about it and they didn’t sit down and think how am I going to tweak it so that this is super great for I don’t know Instagram but not for Snap and now for you know whatever.” Authenticity precedes optimization. The core content resonates because it reflects genuine expertise or experience, not because it was engineered for a specific platform’s algorithm.

This creates a strategic inflection for publishers who built content operations around keyword targeting and search volume data. Sullivan challenges: “If you are providing those expert takes, you know, you’re you’re doing reviews or whatever and you’ve done that in the written form, you still have the opportunity to be doing those in videos and podcasts and so on. Those are other opportunities.” The format expansion isn’t optional — it’s how publishers demonstrate the depth of expertise that AI systems cannot replicate from aggregated data.

Strategic Bottom Line: Restructure content operations to prioritize firsthand expertise and unique analytical frameworks over keyword-targeted informational content, and expand into video and audio formats to demonstrate authentic subject matter authority.

Multimodal Search Architecture: Cross-Format Content Requirements

Sullivan introduces multimodal search — which he admits “I hate the term multimodal. I just hate it. It says nothing” — as a critical optimization consideration. The concept: users search in one format (video, image, voice) and receive responses in another (text, video, structured data). Google’s systems must match queries across format boundaries, creating new discovery opportunities for publishers who provide content-rich assets.

Sullivan demonstrates through personal example: “I was walking around in Portland and some killing some time before I did a talk out there and I saw these geese on the ground. Not in the air but they were on the ground. They’re all over the place and they were eating something or doing something or poking at the ground. So, I did a video of them and then I sent it off to Google conveniently enough because we can do that. You can send a video through the app and I’m like, ‘What are they doing?’ And it came back and it said, ‘They’re eating.'”

The interaction required no textual query formulation — Sullivan simply captured video and asked a natural language question. The system analyzed the visual content, understood the context, and provided a textual response. This pattern extends to image-based searches for product identification, troubleshooting (“I couldn’t even be bothered to type in the name of it. I’m just like what is it? How do I fix it?”), and valuation queries.

The technical implication: Google’s systems increasingly match multimodal queries against content that includes images, video, and text. Publishers who remain text-only lose eligibility for an expanding share of search traffic. Sullivan frames this as opportunity rather than mandate: “If you have been more textual in nature and you haven’t been doing images or video, then by maybe making sure you have more contentrich original content, you potentially are going to have opportunities in some of these multimodal search experiences.”

Strategic Bottom Line: Implement multimodal content strategies that integrate text, images, and video for core topics, ensuring content remains eligible for discovery across query format types and AI-generated answer experiences.

Structured Data as Contextual Signal: The Interpretation Layer

Sullivan addresses structured data with notable restraint, positioning it as a helpful signal rather than a deterministic ranking factor. “It wasn’t like if you didn’t have structured data, that’s it. You’re done for,” he clarifies. The role: structured data provides explicit semantic markers that help Google’s systems understand content relationships and entity attributes, particularly valuable in AI Overview generation.

The strategic nuance: structured data functions as an interpretation layer that reduces ambiguity for algorithmic systems. When content includes schema markup identifying the author, publication date, article type, and key entities discussed, Google’s query fan-out systems can more accurately match that content to related queries beyond the page’s primary keyword target.

Sullivan’s guidance emphasizes proportionality: structured data represents “one of those things like when we were all putting our heads together, it was like, yeah, let’s let’s emphasize that as well because that’s that’s something people can focus on that will help them in both of these kinds of areas.” It’s a technical optimization with cross-format benefits, not a silver bullet for AI Overview inclusion.

The implementation priority: focus structured data efforts on core content assets — comprehensive guides, original research, and high-authority pages — rather than attempting to markup every page on the site. The marginal benefit of structured data increases with content quality; it cannot compensate for thin or commodity information.

Strategic Bottom Line: Implement structured data as a complementary optimization for high-value content assets, prioritizing schema types that clarify entity relationships and content authority signals relevant to your industry vertical.

Quality Conversion Metrics: Redefining Success Beyond Click Volume

Sullivan challenges publishers to redefine success metrics in an environment where AI Overviews and direct answers reduce total click volume but potentially increase per-click value. “You probably want to start measuring quality clicks and quality conversions,” he advises. The shift reflects a fundamental change in user behavior: visitors arriving from AI-generated answers demonstrate higher contextual awareness and clearer intent than those clicking traditional blue links.

The mechanism: AI Overviews function as pre-qualification filters. Users who read a comprehensive AI-generated answer and still choose to click through have confirmed their need for deeper information, alternative perspectives, or transactional action. Sullivan explains: “We think that these formats are putting people into a better state of contextual awareness. We think people understand better what they’re getting into after they read one of these things.”

Google’s internal data supports the engagement hypothesis: “We can understand the time of of visits or whatever and so we can understand they’re spending more time on the sites which we think is a good proxy for people being more engaged.” This aligns with anecdotal reports from publishers who observe that AI Overview traffic converts at higher rates despite representing lower absolute volume.

Sullivan frames the measurement challenge: “Maybe you’re not even thinking about how you do the conversions, right?” He describes his own publishing strategy: “When I used to run a site or two, I like my number one goal was always to make sure that if you arrived there, you left your email because I wanted to then have this continuing connection.” The conversion definition varies by business model, but the principle remains constant: optimize for visitor quality, not just visitor quantity.

The analytics implementation requires tracking beyond pageviews and session counts. Publishers need visibility into time on site, scroll depth, interaction with conversion elements, and downstream behavior — metrics that indicate genuine engagement rather than accidental clicks. Sullivan acknowledges the complexity: “There are different analytic tools that can let you do that. If if you want to go into that deep they’re way beyond the kind of stuff I deal with.”

Strategic Bottom Line: Implement quality-focused conversion tracking that measures visitor engagement depth and downstream actions rather than optimizing solely for click volume, recognizing that AI-mediated traffic may deliver lower volume but higher per-visitor value.

The Librarian Metaphor: Conversational Search as Original Intent

Sullivan resurrects a 1996 metaphor from WebCrawler founder Brian Pinkerton to explain why conversational AI search feels revolutionary despite representing search’s original intended interaction model. Pinkerton described the fundamental limitation of keyword-based search: “It’s like someone walks into a library and they say travel and that’s that was the state of search right and it still is in some ways people give you one or two words.”

The ideal interaction: “The librarian would turn to you and say, ‘Well okay that’s nice did you want to travel anywhere in particular are you interested in the history of travel are you wanting to go on a boat on a plane a train?’ The librarian would engage in conversation with you, kind of get these information from you, if you will, do all those queries that you would kind of do and then say, ‘Here are some things that might be helpful to you.'”

Traditional search engines developed contextual proxies to approximate this interaction. Sullivan provides examples: “If you were to search for pizza on us today, we would understand first of all your location and the location of things related to you and we probably would show you local pizza places without you having to say I want local pizza places.” The system infers intent from location signals, query patterns, and aggregate behavior data.

Conversational AI search eliminates the proxy layer. Users can now initiate with the full context: “I want to make a pizza. Can you tell me the kinds of stoves that would be useful for me to use if I want to do it in my backyard or garden or whatever?” The system handles follow-up questions, maintains context across the conversation, and refines recommendations based on user feedback — the librarian interaction Pinkerton described nearly 30 years ago.

Sullivan’s conclusion: “I think that we’re getting into this more natural way of searching, the way you probably would have wanted to search way back when.” The technology finally matches user expectations rather than forcing users to adapt their information-seeking behavior to system constraints.

Strategic Bottom Line: Optimize content for conversational query patterns by addressing the full spectrum of related questions and context variations users might express in natural language, rather than focusing solely on two-word keyword phrases.

The Keyword Tool Obsolescence: From Term Matching to Topical Authority

Sullivan identifies a strategic obsolescence in traditional keyword research workflows. The historical logic: “They would use various tools over the years to understand what were popular things people were searching for, what were specific terms and sometimes way in the past that was really important because you would be like oh am I trying to be found for this word or that word? Well, which is the more popular way people describe it?”

The example: “Am I trying to be found for generative engine optimization or answer engine optimization? Well, let me track which is more popular and then I’m going to make sure I write using that thing all along.” This approach made sense when exact match dominated ranking systems. Google’s semantic understanding was limited; using the precise term users searched for provided a measurable advantage.

Modern systems eliminated this constraint: “Search engines got smarter and they’re like, ‘Yeah, I know you didn’t use the exact word, but guess what? Sounds we can figure all that stuff out. You didn’t have to be as specific.'” Synonym recognition, entity understanding, and contextual interpretation mean content ranks for semantically-related queries regardless of exact term usage.

Conversational search accelerates this obsolescence. Sullivan poses the absurdity: “Now what do you do like when people are doing like entire sentences? Well, obviously you’ve got to start optimizing for the entire sentence. So right now make a page for every possible sentence that someone might know. Don’t do that please.” The combinatorial explosion of possible natural language queries makes term-level optimization mathematically impossible.

The alternative framework: topical authority. Sullivan returns to the core principle: “What is the general thing that you’re trying to write about and then again are you writing it in a way that a human being would expect it to be written for them?” Content that comprehensively addresses a topic — with clear explanations, supporting evidence, and authentic expertise — ranks for the full semantic cluster of related queries, regardless of specific phrasing.

Strategic Bottom Line: Deprecate keyword-centric content planning in favor of topic-centric strategies that establish comprehensive authority on core subjects, ensuring content satisfies the full range of natural language queries users might express around those topics.

Implementation Roadmap: Transitioning to AI-Era Content Strategy

Sullivan’s guidance consolidates into a four-pillar implementation framework for organizations navigating the AI search transition:

Pillar 1: Content Audit and Commodity Elimination. Identify content assets that provide no unique perspective, firsthand experience, or analytical framework. These commodity information pages face declining value as AI systems answer questions directly. Redirect resources toward original analysis and experiential content that demonstrates authentic expertise.

Pillar 2: Multimodal Asset Development. Expand content formats beyond text to include video, audio, and image-rich presentations. This isn’t about creating separate “video content” — it’s about providing multiple modalities for the same core expertise, ensuring eligibility for discovery across query format types and AI-generated answer experiences.

Pillar 3: Topical Authority Architecture. Restructure content operations around comprehensive topic coverage rather than keyword targeting. Each core topic should receive treatment that addresses the full semantic cluster of related queries, supporting questions, and contextual variations users might express in natural language.

Pillar 4: Quality Conversion Measurement. Implement analytics that track visitor engagement depth, time on site, interaction with conversion elements, and downstream behavior. Optimize for per-visitor value rather than absolute click volume, recognizing that AI-mediated traffic may deliver different quantity-quality tradeoffs than traditional search.

Sullivan emphasizes that these pillars don’t represent new SEO principles — they represent the application of existing principles to evolving interface formats. Organizations that already prioritized user satisfaction, authentic expertise, and comprehensive topic coverage require minimal strategic adjustment. Those who built content operations around keyword manipulation and thin information pages face more fundamental restructuring.

The timeline consideration: Sullivan notes that Google’s AI Overview presence expanded rapidly, but the underlying ranking systems evolved gradually over years. Organizations that delayed strategic shifts waiting for “AI search” to arrive may find themselves behind competitors who maintained focus on core content quality principles throughout the transition.

Strategic Bottom Line: Execute the four-pillar framework as an integrated content strategy evolution rather than a revolutionary pivot, maintaining focus on user satisfaction and authentic expertise as the constant north star across all format changes.