Search teams spent a decade learning to win the “blue links.” Now the blue links are getting skipped.
Semrush has reported that AI summaries can reduce clicks by 20–50%—a blunt number for anyone whose dashboard still treats organic sessions as the primary proof of life. But the same research points to a twist: conversions from AI-summary traffic were reported as 4.4x higher than traditional traffic. Less volume. More intent. That’s the tension hiding in 2026 search.
So the real question isn’t whether SEO is “dead.” It’s whether keyword strategy—built for rankings—can survive a world where answers are generated, summarized, and increasingly zero-click.
Nut graf: This is where GEO (Generative Engine Optimization) enters the picture. GEO is the practice of optimizing brands, products, and entities to be accurately represented and cited inside AI-generated answers (Google AI Overviews, ChatGPT, Perplexity), not just optimizing pages to rank. Experts describe it as an evolution that builds on SEO fundamentals—including E-E-A-T—rather than a replacement. The shift matters now because measurement, content structure, and even “what a keyword is” are changing at the same time.
The metric problem: rankings can’t explain what’s happening
Traffic loss has become the headline, but it’s the wrong KPI to obsess over in isolation. When the search experience turns into an answer box, an overview, or a chatbot exchange, the user may never visit a site at all. One widely cited figure in the GEO conversation is that 60% of U.S. Google searches in 2024 were described as zero-click. Whether a given team accepts that exact number or not, the direction is hard to argue with: more outcomes are happening on the results page.
That’s why enterprise SEO teams are being told to update what “good” looks like. Instead of treating rankings and sessions as the finish line, the better approach is to track what survives the interface change: conversions, LLM referral traffic, and share of voice inside AI answers. It’s not as tidy as a rank tracker. It’s also closer to revenue reality.
BrightEdge survey results captured the mood shift: 57% of marketers were cautiously optimistic about AI search changes, and 68% reported making strategic adjustments in response. Optimism is nice. Adjustments are the tell.
And there’s another reason measurement needs to change: AI-driven discovery tends to compress the funnel. If Semrush is right that AI-summary traffic converts at 4.4x the rate, then “down traffic” can coexist with “up pipeline.” That combination creates cognitive dissonance inside organizations that still reward teams for raw visit counts.
Keywords weren’t wrong. They were incomplete.
Classic keyword lists are built from fragments: “best crm,” “sales enablement software,” “pricing,” “alternatives.” They work because they map to how search engines historically matched strings of text to pages.
But conversational search behavior is moving toward longer, question-based, natural-language queries—often tied to voice and AI-driven experiences. In practice, that means the query is no longer a label. It’s a situation. A constraint. A preference. Sometimes a risk statement.
Ben Popper, writing for WRITER in March 2026, framed the shift plainly: “The website and SEO strategy you’ve been building for the last decade still matter. But you perfectly optimized keywords aren’t enough, at least not alone. They were built for search engines, not conversations.” In the same piece, Popper contrasts fragment keywords with how people actually ask for help inside chat-style interfaces.
WRITER’s article also describes a practical method: taking existing keyword data from tools like Ahrefs, Semrush, or Google Search Console and transforming it into natural-language queries aligned to personas. The example given is a hair restoration company called Hirstute, where 91 keywords were mapped into persona-specific conversational queries with rationale and recommendations. The important detail isn’t the category. It’s the workflow: stop treating the keyword list as the output, and start treating it as raw material.
GEO is “entity work” plus ecosystem work—both are operational
GEO changes the optimization target. Instead of asking, “Which page ranks?” teams have to ask, “How does an AI system describe the company, the product, the category, and the alternatives—and where does it get that information?”
That’s why experts emphasize GEO as an extension of SEO fundamentals, not a clean break. The same quality principles still matter. E-E-A-T still matters. But the surface area expands from on-site content to entity clarity across the web.
Here’s the operational piece that tends to surprise people: AI-driven search visibility is increasingly tied to structured data and centralized business information. In AI-mediated local and on-SERP experiences, inaccurate or fragmented business/location data can lead to exclusion. Not a ranking drop—an absence. The fix isn’t a clever headline. It’s data hygiene, ownership, and governance.
Then there’s the ecosystem angle. One claim cited in an April 7, 2026 podcast discussion is that up to 85% of AI recommendations may draw from third-party content (partners/influencers) rather than brand-owned sites. The research brief flags this as directional, not definitive, and it should be validated by category before budgets move. Still, it points at a strategic truth: “optimize our blog” is too narrow if AI answers are assembled from everywhere else.
In B2B, that “everywhere else” often includes partners, marketplaces, integration pages, analyst write-ups, comparison posts, and influencer commentary. GEO turns brand-building into a demand gen dependency, whether teams like the implication or not.
The conversational upgrade: a practical way to refit the keyword list
For a VP of Demand Gen, the temptation is to treat this as a content rewrite project. That’s rarely the highest-leverage move. The better sequence is: keep the keyword universe, then reshape it into questions that reflect intent and decision mechanics.
A useful internal test is simple: if a keyword can’t be spoken as a sentence, it’s not ready for GEO. Not because AI “only” understands sentences, but because AI answers tend to reward clarity about who the query is for and what would count as a good answer.
Three upgrades typically matter most. One: convert fragments into question-led topics (“best X” becomes “which X works for Y situation, and what should be checked before buying?”). Two: build comparison and constraint language into the content architecture (security, governance, integrations, pricing model—whatever the category punishes teams for ignoring). Three: make the entity legible across surfaces through consistent naming, structured data, and up-to-date canonical facts.
This is also where reporting has to grow up. If the next quarter’s organic deck still leads with rank changes, it will miss the point. GEO performance shows up in citations, presence, referral traces from LLMs, and downstream conversion quality—especially if click volume keeps getting squeezed.
McKinsey has forecast $750 billion in revenue routed through AI search by 2028. That projection is about money, not marketing theory. And it’s the cleanest way to understand why “conversational keywords” aren’t a copywriting trend: they’re an attempt to stay visible in the interface where purchase decisions are being pre-shaped.
The blue links won’t disappear. But the center of gravity has moved. Keywords still matter—just not as fragments on a spreadsheet. In 2026, the teams that win will treat the keyword list like a translation problem: from strings built for indexing to questions built for decisions, with entity clarity strong enough to survive being summarized by a machine.
