Generative AI search adoption remains fragmented, agencies need dual SEO strategy

The biggest mistake agencies can make right now is assuming generative AI is replacing search in one clean sweep. It is not. Search Engine Journal’s read on the market is blunt: adoption is fragmented, about 82% of people have not used generative AI regularly, and 57% still prefer traditional search for important topics. That is the real strategy shift. The winning play is not “SEO or AI search,” it is both, because the audience is splitting by intent, risk, and trust.

Why fragmentation, not replacement, is the real story

Google has pushed AI Overviews from experiment into the core product. Sundar Pichai’s company said the feature began rolling out to everyone in the United States on May 14, 2024, then later expanded it to more than 200 countries and territories and more than 40 languages by May 2025. That is a serious product commitment, not a side quest. But product rollout does not equal behavior convergence, and the usage data keeps proving that point.

The practical takeaway for agencies is simple: classic SEO still has to perform while AI visibility is built in parallel. Organic rankings, comparison pages, and publisher context still matter because a large share of users are not ready to trust a chat-style answer for anything consequential. If you are managing client accounts, especially in categories where the purchase or decision carries risk, you cannot trade away search traffic for a theoretical AI presence and hope the math works out later.

What the click data says about user trust

Pew Research Center’s browsing study gives the clearest picture of what is happening on the results page. The study followed activity from 900 U.S. adults and 68,879 unique Google searches from March 1 to March 31, 2025. In that dataset, about 18% of Google searches produced an AI summary, and around six-in-ten respondents visited at least one Google search page with an AI summary during the month.

The more important number for agencies is the click gap. Pew found that users clicked a search result 8% of the time on pages with an AI summary, compared with 15% when no summary appeared. That is the kind of drop that changes reporting conversations with clients, especially when a campaign has lived for years on informational clicks. Pew also found that only about 5% of AI-summary citations and 5% of standard search results linked to news websites in the study, which helps explain why publishers and content-heavy brands are watching referral traffic so closely.

This does not mean AI Overviews wipe out traffic everywhere. It means the SERP behaves differently when an AI summary is present, and agencies need to segment by query class instead of treating every search as equally exposed.

Which client segments are most exposed, and which are safest

The most exposed accounts are the ones built on broad informational discovery, especially content that answers simple questions without much proprietary value. News publishers, top-of-funnel education sites, and generic how-to content are the easiest to compress into a summary, which is why the click-through hit matters so much. If the page can be paraphrased into a few lines, the user may never need to visit.

The most resilient client segments are the ones where users want verification, comparison, or consequence-aware advice. Financial services, medical, legal, insurance, and other YMYL-style categories benefit from AI-free search behavior because the user is already cautious. The same is true for high-consideration B2B purchases, technical buying journeys, and local service searches where the user wants names, prices, maps, or proof before acting. In those cases, users often tolerate AI for broad exploration, then return to traditional search for validation.

That is why agency strategy should separate query classes into at least three buckets:

• Low-risk exploration, where AI summaries can satisfy curiosity quickly

• Comparison and evaluation, where users still want links, charts, and source context

• Consequential queries, where trust signals, publisher credibility, and direct evidence matter most

If you build content only for the first bucket, you are overexposed. If you build only for the third bucket, you miss the scale of the first. The agencies that do this well will map intent to format instead of dumping every keyword into the same blog template.

What content formats still earn their keep

The content stack has to split the same way. For AI-driven discovery, concise definitions, structured FAQs, and tightly written explainers can help an answer engine understand the page. For traditional search, comparison pages, pricing pages, calculator tools, and original research still do the heavy lifting because they give users something the summary cannot.

That means agencies should keep investing in pages that are hard to collapse into a single AI answer:

• Side-by-side comparison pages

• Product and service pages with specific features, pricing, and proof points

• Original data, surveys, and case studies

• Author and editorial bios that build trust

• Support content with exact steps, screenshots, or workflows

This is where a dual-stack approach earns its name. One stack is optimized for discoverability inside AI summaries and answer-style interfaces. The other is built to win the click when the user wants to go deeper. Search Engine Journal’s point about fragmented adoption makes both stacks necessary, not optional.

What agencies should watch in reporting

A lot of reporting still stops at rankings and organic sessions, which is not enough anymore. If search is fragmenting rather than consolidating around one AI model, agencies need to track visibility at the query level, not just the domain level. That includes classic organic click-through rate, impressions, and non-brand share of traffic, but also whether AI summaries appear on the queries that matter most to a client.

Gartner’s consumer data reinforces why this matters. In 2025, Gartner said 61% of U.S. consumers wanted the option to toggle AI summaries on or off, 41% said generative AI overviews make searching more frustrating than traditional search, and 53% said they distrust AI-powered search results. Then, in January 2026, Gartner said only about one-third of consumers believe GenAI chatbots are as effective as search engines for learning new information. That is not a signal to ignore AI. It is a signal that trust remains uneven and the reporting has to show where users still prefer the old path.

In practical terms, agencies should add these KPIs to the dashboard:

• CTR on pages where AI summaries appear versus pages where they do not

• Query clusters by intent, especially consequential versus exploratory searches

• Share of traffic from comparison and verification content

• Brand lift and direct visits when AI summaries reduce clicks

• Referral mix by content type, not just by source

Tools such as BrightEdge and similar SEO platforms become more useful when they are used to watch both traditional rankings and AI visibility, not one or the other. The agencies that treat AI search as a parallel surface, instead of a replacement surface, will be the ones that keep proving value when the SERP keeps changing shape.

The bottom line is not that AI search failed to matter. It is that the market is moving in layers, and the smartest agency strategy is built for that messiness. Keep the organic machine healthy, build answer-engine visibility where it fits, and never assume every client audience wants the same kind of search experience.