The Night the Clicks Went Missing
Somewhere between your query and your cursor, the answer appeared. Not a link, not a list—an immaculate paragraph stitched from the web’s collective memory. It had citations tucked in like after-dinner mints. It sounded confident, sometimes too confident. And just like that, the click you would have given a publisher turned into a satisfied nod, and a new tab never opened.
If this feels recent, that’s because it is. On May 14, 2024, Google switched on AI Overviews in the U.S., then pushed the feature across 100+ countries by October the same year, normalizing machine-written summaries at the top of results. By 2025, Google was even testing an “AI-only” mode—no classic list of blue links at all, just answers and a prompt to keep chatting. The center of gravity in search moved from “go find” to “here it is.”
The CTR Gravity Well
When answers move up, clicks move down. That’s not a vibe; it’s a measurable outcome. Ahrefs tracked what happens to top organic listings when an AI Overview sits above them and found position-one click-through rates drop by roughly a third—34.5%—versus comparable queries without an AI Overview. For informational queries, the falloff from 2024 to 2025 is stark. This aligns with real-user behavior measured by Pew: people are less likely to click any result when an AI summary is present, and the links tucked inside the summary earn vanishingly few clicks. The short version: summaries satisfy enough intent to suppress curiosity.
But the averages hide messy edges. Semrush, looking at before/after cohorts of the same keywords, saw zero-click rates slightly decrease when AI Overviews were present, a reminder that methodology matters and that intent type, query shape, and the composition of the page all shift the outcome. In other words, AI doesn’t always produce a click apocalypse; it changes who wins the remaining clicks and why.
The Publisher Weather Report
If you pay bills with page views, the weather has turned. Similarweb and others have tallied year-over-year drops in search referrals to news sites since AI Overviews arrived, a story the Columbia Journalism Review summarized as a “traffic apocalypse.” It’s not uniform—some outlets gain on brand and social while others crater—but the aggregate pressure is unmistakable. Meanwhile, AI platforms themselves now drive a growing (still small) stream of visits; TechCrunch put the figure at ~1.13 billion referrals to the top 1,000 sites in June 2025, up ~357% year over year. The faucet is opening, but not nearly wide enough to replace the classic search.
Europe’s regulators and publishers have noticed. Italian publishers filed a complaint with AGCOM, arguing AI Overviews siphon visibility and revenue and may run afoul of the DSA; in the U.S., Chegg took the gloves off and sued Google, claiming anticompetitive harm from those very summaries. The politics of “who gets paid when AI rewrites your work” just left the op-ed page and walked into court.
The New Physics of Being Found
Old SEO told you to earn a position and harvest demand. New SEO—let’s be honest, answer-engine optimization—asks you to be useful in chunk form and credible enough to cite. The shift is structural. In the UK, where the rollout matured early, SISTRIX tracked AI Overviews for roughly 15% of queries in recent months. That creates a permanent “answer box weather pattern” you can’t out-blog. If your information can’t be lifted cleanly, summarized crisply, and attributed unambiguously, the model will grab someone else’s.
This is the part where people declare SEO dead. It isn’t. It’s just acquired a hostile upstairs neighbor. The job is no longer “rank and wait.” It’s “design the canonical paragraph that a model wants to quote, then prove you’re the best destination if the user chooses to leave the summary.” That’s a content design problem and a conversion design problem, not a title-tag problem.
The Geography of Loss—and Opportunity
Informational queries are the soft underbelly; they’re easy to summarize and hard to differentiate. Bottom-of-funnel intent behaves differently. Pricing nuance, implementation trade-offs, regulatory angles, integration gotchas—these don’t compress neatly into a tidy paragraph. When they do, the paragraph often earns the click back to the source that shows its work.
You can see this split in the new referral math. Yes, AI answers skim curiosity clicks, but they also occasionally send qualified traffic downstream, especially in verticals where a cited source signals authority. And yes, AI referrals are rising fast, but even defenders concede the totals are small against Google’s juggernaut. That means your durable wins still live where searchers have to choose, compare, and commit: side-by-side comparisons, “Alternative to X” explainers, pricing/plan architecture, ROI calculators, integration documentation, implementation timelines, and accountability artifacts like SLAs and security notes. Those not only attract the remaining search clicks; they turn AI-summary curious visitors into leads because you’re showing the work the summary can’t.
Design for the Citation, Earn the Click
If answer engines are the first screen, your job is to write for the lift. That means atomic claims with dates and sources, Q&A blocks that survive copy-and-paste into a model’s context window, tables with explicit headings the model can quote without garbling, and short intros that answer “what, why, how” before you get lyrical. When Pew says users click less in the presence of a summary, your counterpunch is to become the source the summary leans on—then make the landing worth the detour.
And no, this isn’t just about content. If the presence of an AI Overview reduces position-one CTR by a third on certain queries, the remaining traffic has to convert like it matters. You redesign forms. You shrink friction. You swap generic CTAs for intent-matching ones. You run CRO like your pipeline depends on it—because it does.
The Laws Are Coming; So Is the Math
The legal perimeter will keep tightening. Publishers will test compensation theories; regulators will weigh transparency and consent. But most businesses won’t win their traffic back in court; they’ll win it—or lose it—in the analytics. Treat AI summaries as a permanent fixture and measure what they actually do to your demand: segment queries that trigger Overviews, monitor citation share by page type, track assisted conversions from AI surfaces, and model pipeline per 1,000 organic sessions instead of celebrating impressions. The point isn’t to “beat” the summary. It’s to build an answer stack where you get paid anyway.
So, Is SEO Still the Holy Grail?
If by “holy grail” you mean a bottomless well of free, compounding clicks—no. The well has a siphon in it now. But if you mean the most reliable way to capture declared intent at a reasonable cost of acquisition, then yes—with conditions. Classic SEO is the front door, answer-engine optimization is the foyer, and your on-site experience is the living room where decisions happen. The winners in 2025 aren’t debating if SEO is dead; they’re learning to live above a noisy bar, building insulation, and inviting the right people upstairs anyway.