Is SEO Dead? What the Data Actually Says

The strangest quarterly report in marketing right now belongs to the SEO manager whose rankings held. Every position defended, every audit green, and organic clicks sliding anyway, month after month. Nothing broke. Nobody got penalized. Google simply started answering the question above the links, and the click that used to prove the program was working stopped showing up. That gap, between a discipline that still functions and a metric that no longer pays, is where the “SEO is dead” argument lives, and it’s the same fault line running through the larger shift from search results to AI answers. Settling that argument with evidence is the whole job at Cite First AI. And the evidence says the obituary and the eulogy are both wrong.

How much traffic has SEO actually lost?

More than half of it, wherever AI answers appear. Ahrefs compared clickthrough rates across 300,000 keywords and found that an AI Overview now costs the top-ranking page 58% of its clicks, measured against a two-year baseline with December 2025 data. A widely quoted 34.5% figure from Ahrefs’ earlier study used a one-year window, so 34.5-to-58 is a change in method, not a measured deepening; never draw it as a line.

The trajectory that is real is the spread. AI Overviews now sit on about one in five results pages, and on nearly six in ten question-shaped queries, on Ahrefs’ late-2025 count of 146 million SERPs. The repricing isn’t a one-time event SEO teams can absorb and move past; it reaches more queries, more countries, and more of the path to purchase every quarter.

Independent measurement agrees. Pew Research Center watched 900 U.S. adults browse in March 2025 and found that when an AI summary appeared, people clicked a traditional result on just 8% of visits, versus 15% without one, and clicked a source inside the summary a vanishing 1% of the time. Different method, different data, same verdict: the answer absorbs the click.

Ryan Law, the Ahrefs content director who ran the clickthrough study, is not a man given to hype: his write-ups read like lab reports. Which makes his conclusion land harder: “Regardless of your feelings about the utility and value of AI Overviews,” he wrote when the December numbers came in, “it seems categorically true that they siphon away the majority of the clicks once available to top-ranking pages.” In a field that hedges everything, the click decline is the one finding nobody disputes anymore, not even Google, which now argues the clicks that remain are better, not that the missing ones will return.

So the mourners have a real corpse. It just isn’t SEO. It’s the organic click as a business model.

Is search dying, or just the click?

Just the click. Google disclosed in March 2025 that it handles more than 5 trillion searches a year, its first volume update in nearly a decade, and its biggest number ever. People aren’t asking fewer questions. They’re accepting more answers without ever leaving the results page.

Hold both facts at once and the “death” framing collapses. Demand for answers is at an all-time high; distribution of visits is in structural decline. A discipline whose inputs (queries) are growing and whose outputs (clicks) are shrinking isn’t dying: it’s being paid in a currency that’s losing value. The work of earning visibility still matters. The click-shaped receipt for that work is what’s disappearing, and where those never-happening clicks actually go is a different question from whether anyone is searching.

This is the demotion in plain terms: for twenty years, SEO was the strategy: rank, get the click, own the session, convert. Now Google AI Overviews, ChatGPT, Perplexity, and the engines coming up behind them sit between the question and your website, and SEO controls only the first half of the journey.

Do rankings still decide who gets cited?

Less than they ever have. When Ahrefs analyzed 863,000 results pages and 4 million cited URLs in March 2026, only 37.9% of AI Overview citations came from the first ten blocks of the same results page, and roughly a third ranked beyond the top 100.

One caveat the hot takes skip: Ahrefs’ 2025 study had put the top-10 share at 76%, but parsing scope, position counting, and Google’s underlying model all changed between the two, so the pair is two snapshots rather than a trend line. But the direction is corroborated across the industry’s tracking, and the mechanism behind it is documented by Google itself: query fan-out. When an AI answer is triggered, Google splits the question into multiple sub-queries, and the pages that win those narrower questions get cited, whether or not they rank for the term anyone was tracking. A page sitting at position 40 for the head keyword can out-cite the page at position 2, because citation is decided passage by passage, sub-question by sub-question. That selection has rules of its own, starting with how AI engines choose what to cite, and Google’s surfaces add a playbook of their own.

The questions themselves are drifting away from the keyword universe, too. When Semrush ran more than a billion lines of ChatGPT clickstream data against its keyword database across seventeen months, between 65% and 85% of prompts matched no traditional search keyword at all. The share phrased like a classic search is climbing, from 19% to 35% in the study’s final four months, but the bulk of what buyers ask an engine still exists in no keyword tool, and the engines answer it anyway.

For anyone still reporting rankings as if they were visibility, this is the uncomfortable math: the rank report and the citation report are now two different documents, and only one of them predicts whether a buyer hears your name.

Ranking is no longer the decision. It’s the audition.

What from SEO still transfers to AI search?

The foundation, almost untouched. Every major AI engine retrieves its candidate pages through a search index before it writes a word: Google’s AI features draw on Google’s index, and ChatGPT’s search layer has historically leaned on Bing’s. If crawlers can’t reach your content and search can’t surface it, no engine can quote it. What changed is what wins after retrieval.

A page an engine can’t fetch can’t be quoted, no matter how good the answer on it is. That’s why the technical half of SEO transfers intact, and why the content half needs rebuilding around evidence instead of keywords. The field learned the second part the hard way. Its founding experiment, Princeton-led GEO research from 2023, reported that adding statistics, quotations, and cited sources lifted AI answer visibility by 30 to 40 percent, and two years of industry advice got built on that number. Then Puerto and colleagues re-ran those exact rewrites and published the result at NeurIPS 2025: of 54 cases, only three improved significantly, and the add-statistics rewrite lowered citation rank in 19 of 24 settings. What survived the collapse is a sturdier principle. Across 252,000 controlled trials in 2026, claims backed by evidence beat unsupported ones in all six models tested. The lever was never adding numbers. It’s making claims a model can check, and the infrastructure underneath is now the price of admission.

Only keyword density leaves the building. The rest of the discipline got new job titles: from strategy to prerequisite, from goal to gate. The discipline of optimizing for language models starts where that gate ends, which is exactly where we built the Citation Cascade Method™.

Should you stop investing in SEO?

No, but stop buying it as a traffic strategy. Keep the technical and topical work that gets you retrieved, then shift the reclaimed budget toward what engines measurably reward: passage-level evidence, off-site mentions, and answers structured to be quoted.

For B2B companies, the reallocation is more urgent, not less, because the buying committee doesn’t research where consumer studies look. Venture firm Menlo Ventures, itself an Anthropic investor, found Anthropic holding 40% of enterprise LLM spend against OpenAI’s 27% in its December 2025 enterprise survey, which means the shortlists that matter to your pipeline are being assembled inside Claude and Copilot sessions on corporate laptops, not just in the ChatGPT app your marketing team tested. An SEO program tuned entirely to Google rankings can be invisible in every one of those rooms and still show green on its own dashboard. The business case for fixing that belongs in front of your board, not your SEO vendor.

Here’s where I’ve landed after 25 years of watching marketing channels get declared dead: every “is SEO dead” debate I get pulled into is really a measurement argument wearing a strategy costume. The people saying yes are staring at traffic charts. The people saying no are staring at revenue: from buyers who never clicked, and from the ones who did. Semrush’s July 2025 traffic study, which covers digital-marketing topics and discloses little of its methodology, measured the average AI search visitor converting at 4.4 times the value of a traditional organic visitor; Ahrefs’ own funnel, fully disclosed but a single site, put the multiple at roughly 23x. Both charts are real. Only one of them counts money. The discipline that connects them, earning the citation when the click never comes, is answer engine optimization, and it inherits SEO’s foundation the way a house inherits its plumbing: invisibly, and only noticed when it fails.

SEO isn’t dead. It’s load-bearing. Just stop expecting the load-bearing wall to also be the storefront.

Before the next budget review, ask the engines about yourself

Your CFO is going to see the traffic chart eventually. The only question is whether you walk into that meeting with the second chart, the one showing where your buyers’ questions are being answered now, and whose name comes back. Cite First AI runs that analysis as the first step of every engagement; it’s how InCorp went from invisible to cited in the engines its customers actually ask. Book a strategy call and bring your hardest keyword; the free mini audit will show you what the answer layer says about you before your buyers hear it.

FAQ: SEO in the age of AI search

Is SEO dying?

The click-harvesting version is. AI Overviews cut position-one clickthrough by 58% (Ahrefs, December 2025), and the loss deepened as AI answers spread. The retrieval foundation (crawlability, indexation, topical authority) is not dying; AI engines depend on search indexes to find the pages they quote.

Does SEO still work in 2026?

Yes, as a prerequisite rather than a strategy. Search volume is at record highs, over 5 trillion Google searches a year, and pages must still be indexed and retrievable to be cited by AI engines. What no longer works is treating rankings as the finish line and clicks as the payoff.

Will AI replace SEO?

AI replaces the interface, not the infrastructure. Engines like ChatGPT, Gemini, and Copilot still retrieve candidates from search indexes, so SEO’s technical work survives. What AI replaces is the outcome: citations in answers increasingly decide brand visibility where ranked links once did.

How do I know if AI engines are citing my brand?

Not by asking them once. AI answers vary by user, session, phrasing, location, and model version, so what you see isn’t what your buyers see. Tracking platforms like Otterly.AI and Ahrefs Brand Radar monitor citations across engines at scale, or book a call for a free mini audit of your AI visibility.

What are the key statistics on SEO and AI search in 2026?

The numbers that frame the debate, from our running AI search dataset:

• Google handles 5+ trillion searches a year (Google, March 2025)

• AI Overviews reduce position-one CTR by 58% (Ahrefs, December 2025)

• Users click results on 8% of visits with an AI summary vs. 15% without (Pew Research Center, 2025)

• 37.9% of AI Overview citations come from the first ten blocks (Ahrefs, March 2026; the 2025 figure of 76% used different methods)

• Adding statistics without substance lowered citation rank in 19 of 24 settings (Puerto et al., NeurIPS 2025); evidence-backed claims won in all six models tested (252,000 trials, 2026)