No, SEO is not dead. What died is the assumption that a high ranking automatically converts into traffic. Search engines and AI assistants still have to find, crawl and understand your pages before they can use them, and that discovery work is SEO. The change sits at the other end: an answer generated above the results can satisfy the searcher before any link is clicked, so position and visits have come apart. The discipline did not end. Its scoreboard grew a second column.
If SEO isn't dead, what actually died?
Three things, roughly. The first is the reliable link between rank and clicks. A page can hold its position while an AI Overview or a chat answer above it hands over the same information, and the click never happens. The second is the value of thin content written purely to occupy a slot: a model summarising five sources has no reason to quote the one that says the least. The third is the idea that SEO is a separate task you bolt onto a finished website, run by someone who never sees the build. When the question is whether a machine can read, parse and quote your page, the answer is decided in the markup, the server response and the writing, not in a monthly report.
None of that means the fundamentals stopped working. Crawlability, clean HTML, page speed, internal links, real topical depth: every one of those still decides whether you show up at all, in a blue link or in a generated answer. If anything they matter more now, because a model that cannot fetch your page cannot cite it either, and there is no consolation prize further down the results.
What is zero-click search, and how bad is it?
Zero-click means the searcher gets what they came for on the results page itself, from a featured snippet, a knowledge panel or an AI-generated summary, and never visits a website. It predates AI search by years (weather, currency conversion, opening hours) but generated answers extended it into questions that used to require a real page: how something works, what to look for, which option suits a particular situation.
How bad it is depends entirely on query type, which is why blanket statistics are close to useless. Broad informational queries lose the most. Queries with intent behind them (someone comparing suppliers, checking whether a business is credible, ready to book or buy) still produce visits, because the answer is not the destination. The practical move is to read your own Search Console data instead of an industry average: find the queries where impressions hold steady but clicks fall, and you are looking at your own zero-click problem, query by query, with names attached.
Does ranking still matter if nobody clicks?
Yes, for a reason that is easy to miss. The sources an AI answer draws on are not picked at random. Assistants that search the live web run queries and read what comes back near the top, so ranking well is often what puts you in the candidate pool in the first place. Ranking has partly become a qualifying round for citation rather than the prize itself.
What changes is what you measure. Position alone can look healthy while the value underneath it drains away. Two numbers now sit on every query that matters to you: click-through traffic, which Search Console still reports honestly, and citation share, meaning whether an assistant names you when it answers. There is no clean tool for the second one yet. Running your target queries against ChatGPT, Perplexity, Gemini and Copilot on a schedule and recording what comes back is unglamorous, and it is still the most accurate method available. We do exactly that for kiwastudio.com and for client sites, arena by arena.
What does AI search reward that classic SEO ignored?
Being quotable in isolation. A ranking system delivers your page whole and lets the reader find the relevant part. A model extracts sentences and rebuilds them inside someone else's answer, so a paragraph that only makes sense after three paragraphs of build-up rarely survives the trip. Writing that states the answer plainly, near the top, in a sentence that stands on its own, gets used. Writing that circles for 200 words before committing does not.
It also rewards being unambiguous about facts. Structured data that states what you are, where you operate and what you sell removes guesswork, which is why schema markup earns its place even though it buys no citation by itself. Same logic behind llms.txt, a plain-text brief telling a model what a site is and which pages matter. And it rewards content carrying something nobody else has, because summarising five identical pages gives a model no reason to name any of them.
Which SEO tactics stopped working?
Volume plays, mostly. Publishing dozens of near-identical pages to catch keyword variants now competes against a system that can generate that text on demand, instantly, at no cost. Keyword stuffing was already dead and is now conspicuously so, since a model reading for meaning is unimpressed by repetition. Bought links remain what they always were, a liability with a delay on it. Thin location pages that swap one town name for another convince nobody, human or machine.
What replaces them is slower and less scalable: fewer pages, each genuinely worth reading, each answering one real question properly. That is a harder brief for a content mill and a fairer fight for a small business that actually knows its trade.
What should a business do differently, starting now?
Start with what a machine sees. Fetch your own page with JavaScript disabled and check whether the content is there. A surprising number of modern sites return an empty shell that only fills in once a framework boots, which is fine for a browser and useless for a crawler that does not wait. We build client sites hand-written and framework-free for this reason among others: the content is in the first response, every time, for every reader.
Then fix the writing. Take your most important pages and move the answer to the top, in one or two sentences that would still make sense if a stranger pasted them somewhere else. Add structured data that says plainly what your business is. Keep the SEO groundwork you already have, because it is the floor all of this stands on. Then track both columns: Search Console for clicks, manual query checks for citations. Some pages will lose traffic and gain mentions. You cannot see that trade if you only watch one number.
So what is the future of SEO?
Less a funeral than a merger. The technical half (crawlability, structure, speed, clarity) is now shared infrastructure serving two audiences at once, a search index and a language model. The content half has to clear a higher bar, because it competes with a machine that can produce adequate text about anything for free. What a machine cannot produce is a first-hand account of how something works from someone who does the work, which is the thing worth writing and the thing worth citing.
Anyone announcing SEO's death is usually selling the replacement. The honest version is duller: the same craft, judged by a second standard, with a scoreboard that has not finished being built.
Same craft, second scoreboard.
Get a GEO audit