Is SEO dead now that AI answers questions directly?
No — but the job changed from ranking on a results page to being the source an answer engine quotes. Here's what actually transfers.
No, SEO is not dead — but the thing it optimizes for has moved. For thirty years the job was to rank a page so a person would click it. Now a growing share of questions get answered directly by ChatGPT, Perplexity, Gemini, and Google's AI Overviews, and the job is to be the source those systems retrieve, trust, and cite. Same discipline, new referee.
I've been through every version of this argument. I was building online services before graphical browsers existed, and I've watched "search is dead" declared about directories, about portals, about Google itself when universal results ate the ten blue links, and about mobile. Each time, the people who quit were replaced by the people who adapted. This shift is bigger than those — but it rhymes.
What actually died
The thing that's dying is the informational click. If your traffic strategy was ranking thin "what is X" articles and monetizing the visit, AI answers are eating that alive, and no tactic brings it back. The answer engine reads the page so the user doesn't have to.
What did not die is the underlying question every search system has asked since AltaVista: who can I trust on this topic? Retrieval-augmented systems still have to pick sources. When Perplexity answers, it cites. When ChatGPT browses, it chooses pages to read. When Google assembles an AI Overview, it pulls from documents it already trusts. Someone gets picked. The work is making sure it's you.
What transfers from classic SEO
More than people think. Crawlability, clean semantic HTML, fast pages, canonical URLs, structured data, a coherent internal link graph — the machinery that made you legible to Googlebot makes you legible to every AI crawler. I run these checks across my own network of sites, and the failures that keep pages out of AI answers are mostly the same failures that kept them out of the index ten years ago: content that only exists after JavaScript runs, no dates, no author, no schema, ambiguous entities.
What's new
Three things, in order of leverage.
Answer-shaped writing. Language models lift the first clean, self-contained answer they find. Open the page with a 40–60 word paragraph that answers the headline question outright — quotable without edits. Then earn the rest of the page. If your answer is buried under six paragraphs of wind-up, the model quotes someone terser.
Entity clarity. AI systems resolve entities, not keywords. Your name, your company, your bio need to be identical everywhere they appear — site, LinkedIn, GitHub, directories — and tied together with Person and Organization schema pointing at one canonical ID. A fragmented identity doesn't get retrieved, because the model was never sure the fragments were the same thing.
Being the primary source. Models are trained and grounded on the open web, and they can tell recycled summaries from firsthand material. Publish what only you know: your numbers, your failures, your logs. I can write about running a 17-paper AI newsroom network as one operator because I do it every day. Nobody outranks you on your own experience — in the old search or the new one.
The honest caveat
Nobody outside these companies knows exactly how retrieval weighting works, and anyone selling you a guaranteed "#1 in ChatGPT" placement is selling smoke. What I've described is the pattern I see across the sites I run and grade: pages structured as answers, published by clearly-identified entities, with firsthand substance, get cited. Pages that are none of those things vanish from the conversation entirely.
SEO isn't dead. It just stopped being about the click and went back to being about the thing it was always supposed to be about: being the credible answer.