Is SEO dead? No. AI assistants answer questions by running web searches and reading the pages that rank, which means the craft of being findable, crawlable, and worth citing is now load-bearing for two audiences instead of one. What actually died is lazy SEO: the thin content, the keyword-stuffed filler, the publish-and-pray volume plays. AI didn't kill search. It raised the bar, and everything that couldn't clear the new bar is what people are mourning.
We run an SEO agency, so discount our take accordingly. But we'll make the case with mechanisms you can verify yourself, not vibes.
Why do people keep saying SEO is dead?
Three real changes, each misread as a funeral.
- AI Overviews sit on top of results: when Google answers the question at the top of the page, fewer people scroll to the links. Informational clicks on many queries are genuinely down. That one's real.
- Assistants answer without a search box: a growing slice of research that used to start at Google now starts in a chat window. If your model of SEO is "rank, get click, count session," that slice looks like pure loss.
- The old volume playbook stopped paying: sites built on mass-produced, low-effort content watched traffic evaporate. For businesses whose whole strategy was thin content at scale, SEO really is dead. Theirs, anyway.
Notice what all three have in common. They punish sites that were optimized for clicks instead of answers, and none of them punish being the source the answer comes from. That distinction is the entire argument.
There's also a simpler explanation for the volume of obituaries. "SEO is dead" is itself a headline that has ranked reliably every year for over a decade. Mobile killed SEO. Voice killed SEO. Social killed SEO. Each time, the channel absorbed the change and the practitioners who adapted kept compounding. This time is different, structurally. But "different" and "dead" keep getting confused, usually by people selling whatever comes next.
What does AI search run on?
Search. That's the part the obituaries skip.
When you ask ChatGPT, Gemini, or Perplexity anything that depends on current reality (prices, providers, products, "best X for Y"), the model doesn't recall an answer. It searches for one. It splits your question into multiple queries, runs them against ordinary search indexes, reads what ranks, and often runs follow-up searches to vet the specific brands it found. The first behavior is query fan-out; the follow-ups are grounding queries. Both are searches. Both read pages. Both reward the same things ranking always rewarded, plus clarity.
You can watch this happen. Ask an assistant a commercial question in your niche and look at the citations. Those are pages, ranking in indexes, selected because they answered cleanly. Dead channels don't have a selection process.
So the plain reframe: search didn't shrink, it grew a second consumer. Humans scan results. Machines assemble answers from them. A page that ranks and answers directly gets clicks from one and citations from the other. A page that does neither is invisible twice.
What died, and what survives?
Worth being specific, because "SEO" was always a bundle of practices, and the bundle just got sorted.
| Died | Survives and compounds |
|---|---|
| Thin content published at volume | Fewer pages that answer completely |
| Keyword stuffing and word-count theater | The keyword where it counts: title, meta, slug, H1, first sentence |
| Writing for crawlers | Writing answers machines can lift and cite |
| Generic "ultimate guides" restating page one | Real expertise: specifics, numbers, trade-offs, experience |
| Ranking as the finish line | Ranking as the entry ticket to being read by AI |
| Traffic as the only metric | Citations, recommendations, branded search, and traffic |
Read down the right column. That's not a new discipline. It's what good SEO practitioners preached for a decade while the industry cargo-culted word counts. The AI era didn't refute the craft. It audited it, and it flunked the pretenders.
Three surviving skills matter most.
Answer-first pages come first. Machines read to extract. A page that answers the title's question in its first three sentences is cheap to cite; a page that buries the answer under 800 words of preamble is expensive. Every page on this site opens with the answer for exactly this reason.
Real expertise is second. Synthesis engines are allergic to content that merely re-synthesizes them. What can't be generated from the index is what's worth citing: your data, your pricing, your trade-offs, your experience. If a language model could have written your page without visiting your site, why would it cite you?
The third is being the page the answer is built from. That's the new win condition. Not just ranking for the query: being the source an assistant quotes, links, and names when it answers the question. A stricter standard than ranking, and a bigger prize, because a recommendation carries more intent than a blue link ever did.
But if AI answers everything, why does the click matter?
Two reasons, one measurable and one not.
The measurable one: clicks didn't go to zero, they got denser. The person who clicks through an AI answer wanted more than a summary. They're comparing, verifying, buying. You can track exactly what assistants send you; we published the GA4 and Search Console setup in how to measure AI search traffic.
The unmeasurable one is bigger. When an assistant tells a buyer "consider these three companies," and you're one of them, no analytics event fires. The buyer shows up later as direct traffic or a branded search, and the AI's role is invisible. Concluding "AI sends us nothing" from a referral report is measuring the shadow and ignoring the object. Absence from analytics is not absence from the buying process.
So what should you do?
Not "more SEO." Sharper SEO, pointed at both audiences.
- Audit yourself the way a machine would. Ask ChatGPT and Perplexity the questions your customers ask. See who gets named, who gets cited, and whether you exist in the answer.
- Rebuild your money pages answer-first. Question in the title, answer in the first sentence, keyword in title, meta, slug, H1, and opening line.
- Publish what only you can publish. Your pricing, your process, your fair comparisons. Specificity is why pages get cited.
- Keep the fundamentals boringly solid. Crawlable site, clean structure, consistent entity info everywhere your business is described. Machines cross-check.
- Measure both layers. Rankings and traffic, plus citations, AI referrals, and branded search growth.
That sequence (audit, rebuild, harden) is our Reforge Method. We built the agency on the premise of this article: the blade still works, it needed reforging, not replacing.
Is SEO dead? The short version
No, and you can test that answer yourself faster than you can read another obituary. For the fuller map of how the old discipline and the new one fit together, read GEO vs SEO and what is GEO. Then run step one of the list above this week: ask ChatGPT and Perplexity the five questions your customers ask most, and note who gets named. If you're in the answers, harden your position. If you're not, you've just found next quarter's work.