Is SEO dead now that AI answers questions directly?
No. AI assistants reach the open web through crawlers, so the technical foundations of SEO — crawlability, speed, clean markup, correct status codes, genuine authority — are now prerequisites for two channels instead of one. What has shrunk is the value of ranking for informational queries that an assistant can answer without a click.
The prediction was that generated answers would collapse organic traffic and take the discipline with it. What happened instead is more specific and more useful: the two halves of SEO that were always bundled together came apart, and they are now worth very different amounts.
The half that got more valuable
Everything a crawler needs. Correct status codes, canonical tags that agree with each other, pages that render without JavaScript, structured data that resolves your organisation as a real entity, a site fast enough not to time out mid-fetch.
This work used to buy you one thing: a shot at ranking. It now buys you a shot at ranking and eligibility to be read by every assistant that fetches live pages. The same fix, twice the return. This is why a GEO engagement so often turns out to be technical SEO wearing a new hat — the audit finds the same broken things, they just matter more now.
The half that got less valuable
Ranking for informational queries. "What is a load cell", "how to calibrate an industrial scale", "difference between induction and resistance furnaces" — these resolve inside the answer box. You can rank first and receive nothing.
Which keywords still send traffic?
Transactional and navigational ones. Someone comparing suppliers, checking a price, or looking for a specific company still clicks through, because the decision needs the source. Purely informational queries — definitions, how-tos, quick facts — increasingly resolve inside the answer and send nothing.
This does not mean stop writing informational content. It means stop measuring it by sessions. Its job changed: it is now the material an assistant reads in order to have an opinion about you. Being the source of the answer has value even when it produces no click — but only if you are attributed, which is a structural problem rather than an editorial one.
What this means for a budget
How should I split budget between SEO and GEO?
They are not separate budgets in practice. Roughly seventy per cent of a first GEO engagement is classical technical SEO work, because that is what the audit finds broken. Treat GEO as an additional layer of requirements on the same site rather than a parallel channel with its own line item.
The genuinely new spend is small: entity and identity work, llms.txt, structured data beyond the basics, and the monthly cost of actually asking the assistants whether they name you. Everything else on the list is work you were supposed to have done anyway.
The measurement problem
The uncomfortable part is attribution. When an assistant recommends you and the buyer arrives by typing your name into a browser, analytics records direct traffic and gives the credit to nothing. Brand-search volume and assisted conversions move; the channel that caused them is invisible.
The only honest response is to measure the input directly — ask the models the questions your buyers ask, and record whether you are named. That is a different instrument from a rank tracker, and it is the one worth adding.