“Is SEO dead?” has been asked for as long as SEO has existed, and the answer has always been the same: no, but it changed. The current version of the question is different in one respect — this time the change is visible in the answers themselves. When an assistant names a shortlist of businesses, the ten blue links are no longer the only route to a customer.

So it is worth being precise. Some of what SEO was is genuinely finished. Most of it is not only alive but more important than before, because a machine reads your site more literally than a person ever did.

What is actually over

  • Papering over a thin site with keywords. A model summarising your page cannot be fooled by repetition; it needs a real answer. This was always poor SEO; now it fails faster.
  • Treating the ranking as the outcome. A position-one page is not a citation. In my baseline the AI-recommended set barely overlapped with the Google top results, and a ranking without a summarisable answer can be skipped entirely.
  • Buying links or citations to fake authority. Answers lean on corroboration between sources that are actually trusted. Manufactured consensus is brittle.
  • Publishing at a fixed monthly quota. Volume without a specific answer is noise, both to a reader and to a model.

None of that is a loss. It was the part of SEO nobody should have been doing.

What is not only alive but more important

  • Knowing what your market actually searches for. The starting question is unchanged, and it is still the one most businesses skip.
  • Answering real questions on dedicated pages. The page that answers directly is the page that gets cited — by Google, by AI Overviews, and by assistants.
  • Technical clarity: structure, speed, structured data. Machines need to parse your facts. This is SEO’s technical half, and AI search raised its value.
  • Earning mentions and links by being worth mentioning. Off-site reputation is now directly load-bearing for AI citation, not just a ranking signal.
  • Measuring against a baseline. Position tracking still matters; it is just no longer sufficient. Citations and mentions are the new columns.

Look at that list next to the classic one and you will notice they are almost the same. The discipline did not die; the surface multiplied.

Why “AI search” and “SEO” are the same job now

It has become fashionable to sell AI visibility as a separate programme from SEO. That is mostly a labelling exercise. An assistant needs exactly what a search engine needs in order to understand you — clear pages, real answers, structured facts, a current site — plus the corroboration that makes you worth naming. Those are the same foundations. The difference is the measure: instead of, or as well as, a position, you count citations and mentions.1

The businesses that will do well are not the ones that pivoted to a new acronym. They are the ones that already had the foundations — a site that answers questions and states its facts — and added a way to count being cited.

What to do if you are starting now

For a small business, the sequence is unchanged, which is the reassuring part:

  1. Find out what your market actually asks. Research first, before a page is written or a penny is spent on a redesign.
  2. Build pages that answer those questions directly. One question, one page, answer up front, facts in structured data.
  3. Earn mentions where the answers are sourced. Directories, comparisons, reviews, press, the places assistants already trust.
  4. Keep it current. Dated, reviewed, updated. Freshness is a signal on every surface.
  5. Measure citations as well as rankings. The same question set, counted over time.

That is the whole method, and it is the same whether you think of it as SEO, AI SEO or GEO. The names will keep changing. The work — being the clearest answer to a real question — will not.

If you want the mechanism in full, read how AI assistants choose which businesses to recommend. GroundWork is the research that tells you which questions are worth answering; Build is the site that answers them clearly; Coverage is the ongoing work of staying cited.

  1. KOODOS AI-visibility baseline, 17 September 2026: 15 questions × Perplexity, ChatGPT, Gemini and Google AI Mode, 60 completed calls. Raw data in docs/research/ai-visibility-2026-09/. ↩