Is SEO dying, or becoming the infrastructure for AI search?

Why SEO is dead is the wrong question.

Every major change in search produces the same prediction. The interface changes; the need to discover, evaluate and retrieve information does not.

Featured snippets were supposed to kill SEO. Voice search was supposed to replace it. Social discovery was supposed to make it irrelevant. Generative AI has now inherited the claim.

Google’s 2026 guidance says its generative AI features remain rooted in core Search ranking and quality systems. To be eligible for visibility in those features, a page must still be indexed and eligible to appear in Search with a snippet.[1]

What changes is what happens after discovery. A page can rank without being cited. It can be cited without the brand being named. A brand can be named without receiving a visit. A recommendation can influence a purchase before the buyer reaches the website.

What SEO as infrastructure means.

Infrastructure makes participation possible. It does not determine every outcome built upon it.

A road network enables distribution, but it does not decide which product a customer buys. Search infrastructure enables retrieval, but it does not guarantee that an AI system will select, cite or recommend a particular source.

1. 01

   Crawl

   Can search systems access the page and its important resources?

2. 02

   Index

   Can the system understand the canonical page and include it in the searchable index?

3. 03

   Retrieve

   Is the page relevant enough to enter the candidate set for this information need?

4. 04

   Select

   Does the source provide the evidence, perspective or format the response requires?

5. 05

   Cite

   Is the source displayed as visible support?

6. 06

   Mention

   Does the answer name the company, product or expert?

7. 07

   Recommend

   Does the system suggest the brand for the buyer’s situation?

8. 08

   Visit

   Does the user click through after receiving the synthesised answer?

Traditional SEO operates heavily across the first three stages. AI search adds new selection layers. SEO supports entry into the system; it should no longer be the only measure of success within it.

Ranking high does not guarantee selection.

Ahrefs analysed 863,000 search-result pages and four million URLs cited by Google AI Overviews, comparing cited URLs with the conventional results for the same query.[2]

Limited to standard organic links the pattern stayed broad: 37.1% of cited URLs ranked in the top ten, 26.2% in positions 11 to 100, and 36.7% outside the top 100.

SEO is dead and ranking first guarantees AI visibility are both unsupported by this evidence.

Why AI cites pages that rank below you.

A conventional result answers the submitted query. A generative answer may expand that query into several related information needs.

Google describes this as query fan-out: the system issues multiple related searches across subtopics and sources before constructing an answer.[1]

If AI answers it, what happens to traffic?

Pew Research Center analysed 68,879 Google searches made by 900 participating US adults in March 2025. Of those, 12,593 produced an AI summary.[3]

The study covers one market, one search engine and a defined period, so it should not become a universal forecast. It does show why conventional organic traffic cannot describe the full value of search visibility.

Influence can occur before the attributable visit. Traffic remains commercially important, but it cannot reveal whether a brand was repeatedly included, excluded or misrepresented during the research journey.

Track ranking and citation apart.

Do not replace SEO measurement with one opaque AI visibility score. Connect distinct layers instead.

Google’s generative AI performance reports expose impressions, pages, countries, devices and dates.[4] Bing reports citations, cited pages, grounding queries and citation share, and says explicitly that those metrics do not represent rankings, authority, clicks or engagement.[5]

The platforms are separating search performance from citation performance. Your reporting should do the same.

What SEO and content teams should do now.

Six moves, in the order they pay off.

1. 01

   Preserve the discovery foundation

   Keep investing in crawling, indexing, internal links, page experience, relevance and genuinely useful content.

2. 02

   Add citation analysis to rank tracking

   Record what ranks, what is cited, which domains appear only in answers and which formats are selected.

3. 03

   Study the wider information need

   Map every question required to produce a useful answer, not only the wording of the original query.

4. 04

   Improve evidence, not just coverage

   Add original research, product facts, current pricing, expert explanation, first-hand cases, transparent comparisons and clear limitations.

5. 05

   Separate citation from brand visibility

   Measure citation, mention, recommendation, accuracy and sentiment as distinct outcomes.

6. 06

   Connect visibility to commercial outcomes

   Monitor AI referrals, branded searches, assisted conversions, sales-call mentions and pipeline, without presenting visibility as revenue proof.

The measurement model is getting larger: eligibility, ranking, retrieval, citation, mention, recommendation, visit, conversion.