Does Traditional SEO Still Matter for AI Search? An Honest Answer

Someone senior has read that AI search is replacing traditional search and wants to know whether the SEO budget can be redirected. Someone else insists nothing has changed and AEO is a rebrand. Both positions are wrong, and the cost of acting on either is high. Here is the honest answer.

The short version

Traditional SEO still matters for AI search, but not for the reason most people give, and not in the form most agencies sell. The parts of SEO that concern being crawled, understood and structurally clear matter more than ever. The parts that concern winning a blue-link position matter less than they did. The parts that were always manipulation matter negatively.

The mistake is treating SEO as one indivisible thing. It never was.

Why traditional SEO still matters

Several AI systems retrieve from the search index

Google’s AI Overviews are generated over Google’s index. Several assistants use a commercial search API as their retrieval layer. If a page isn’t indexed, or is indexed badly, it isn’t in the candidate pool the engine draws from. Everything that makes a page findable therefore remains a precondition, not a legacy concern.

Technical health is the same technical health

Crawlability, sane site architecture, working internal links, correct canonicalisation, fast rendering, content that exists in HTML rather than only after a client-side render, clean status codes. AI crawlers are typically less patient than search crawlers, not more. A site that frustrates Googlebot will frustrate them comprehensively.

Structured data helps machines resolve entities

Schema markup doesn’t force a citation. It does make your organisation, authors, products, dates and FAQs unambiguous to a machine, which supports the entity resolution that underpins whether an engine can confidently name you at all.

Authority signals overlap heavily

The corroboration that helped a page rank — independent references, topical relevance, consistent entity data, real expertise — is largely the same evidence generative systems use to decide which sources to trust. The evaluation differs; the underlying inputs don’t.

Where the two genuinely diverge

Some SEO practice is now neutral or counterproductive for AI visibility.

• Position obsession. A page can be cited in an AI answer without ranking in the top ten, and can rank first without being cited at all. Rank tracking alone no longer describes your visibility.
• Keyword density and exact-match phrasing. Language models match meaning, not strings. Repeating a phrase adds nothing and degrades readability, which does cost you.
• Long preambles before the answer. Written to increase dwell time, they actively harm extraction. A retrieval system pulling a chunk from the top of your page should find the answer there.
• Content volume as a strategy. Publishing many thin pages to cover keyword variants dilutes topical signal and, where the content is low quality, actively suppresses it.
• Click-optimised titles. Curiosity-gap headlines that hide the subject make the page harder for a machine to classify.

What AEO adds that SEO never covered

These are additions, not replacements.

• Chunk-level self-containment. Each section must answer its own question without depending on the paragraphs above it, because retrieval operates on fragments.
• Answer-first construction. Conclusion in the first sentence under each heading, explanation after.
• Question-shaped headings. Matching real phrasing rather than keyword fragments.
• Extractable specifics. Numbers, steps, requirements, limits and definitions engines cannot synthesise from generic sources.
• Off-site accuracy management. How third parties describe you feeds directly into generated answers. Traditional SEO treated this as PR’s problem.
• Prompt-set measurement. Testing a fixed set of questions across engines is a measurement discipline with no SEO equivalent.

So how should the budget split?

There’s no universal ratio, and anyone offering you one is selling. The split should follow where your buyers actually are, which you can determine.

1. Check your analytics for referral traffic from AI sources. Imperfect, but directional.
2. Add a source question to your enquiry form and read the answers for a quarter.
3. Ask your sales team how often prospects arrive citing something an AI told them.
4. Run your key commercial queries in Google and observe how much of the result page an AI Overview occupies before an organic link appears.

For most organisations the honest conclusion today is that classical search still delivers the majority of discoverable demand, while AI-mediated discovery is growing and is disproportionately common among earlier-stage, research-heavy buyers. That argues for reallocating a meaningful slice of effort, not for a wholesale switch.

The practical position

Treat AEO as a layer on a healthy technical and editorial foundation, not a substitute for it. Concretely:

• Keep every technical SEO fundamental. Non-negotiable, and it serves both.
• Keep earning genuine authority signals. They serve both.
• Stop optimising for position as the goal, and start measuring citation and mention alongside rank.
• Rewrite priority pages answer-first and chunk-clean. This costs little and improves human readability too.
• Add the measurement discipline: a fixed prompt set, tested on a schedule.
• Retire the tactics that were always about gaming rather than clarity.

The uncomfortable truth for agencies and the reassuring one for clients is that most of what genuinely worked in SEO — clear writing, real expertise, sound technical hygiene, honest specificity — is exactly what works for AI search. What is dying is the layer of manipulation that grew on top of it.

If you want to see where you actually stand across both, our ARIA citation tracker measures citation and mention across engines, the AI citability scorer grades individual pages on extractability, and our AI visibility work sits deliberately on top of technical fundamentals rather than replacing them.

Frequently asked questions

Should we stop tracking keyword rankings?

No, but stop treating rankings as the whole picture. Track rank alongside citation rate and mention rate, and expect them to diverge. Pages that rank poorly but get cited frequently are telling you something useful about how you write.

Does schema markup make AI engines cite us?

Not directly. Schema helps machines correctly identify your entities, authors, dates and content type, which supports confident attribution. It is a clarity mechanism, not a citation lever.

Is AEO just SEO with new terminology?

It overlaps substantially but is not identical. The shared ground is technical health and genuine authority. The genuinely new parts are chunk-level answer construction, off-site accuracy management, and prompt-set measurement across engines.

Can a page be cited by AI without ranking well in Google?

Yes, and it happens regularly — particularly on niche procedural and specification questions where a well-structured page answers precisely, even if it lacks the link authority to rank highly. That’s one of the more encouraging features of the current landscape for smaller organisations.