Google says AI search still rewards unique, helpful SEO fundamentals

On May 21, 2025, Google Search Central said AI search performance still depends on foundational SEO. Eli Schwartz has made the same argument from the marketer’s side: if a page is not unique, helpful, and ready for an agent to use, a dashboard full of mentions will not make it valuable.

Google’s AI search still runs on SEO fundamentals

That foundation includes unique valuable content, page experience, crawlability, indexability, and structured data that matches visible content. From Search’s point of view, optimizing for generative AI search is still SEO, not a separate discipline with its own rules and shortcut stack.

A lot of AEO and GEO tooling reduces AI search visibility to checklists and scorecards, but Google’s documentation points the other way. AI Overviews and AI Mode are rooted in core Search ranking and quality systems, and Google says these experiences use retrieval-augmented generation and query fan-out to pull relevant pages from the web and generate answers with clickable links. If your content cannot be crawled, indexed, understood, and matched to what the page actually shows, it is starting behind the system instead of gaming it.

The preview controls still matter too. Google’s guidance calls out nosnippet and noindex as deliberate visibility controls, not afterthoughts. Teams that want AI citations but also need to protect certain content, control how pages appear, or prevent thin material from being surfaced as a source of record still have those controls.

Agent-ready is now part of the job

The newest wrinkle is that Search is no longer only a blue-link system with an AI layer on top. At Google I/O 2026, Google said AI Mode had surpassed one billion monthly users, with queries more than doubling every quarter since launch. The company also called its new AI-powered Search box the biggest upgrade in over 25 years.

Google also said users can now use agents directly in Search just by asking a question, and can continue conversational follow-ups from AI Overviews into AI Mode. That makes Schwartz’s “agent-ready” framing practical, not theoretical. A page now has to do more than rank or get cited. It has to support an answer path, a follow-up path, and in some cases a task path where a search agent can keep moving without losing the thread.

That does not mean every brand needs to build for some abstract AI future. It means the content has to be precise enough for machines, useful enough for people, and structured enough to survive conversational retrieval. If your page can answer a question but not support the next step, you are leaving the interaction half-finished.

Stop measuring the wrong thing

Schwartz has been blunt about the dashboard economy. In a January 21, 2026 Previsible interview, he criticized the industry’s focus on AI visibility dashboards that count ChatGPT mentions and citations. His objection was not to measurement itself. It was to measurement that cannot prove business value.

His standard is stricter: metrics have to connect to qualified traffic, pipeline, and revenue. Otherwise, AI visibility becomes a numbers game that looks impressive in a meeting and disappears when someone asks which opportunities were actually influenced. Schwartz also likened the current obsession with AI visibility tracking to old rank-tracking schemes.

In Google’s guidance, AEO and GEO are just labels for optimizing visibility in AI search experiences. From Google’s perspective, it is still SEO.

What to do instead

In an April 15, 2026 AirOps webinar, Schwartz put the operating model into four parts: customer understanding, funnel selection, asset creation, and brand visibility. It starts with the buyer, not the tool. Deep customer research still separates useful content from generic output, and funnel choice decides whether a page should educate, compare, convert, or support.

Use those pillars as a practical filter:

• Build pages around real customer questions, not keyword clusters copied from a dashboard.

• Publish unique, valuable content that says something specific your competitors did not say first.

• Keep pages crawlable and indexable, then make sure structured data matches what the visitor can actually see.

• Use nosnippet and noindex deliberately when you need control, not as a reaction after the page is already messy.

• Track qualified traffic, assisted conversions, pipeline, and revenue, not just citations or mentions in an AI answer.

AI Overviews and AI Mode are built on the same ranking and quality systems that have always favored strong content and technical accessibility, then layered with retrieval and generation to produce answers and links. If the content is thin, vague, or locked away from crawlers, the AI layer does not rescue it.