Google News

Google officially positions AI search optimisation as SEO, not a separate practice

Updated:

Google has published an AI optimisation guide that consolidates existing SEO best practices under an AI-focused framework, explicitly confirming that traditional SEO remains the foundation for visibility in Google’s generative AI search features.

The guide marks Google’s position in a debate that has persisted since generative AI search emerged. Rather than introducing novel tactics or requiring special engineering, Google confirms that sites earning visibility in its AI-generated answers use the same fundamentals that earned traditional search rankings: unique perspectives, clear structure, accurate information, and credible authorship. This is Google’s view of Google’s systems; other AI platforms may weight signals differently.

What Google recommends

The guide emphasises distinctive, non-commodity content over generic information. Commodity content (information found across dozens of websites with little differentiation) underperforms in AI systems even more than it does in traditional search. Sites should develop unique perspectives, original research, practitioner expertise, or angles that add genuine value beyond surface-level definitions.

Support text with high-quality images and video where relevant. Avoid creating content variations solely to manipulate rankings (a tactic that signals low editorial quality to AI retrieval systems, which are trained on the same quality signals human raters use).

Technical foundations remain unchanged. Pages must be crawlable, indexed, and eligible for standard Google Search results. Semantic HTML (proper heading structure, list markup, definition pairs) and strong page experience across devices matter as much for AI features as they do for traditional search. This is particularly important because passages extracted from well-structured content are easier for AI systems to present cleanly in synthesised answers.

Google also recommends using existing tools for business visibility: Merchant Center for product information, Google Business Profiles for local and entity data, and emerging tools like Business Agent for customer-facing interactions.

What Google says to skip

The guide explicitly debunks several popular “AI optimisation” tactics that have been promoted as essential by some service providers:

  • llms.txt files: creating special machine-readable files is unnecessary for Google’s AI systems, which can extract and parse web pages directly. This is Google’s position on Google’s products. Anthropic publishes its own llms.txt for its documentation and co-developed the llms-full.txt format; whether Claude’s inference systems actively read other sites’ llms.txt files at query time is unconfirmed.
  • Content chunking: artificially breaking paragraphs into bullet points or short sentences to optimise for AI doesn’t work. It signals poor editorial quality and reads unnaturally to human visitors. AI systems favour content written naturally for people.
  • AI-specific rewrites: rewriting existing content specifically to trigger AI retrieval signals is ineffective. AI retrieval systems evaluate the same quality signals that human readers do. Content rewritten to sound “for robots” underperforms.
  • Schema markup designed for AI: adding schema markup (FAQ, HowTo, structured data) to content that doesn’t genuinely fit those patterns adds noise rather than signal. Google explicitly warns against schema stuffing.
  • Inauthentic brand mentions: manufacturing mentions of a brand across the web to game citations doesn’t work and violates spam policies.

The underlying message: tactics that treat AI optimisation as a separate game to be won independently of content quality will underperform. Sites earning consistent AI citations are those that would have ranked well in traditional search anyway.

What this means

Google’s guidance takes a clear position in a debate that has fragmented the SEO industry: from Google’s perspective, Answer Engine Optimisation (AEO) and Generative Engine Optimisation (GEO) are not separate disciplines. There is no “GEO playbook” distinct from SEO, at least not for Google’s AI features. Bing, Perplexity, and Anthropic operate different retrieval systems with different signals; the guidance here is specific to Google.

For sites already doing SEO well: you already meet Google’s AI optimisation requirements. No additional engineering, special files, or content rewrites needed. If your site ranks for traditional search because it has unique expertise, clear structure, and trustworthy authorship, those same qualities earn AI citations.

For services promoting AI-specific tactics: this guidance signals that llms.txt generators, “AI optimisation” audits, and content-chunking consultants are not addressing a need Google’s systems have. Google speaks for Google; other AI platforms have not issued equivalent guidance and may weight these signals differently. Publishers should evaluate tactics against the specific platform they are trying to appear in, not assume a single playbook covers all surfaces.

For the measurement side: AI visibility comes from the same content signals as traditional visibility. The unit of retrieval has shifted from pages to passages, which means structure and clarity matter more than before. But the fundamentals: accuracy, authority, and comprehensiveness remain unchanged. Sites should continue measuring SEO success through traditional channels (rankings, organic traffic, conversions) whilst layering in AI citation tracking via Search Console and GA4’s AI Assistant channel.

Sources

More news