At Search Central Live Milan (mid-June 2026), Google drew a sharp line: AI cites non-commodity content (proprietary data, original analysis, expertise) and synthesises around commodity content (generic, reproducible identically everywhere) without citing it. Google also confirmed that citation depends on signals across your whole site, not page by page. The takeaway: produce what AI can't manufacture on its own.
Direct answer: if your content looks like thousands of other pages, generic definitions, basic how-tos, broad category overviews, Google has no reason to cite your source over anyone else's in its AI answers. That was the core message Google Search's team hammered home on stage in Milan in mid-June 2026, at its Search Central Live conference, according to the detailed recap published by Barry Schwartz (Search Engine Roundtable), who attended the event.
The term Google used is now the reference: "commodity content" versus "non-commodity content." It's probably the most useful lens of the year for understanding who wins, and who disappears, in generative search.
Commodity vs non-commodity: Google's definition
Commodity content is interchangeable: it appears in near-identical form across thousands of sites. A dictionary-style definition, a basic tutorial, a broad overview of a topic. For Google, there's no reason to cite any one source: the information is undifferentiated, so AI summarises it directly without sending anyone a click.
Non-commodity content, by contrast, offers something you can't find elsewhere in the same form: original analysis, proprietary data, expert opinion, specific case studies, detailed process documentation. These are the formats that give Google a reason to cite your source rather than synthesise around it.
The line to remember: if an AI can generate your content without reading you, it doesn't need to cite you. Citation is the price AI pays for information it can't produce on its own.
Google also reiterated its hard line on Scaled Content Abuse: synthetic or programmatic text produced at scale, with no proprietary data or added value, is a target for its anti-spam systems. That's exactly the trap that caught sites which industrialised generic AI writing in 2024–2025. Google had already settled other gimmicks too: "chunking" content for AI adds nothing extra.
Site-wide signals, not just the page
The second major point from Milan: site-wide signals. AI Overviews don't evaluate a page in isolation. It's the overall trust and quality picture of your entire site that determines whether Google is willing to cite you at all. A great page buried in a site stuffed with commodity content starts at a disadvantage.
In practice, you can no longer "optimise one page" and expect to break into AI answers. Editorial reputation is built at the domain level, a point we develop in our guide to Google's official GEO playbook.
What about the click from AI?
Google showed an official slide that cuts against the "zero-click" narrative: "We've seen when people click from AI Overviews, they're more likely to spend more time on the site." In other words, clicks coming from AI are more qualified. Lower volume, perhaps, but visitors already filtered by the generative answer.
In parallel, Google is rolling out its AI Reporting (Beta) section in Search Console to isolate impressions and clicks tied to AI Overviews, AI Mode and Discover, plus an AI Settings panel to explicitly include or exclude your site. We broke down this new measurement tool in our analysis of Search Console's generative AI reports.
What this means for small and mid-sized businesses
The Milan verdict is good news for anyone producing real content, and a sentence for low-grade industrial writing. Three priorities to act on this week:
- Inject non-commodity into every piece. Data only you have (your numbers, your customer feedback, your field tests), a sharp opinion, a concrete worked example. If a sentence could come straight out of any blog, rewrite it.
- Clean up your whole-site signals. De-index or merge the commodity pages that dilute your authority. Thirty dense pages beat three hundred generic ones.
- Measure before you act. Turn on Search Console's AI reports to see where you're already cited, and where you're invisible.
To structure all of this without losing your nights to it, see our data on structured data and SEO in 2026.
Cicéro's take: Google didn't invent a new rule, it put a name on one. GEO isn't technical tinkering; it is producing what AI can't manufacture on its own. Commodity content was already dead for traffic; now it's invisible for citation too. That's exactly the work Cicéro exists to do.
What this news doesn't cover
Google published no numeric scale: "commodity" and "non-commodity" remain qualitative notions, not a measurable score in a tool. The statements reported come from an in-person event (Search Central Live Milan) and an attending journalist's recap, not a line-by-line official blog post. Finally, Google's observation about time-on-site after an AI click is an aggregate internal trend, with no detailed public data by sector. Treat it as a strong strategic direction, not a metric to reproduce.
Sources
- → Search Engine Roundtable: recap of Search Central Live Milan: chunking, site-wide signals, commodity content, paywalls, clicks from AI Overviews (June 2026).
- → Google Search Central: official guide "Optimizing your website for generative AI features": GEO is still SEO.
- → Google Search Console: generative AI performance reports (AI Reporting Beta) and AI Settings.
Growth and SEO content strategist, I founded Cicéro to help businesses build lasting organic visibility, on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.
LinkedIn