Google I/O Didn't Kill SEO. It Split It In Two.

After Google I/O, the industry started debating. TechCrunch said traditional search results were over. Time talked about big changes. LinkedIn added its usual takes, and the old "SEO is dead" story came back. Google, however, said nothing major had changed and that AI Search still uses the same indexing and ranking systems.

Before looking at Google's statements, it's worth changing the focus. Both sides are actually debating the wrong issue.

To understand the real change, look at what Google announced at I/O 2026. It wasn't about search ending or staying the same. Instead, the SEO market quietly split into two groups with very different economics. Most strategy plans made this week haven't noticed this yet.

What Actually Shipped

The recent updates are significant. Now, the Search box can take images, files, videos, Chrome tabs, and text. It can handle longer prompts and gives AI suggestions. Gemini 3.5 Flash is now the default model worldwide. AI Mode has one billion users each month, and the number of queries is doubling every quarter. Queries in AI Mode are three times longer than regular ones. Follow-up questions are up 40% month over month, and planning queries are growing 80% faster than others.

The biggest update is about information agents. These tools keep an eye on the web for users and send them updates via Google. They can watch things like apartment listings, new products, price changes, or anything else the user wants. This summer, these agents will launch for Google AI Pro and Ultra subscribers in the U.S. They come with new UI features, mini apps, and dashboards.

That last product category is where the economic model breaks.

The Bifurcation Nobody Is Naming

For the last year, AI Overviews have quietly run a sorting experiment on the open web. A field study showed Overviews reduced organic clicks on triggered queries by 38%, with no measurable drop in user satisfaction. People got what they came for and didn't click through. Google's Robby Stein said that if users don't engage with an Overview, the system may stop showing one for that query, meaning the AI surface self-corrects toward queries where it replaces the click.

I/O didn't change that trend. Instead, it sped it up and added another layer.

Here's how the split is becoming clearer:

Commodity content

AI now handles content that answers a user's question without sending them to a website. This includes things like store hours, return policies, definitions, comparisons, recipes, or guides found in many places. AI Mode and Overviews use this content. The page is still indexed and helps form the answer, but it doesn't get any clicks.

Source-grade content

On the other hand, there's original analysis, primary research, unique data, first-hand expertise, and anything the model can credit, rather than just summarising. Google's optimisation guide, released four days before the keynote, is clear: only non-commodity, self-created content can be cited rather than just used by the AI.

Most content made in the last decade falls between these two types. This includes rewrites with small changes, roundups that collect information, and explainers on topics that are already well known. This kind of content has been losing its impact, and Google I/O accelerated that. 

AI that answers single questions is one issue. Information agents are a completely different challenge.

An agent doesn't just replace a search - it replaces the whole relationship. The user tells Google what to track, Google checks the web often, and then sends a summary inside its own platform. The publisher's content is used, packaged, and might never get a visit. This could mean lower ad revenue for publishers, since fewer people will visit their sites. The keynote didn't mention the creators behind the content. All of this shows that users are becoming more loyal to Google agents instead of specific websites, making Google the main brand and turning sites into just sources.

The real risk isn't that links will disappear. It's that people will want them much less.

The Measurement Problem Is Worse Than It Looks

Search Console can't separate AI Mode impressions from regular organic ones. It also can't show when your content appears in AI Overviews. You only see total numbers and have to guess where they came from. Agent-driven content introduces another problem: if an agent reads your page and sends a summary to a user, there might be no visit recorded. The content was used, but analytics won't show it.

Marketers are about to decide what content to keep, what to cut, and where to spend more, but their tools can't measure what really matters. Google hasn't said if or when it will show agent-driven impressions in any reports, so this gap will likely stay open for a while.

The key question is: For each piece of content, which side does it belong to, and can you show proof?

If your content is a commodity, accept that it won't get clicks anymore and decide if it's still valuable for brand exposure, internal links, or just volume. If it's source-grade, invest even more. Original data, unique research, named experts, and first-hand reporting are now the only types that are safe from being summarised away.

The content in the middle needs a clear decision, not just an update.

And the measurement gap needs to be addressed. Things like brand search growth, direct traffic, branded query share, and tracking citations in AI are not perfect, but they're better than just using Search Console for now.

Both sides are technically right. The basics of SEO are still there, but how traffic works has changed. That's the main point from the keynote.

Featured Image: Google Gemini