The Most Important Thing Google IO Didn’t Announce: Certainty

There was an assumption quietly sitting underneath most IO predictions (including mine).

After Google IO, AI Mode becomes the default, and we optimize around that.

That would have been clean. Not easy, but clean. One dominant interface. One set of behaviors. One main “SERP” to study and defend in a roadmap.

Instead, Google signaled something messier: more AI, more places it can show up, and fewer guarantees about which version of “Google” your user will actually use in the moment. Google has been explicit that AI Mode is where frontier capabilities land first and that features may “graduate” into core Search over time. That’s not a switch. That’s a moving shoreline.

So here’s the real question: how do you build an AI SEO strategy when Google’s product direction is “more AI, but… TBD”?

If you only look at announcements, it’s tempting to treat this as “Google went all-in on AI.” And yes, they did. They’ve been expanding AI Overviews and introducing AI Mode as a dedicated, more conversational experience.

But the more important product signal isn’t that AI exists. It’s that user choice is becoming the first click.

Historically, most queries flowed through a single dominant lane: classic Search results with a handful of familiar modules. Now, the experience is increasingly a set of adjacent surfaces: classic Search, AI summaries and AI Mode experiences, voice-driven answers, image-first discovery flows (Lens and image search), and whatever variant of “assistive” interaction Google is testing this month.

When users pick the interface first, the interface determines the retrieval and ranking system you’re competing in. That means you’re no longer optimizing a single results page. You’re optimizing choice architecture.

And choice architecture is not something most SEO teams are staffed, tooled, or measured to manage.

The old reality was simple in a depressing way: rankings are opaque.

The new reality is worse: the SERP itself is variable. The same query can produce meaningfully different outcomes depending on whether the user stays in classic Search, triggers an AI Overview, pivots into AI Mode, asks the question by voice, or starts with an image.

That variability isn’t hypothetical. Even within AI experiences, Google has been iterating quickly. Third-party tracking has shown AI Overviews expanding aggressively and then pulling back in certain query classes, which means the presence, prominence, and behavior of AI surfaces can change fast enough to break your assumptions mid-quarter.

If you’re trying to explain performance to leadership, “the interface changed” becomes a recurring root cause. And it will be true. Which is precisely the problem.

Traffic can drop even if demand didn’t.

Not because your pages got worse. Not because you “lost rankings.” But because attention got re-routed into a different experience: an AI Overview that answers enough to end the session, an AI Mode path that keeps the user in a conversation, a voice answer that never shows a ten-blue-links click, or an image-led discovery flow that shifts the click to a different kind of asset.

This is where SEO dashboards start lying to you by accident. Wins can look like losses. Losses can get mislabeled as “seasonality.” And attribution gets noisier right when the org wants more certainty, not less.

Roadmaps that assume one dominant SERP layout get punished.

The team that builds a content plan around “this keyword’s SERP has X modules” is playing a game where the board changes shape. You can still win, but you won’t win by chasing the last visible movement.

You win by building infrastructure that survives churn.

That’s the part most I/O discourse skips, because it’s not as fun as screenshots.

More surfaces can mean more entry points.

If your brand is strong and your content is actually useful, fragmentation can create new ways for users to encounter you. AI Mode and AI Overviews are, in theory, citation systems as much as they are answer systems. Google has positioned AI Mode as a place where deeper exploration happens via follow-ups. That’s an opening for content that’s structured, verifiable, and easy to extract.

It also increases the upside of rich media and entity clarity. Image-led discovery and multimodal search reward assets that aren’t “just another blog post.” The teams who already invested in original visuals, product data, and clean information architecture are not starting from zero.

And, importantly, teams that treat SEO as infrastructure can outlast interface churn. That isn’t motivational poster talk. It’s systems behavior. When the distribution layer changes, eligibility, trust, and technical hygiene become the stable ground.

Surface fragmentation increases operational load.

More formats. More markup. More content types. More “should we make a version for voice?” conversations. More “do we need a video?” debates. More tooling sprawl.

Small teams can’t optimize everything, so they default to whatever looks easiest to ship and easiest to measure. That often means chasing vanity metrics right when vanity metrics are becoming less meaningful.

And this is the trap: when uncertainty rises, the org wants dashboards that feel firm. SEO teams can end up over-reporting what’s visible (rankings, impressions, average position) and under-reporting what’s real (qualified demand, assisted conversions, retention, brand lift).

Our manifesto has a blunt way of saying it: AI is not a shortcut; it’s a multiplier of existing quality or existing debt. If you’re already operating on fragile measurement and fragile infrastructure, this new era will not politely wait for you to catch up.

AI did not remove crawlability. It did not remove indexing constraints. It did not remove trust.

Google can’t cite what it can’t reliably fetch, parse, and understand. Your internal linking still controls topical understanding. Your site’s quality signals still gate distribution.

The tactical details evolve, but the order of operations doesn’t. Eligibility first. Then relevance. Then trust. Then distribution.

If your AI SEO strategy is “write more content that sounds like an answer,” you’re building the wrong thing.

AI surfaces reward content that’s easy to extract, cite, and verify. That means definitions that don’t meander. Claims that come with evidence. Clear frameworks. Original data. Visible expertise. Pages that behave like reference material, not like word count exercises.

Google’s own messaging around AI Mode emphasizes deeper reasoning and follow-ups. That implies a retrieval stack that needs durable, trustworthy building blocks. Your job is to manufacture those building blocks.

This is the least “SEO” sounding advice and the most important.

Social isn’t a nice-to-have. It’s a hedge against interface volatility.

Online PR and partnerships don’t just earn links. They create demand. Demand makes everything else cheaper. It raises brand search. It raises click propensity. It gives your content a life outside the question of which Google surface someone used on Tuesday.

If you’re solely dependent on a single UI you don’t control, you’re not doing strategy. You’re doing hope.

Rankings are not “the truth.” They’re one sensor, and the sensor is getting noisier.

Move reporting toward outcomes: qualified sessions, assisted conversions, pipeline impact, retention, and brand search lift. Separate demand from capture. Separate interface shifts from actual business performance.

This isn’t about hiding bad news. It’s about not hallucinating a story from incomplete signals.

First, map your asset types to likely surfaces. Not in a spreadsheet fantasy way. In a brutally honest way. Which pages are actually voice-friendly? Which assets are genuinely image-led? Which pages contain tight, quotable definitions and evidence that an AI system could lift cleanly without misrepresenting you?

Second, fix eligibility before you chase format. Pay down technical debt that blocks consistent crawling and indexing. Reduce index bloat that dilutes your signals. Add structured data where it genuinely improves comprehension and extraction, not because it looks like progress.

Third, create one or two anchor assets per priority topic. Not “ten supporting posts.” One page that is the best page on the internet for that job. It should carry proof. It should be easy to cite. It should feel inevitable. That is the kind of asset that travels across surfaces.

Fourth, stand up a repeatable distribution system. A newsletter that syndicates your best work. A social loop that repackages it without turning it into content confetti. A PR motion that puts your data and expertise in front of the people who shape demand.

Fifth, update reporting so it can survive ambiguity. Your dashboard should separate demand signals, SERP capture signals, and business value signals. When traffic drops, you should be able to answer: did demand fall, did the interface shift, or did we actually lose relevance and trust?

The most destabilizing change isn’t that Google went all-in on AI.

It’s that we don’t know which surface our users will choose, when, or why. And the more AI expands, the more that uncertainty compounds.

So stop trying to predict the next SERP shape. Build assets that remain useful when the interface changes. Build systems that measure value when visibility gets weird. Build channels that don’t depend on a single button.

If AI is a multiplier, uncertainty is too. Multiply the right things.

Vector Database: A database optimized for storing and searching embeddings.

Learn More

One more thing: AI is only as good as it’s operator, and if you are reading this newsletter, you are better than most!

Till next time,

PS: Let me know what you think of this issue, or anything else here: [email protected]