Please Do Not Ruin AI SEO, This Time Around

If “AI Mode” becomes the the default, are we about to relive the first thirty years of SEO, except with fewer clicks to go around?

Google has already normalized AI summaries at the top of results with AI Overviews, and it’s been explicit that newer AI-first experiences are part of the direction of travel. That’s not a rebrand. It’s a redesign of the SERP’s center of gravity.

When the interface changes, the incentives change. And when incentives change, SEO culture tends to rot in predictable ways.

This newsletter’s bias is simple: if you can’t explain the work in plain language, you probably can’t defend it when it breaks.

The first era of SEO produced a lot of smart operators, and a lot of institutional bad habits. The danger of AI-first SERPs isn’t that “SEO is dead.” The danger is that we respond by doubling down on the same mistakes, just with new labels.

If clicks compress, attention fragments, and answers get synthesized, the easy move is to chase whatever looks like progress. Rankings. Visibility scores. Domain numbers. “Cited in AI.” All the shiny dials that make a dashboard feel alive.

But AI-first SERPs don’t forgive vanity. They punish it.

Multiple analyses have suggested AI Overviews can materially reduce organic CTR for classic blue-link positions, even when you rank well. If your reporting still treats “we moved from #4 to #2” as the win, you’ll keep celebrating while pipeline stays flat.

That’s how you lose trust internally. And once you lose it, you don’t get it back with a better rank tracker.

Rankings are not revenue. They’re a proxy that used to correlate with revenue often enough that we got lazy.

AI-first SERPs weaken that correlation. Sometimes dramatically. So the hard reset is this: stop valuing rankings over ROI.

That doesn’t mean rankings don’t matter. It means you treat them like a diagnostic, not the outcome. The outcome is qualified demand you can attribute, defend, and repeat: assisted conversions, sales cycle acceleration, retention, fewer support tickets, lower CAC.

The second vanity trap is swapping distribution for “Domain Authority” and calling it strategy.

DA-style scores can be useful as directional signals, but they are not distribution. Distribution is your ability to place credible stories and useful resources in front of people who already have attention: publishers, analysts, creators, communities, and partners.

If AI answers increasingly synthesize from sources that already carry trust, distribution becomes the compounding asset. Not the score. Not the spreadsheet of “links we want.”

This is also where spam creeps back in. As AI surfaces become gameable, some teams will chase shortcuts. Google is already signaling that attempts to manipulate AI-based responses fall under its spam policies. The lesson is old: shortcuts don’t just stop working. They turn into liabilities.

When the SERP changes, the industry instinct is to reverse engineer it.

You will see more correlation studies. More “we analyzed 10,000 AI answers.” More checklists dressed up as certainty. Some of that research will be useful. Much of it will be noise.

Because you can’t solve the black box problem by staring harder at the box.

AI-first SERPs add layers of uncertainty: retrieval choices, summarization behavior, citation selection, and shifting UI layouts. Even in the AI Overviews era, reputable studies disagree on the shape of the impact across queries and industries, some find major CTR losses, others argue the effect varies and may even change user behavior in complex ways.

Here’s the practical stance: stop trying to reverse engineer a system that you and everyone else don’t fully understand, and instead build things that survive model churn.

That starts with content that is actually helpful when read by a human who doesn’t want to work for their answer.

AI Mode will reward clear structure, precise language, and verifiable claims. But that is not the same thing as writing content that reads like a thesaurus.

“Query fan-out” does not mean keyword stuffing with better branding. It means users explore adjacent questions in a single session. Your content should anticipate those adjacent questions and answer them cleanly, with evidence, examples, and boundaries.

If you’re “optimizing” by inflating copy with synonyms, you’re training your team to sound smart and be unhelpful. That habit aged badly the first time. It won’t age better in an AI SERP.

In the old world, you could sometimes win with mediocre brand strength if your technical foundation was solid and your link graph was aggressive.

In an AI-first world, brand becomes a filter. Not because Google has feelings, but because systems reach for sources that appear trustworthy, cited, and consistent.

That requires SEO to stop acting like a siloed department that touches only metadata and blog posts.

Push SEO best practices into PR, social, paid media, and brand strategy. Not as “SEO asks,” but as shared infrastructure.

When PR lands coverage, are you making it easy for search systems to understand what the company does, who it’s for, and what claims are defensible? When social creates narratives that spread, are you capturing that demand on pages that convert? When paid discovers messaging that works, does organic inherit it, or ignore it?

AI Mode doesn’t eliminate marketing. It makes coordination non-optional.

Volatile ranking shifts and unfamiliar SERP layouts will keep making teams feel like the floor is moving. That part is real. But the response is a choice.

You can chase new vanity metrics and call it adaptation. Or you can use this reset to make SEO more profitable and more durable than it has ever been.

The durable version of SEO in an AI-first SERP is boring in the best way: measurement tied to business outcomes, content written for humans under time pressure, and brand distribution built through relationships and credibility.

That’s the work that survives interface changes.

And it’s the work that, months from now, you’ll still be able to explain in plain language, without flinching.

Context Window: The amount of text an AI model can consider at one time.

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]