GEO Without Technical SEO Is Hot Air

A debate keeps resurfacing in SEO circles over the past few months: can a brand skip classic SEO entirely and invest only in GEO, optimizing to be cited by ChatGPT, Gemini, or Perplexity instead of showing up in ten blue links?

Google answered the question directly, and the answer is no. The technical foundation (a site a crawler can explore, understand, and index without friction) remains the condition of entry for hoping to be cited in generative search. That shouldn’t surprise anyone who has followed how these models actually construct their answers: they lean heavily on the same indexes and signals as classic search. A site that doesn’t index well doesn’t exist any more for a generative model than it does for a traditional search engine.

What worries me more than this debate is a different pattern repeating in front of my eyes.

SEO’s original sin is coming back under a new name

I spent part of this summer checking our own audit corpus, years of recommendations, against primary sources: official documentation, public statements from Google spokespeople, published independent studies. The exercise wasn’t comfortable. We found several thresholds presented as established fact that never had a real source behind them.

The most telling one: a widely repeated doctrine holds that beyond three clicks from the homepage, a page structurally underperforms. A precise threshold, cited as a technical fact. No Google statement supports that number. The actual position is far more nuanced: pages closer to the homepage carry slightly more weight in certain signals, with no magic three-click cutoff. On a catalog of several thousand products, it’s perfectly normal for the vast majority of pages to sit at a depth of three clicks or more. Presenting that as alarming, with a precise percentage attached, gives a false sense of rigor to what is, in reality, an entirely ordinary site structure.

We found the same thing elsewhere: the idea that a file like “llms.txt” is necessary to be well-treated by language models circulates widely, even though no engine requires it and a Google spokesperson has explicitly compared it to the meta keywords tag, an artifact the industry got in the habit of checking for, even though no engine actually uses it.

The folklore doesn’t stop at classic SEO

The same reflex is already reproducing itself inside GEO, not just in how GEO relates to classic SEO. A “citation score” or “share of voice in AI answers” number increasingly shows up in sales proposals, presented with the same reassuring precision as a conversion rate. In the vast majority of cases, that score rests on a sample of a few dozen queries run once against a single model, with no published methodology on how the queries were chosen or how much the result varies from one run to the next. A language model doesn’t answer the same question the same way twice with perfect consistency: the same score calculated on a Tuesday and again on a Thursday can shift meaningfully purely from that noise, with no real change having happened on the site at all.

That doesn’t mean measuring presence in generative answers is useless, only that the precision on display far exceeds what the method can actually support. A score presented to one decimal point gives the impression of a lab measurement when it’s really a noisy snapshot on a tiny sample. That’s the exact same gap between displayed precision and actual rigor that made SEO’s folklore both successful and damaging for fifteen years.

Why this isn’t a minor detail

A client technical enough to verify even one of these claims can bring down the credibility of an entire audit, including the findings that are actually accurate and well-founded. That’s the exact risk GEO is now reproducing, on fast-forward.

Generative search is young enough that almost no rigorous independent research yet exists on what actually influences a citation in an AI answer. That vacuum attracts the exact same reflex that produced SEO’s folklore: tool vendors publish “AI visibility” scores with no published methodology, agencies cite citation thresholds with no verifiable source, and the whole market starts repeating claims that sound precise without ever being so. The difference this time is that it’s happening at a speed SEO never saw, because everyone knows something has changed and nobody wants to be the last to react.

The discipline that protects an agency

The fix we applied to our own corpus is simple to describe and hard to maintain over time: every claim in an audit gets tagged by its actual origin. An official Google statement doesn’t carry the same weight as a well-conducted independent study, which doesn’t carry the same weight as house doctrine built on experience, which certainly doesn’t carry the same weight as a vendor claim repeated without verification.

House doctrine remains perfectly defensible: experience accumulated across hundreds of accounts is worth something, even without a study behind it. The problem is never having a point of view. The problem is presenting it as a fact established by Google when it isn’t one. The distinction looks minor on paper. It isn’t, in front of a client who checks.

What this means for GEO, concretely

Before paying for an “AI visibility” audit or tool, the question to ask isn’t “what’s my score.” It’s: what is this score actually based on, and can it be independently verified? The same rigor you’d demand of a classic SEO audit applies fully to GEO, with the added catch that the ground is even younger and therefore even more prone to shortcuts.

And before even talking about citations in generative search, the real priority remains the simplest to verify and the most often skipped: does the site crawl, index, and parse cleanly, with no technical friction? That’s not glamorous, it’s not new, and that’s exactly why most brands look for a reason to skip that step and jump straight to whatever feels more urgent.

Concretely, before ever opening a GEO file, three checks are enough to establish whether the technical foundation holds: does the site index in full with no unintentional blocks (robots.txt, forgotten noindex tags, JavaScript-rendered content that never gets rendered for crawlers); are the pages that actually matter reachable within a reasonable number of clicks from the homepage, regardless of the overall percentage; and does the content answer a precise question directly, in plain language, instead of circling the topic before getting to the point. None of these three checks require a proprietary tool or an invented score, just a careful read of the site the way a crawler actually sees it.

Google just confirmed you can’t skip it. GEO didn’t invent the technical rigor SEO was missing, it inherited it, folklore and all.