GEO didn't replace SEO. It raised the bar.

What the research actually says about AI search visibility for eCommerce.

Generative Engine Optimization got its name in a 2023 Princeton paper. In the three years since, it has acquired several sibling acronyms — AEO, LLMO, AIEO — and a small industry of specialists selling each one as a separate discipline.

It's worth knowing what the research actually found, because it points somewhere less dramatic and more useful than the marketing around it.

What GEO is

Generative Engine Optimization is the practice of structuring content so AI platforms — ChatGPT, Perplexity, Google's AI Overviews — cite, recommend, and surface your products when buyers ask questions.

Wikipedia defines it as structuring digital content to improve visibility in responses generated by AI systems, influencing how large language models retrieve, summarize and present information.

It is not a replacement for traditional SEO. It is not a prompt you add to your product descriptions. And it is not something you configure once and forget.

What the research found

This is where it gets interesting, because several studies set out expecting to find AI-specific techniques and mostly didn't.

The consistent finding is that whether your content gets retrieved and ranked in the first place matters far more than any AI-specific tweak to the content itself. Retrieval position, topical relevance, content recency, structure, authority — the things that govern classic search — dominate. Formatting tricks and tone adjustments have negligible effect.

Even the foundational Princeton research points this direction once you read past the headline. The interventions that moved visibility most were evidence signals: citations, statistics, quotable substance. In other words, the things that make content genuinely credible — which good content strategy has demanded for a decade.

Hand that list to an SEO in 2010 and they'd shrug. Yes. That's the job.

Why the fundamentals matter more now, not less

Here's the shift that is real.

Under traditional search, a mediocre product page ranked mediocre-ly. It still existed. Someone scrolling to position fourteen could find it.

In a generated answer there is no position fourteen. The system picks a handful of sources and synthesizes. You're either in the set or you're not.

So the same weaknesses that used to cost you a few positions now cost you the entire citation:

• Product descriptions copied from a manufacturer feed — identical to fifty competitors, with no reason to select you

• Category pages that are a headline and a product grid, offering no context to interpret

• Missing or incomplete structured data, leaving machines to guess at price, availability and specifications

• No third-party authority, so nothing corroborates that you're worth recommending

None of these are new failures. They're the same failures, with the tolerance removed.

The five things that matter most

1. Complete, original product content. Not rewritten manufacturer copy — actual description of what the product is, who it's for, and how it compares.

2. Structured data across the catalog. Price, availability, reviews, specifications. This is how machines read your catalog without inference, and it's the highest-leverage technical work available to a merchant right now.

3. Content that answers real questions. Sizing, comparisons, use cases, honest FAQs. Traditional search rewards it; AI answers are assembled from it.

4. Entity clarity. Consistent business name, clear about page, unambiguous description of what you sell and who you serve. Language models think in entities. If your brand is a fuzzy blob in the index, you lose ties you should win.

5. Third-party authority. Reviews, niche directories, earned mentions, community presence. Domain Authority is a useful proxy, but the underlying point is simpler: your site tells machines what you sell, everyone else tells them whether you're any good.

One job, not four

The reason this matters isn't competitive urgency. It's clarity.

Every new acronym implies a new discipline, a new budget line, a new specialist. For an enterprise with a twelve-person marketing team, fine. For the owner of a 4,000-SKU store who is also the CFO and the head of shipping, four acronyms produce one outcome: paralysis. It gets deferred, and deferred is how catalogs quietly decay.

The more useful framing is that there's one job. Make your business legible, credible and useful to every system — human or machine — that sits between you and your customer. Whether a shopper asks Google or asks ChatGPT, the answer is assembled from your product pages. Roughly 80% of the work is shared.

That's the framework we call Universal Search Optimization. The remaining 20% — where surfaces genuinely diverge, and where measurement gets genuinely hard — is worth understanding too, and we've written about it honestly there.

A word on timing

GEO is early. The paper that defined it is three years old. The measurement tools are immature, and anyone claiming precise attribution for AI-surface visibility is ahead of what the data supports.

That argues for doing the foundational work — which pays off on every surface today — rather than buying a specialist discipline for a surface nobody can yet measure well.

The merchants who spent two decades building crawlable sites, structured data, useful content and real authority didn't get disrupted by AI search. They got validated by it.

The bottom line

GEO didn't create new requirements. It removed the tolerance for skipping the old ones.

The question was never whether to do GEO or SEO. It's whether your product content is good enough to earn both — and, harder, whether you can keep it that way across a whole catalog while the market moves.

That second part is where most good stores quietly lose ground. We wrote about what that looks like over fifteen years, with one store's real numbers, in Don't mistake motion for action.

Further reading

• Generative Engine Optimization — Wikipedia

• GEO: Generative Engine Optimization — Princeton / arXiv

• What is Generative Engine Optimization? — Search Engine Land

• Universal Search Optimization — CataSEO

Keeping a full catalog complete and current — descriptions, metadata, structured data, alt text — is what CataSEO does, inside Shopify, BigCommerce, WooCommerce and Wix.