Generative engine optimization spent 2024 as a concept people talked about at conferences. In 2025, it became something people actually did. The question shifted from "should we think about this?" to "what specifically should we do?"
Having worked with ecommerce retailers throughout this transition, here's our honest assessment of what worked, what didn't, and what matters heading into 2026.
The techniques that actually moved the needle
Structured data got more important, not less
The biggest surprise of 2025 wasn't some novel GEO technique - it was that structured data, which has been an SEO best practice for a decade, became significantly more impactful.
AI models pulling information for generative answers lean heavily on structured, machine-readable content. Product schema, FAQ schema, how-to markup, and organization schema all became stronger signals. Retailers who already had clean structured data saw disproportionate visibility in AI-generated answers.
For ecommerce specifically, product schema with accurate pricing, availability, reviews, and specifications became table stakes. Retailers with incomplete or inaccurate structured data were effectively invisible to generative search results.
Content that answers specific questions outperformed content optimized for keywords
This is the core philosophical shift. Traditional SEO optimizes pages for keyword clusters. GEO rewards pages that directly answer questions that AI systems are trying to resolve.
The practical difference: a category page optimized for "men's hiking boots" (keyword) performs differently from a category page that answers "what are the best hiking boots for wet conditions under $200" (question).
Retailers who restructured their content around answerable questions - in product descriptions, category pages, buying guides, and FAQ sections - saw measurable improvements in AI search visibility. This aligns with what Google has been saying about helpful content for years, but generative search made it concretely measurable.
Internal linking quality became a ranking factor for AI citations
When an AI system cites a source, it's making a judgment about authority. One of the strongest signals of authority for ecommerce sites turned out to be internal linking structure - specifically, whether the site's own link structure treated a page as important.
Pages with strong, contextually relevant internal links were cited more frequently in AI-generated answers than orphaned or weakly linked pages. This makes intuitive sense: if your own site doesn't link to a page in meaningful ways, why would an AI system treat it as authoritative?
Our Linking Agent was originally built for traditional SEO, but the impact on generative search visibility was one of the clearest results we saw in 2025.
What turned out to be noise
"AI search optimization" as a separate discipline
Early 2025 saw a wave of consultants and tools positioning GEO as fundamentally different from SEO - requiring new skills, new tools, and new budgets. By mid-year, it was clear that the overlap between good SEO and good GEO was about 90%.
The fundamentals haven't changed: well-structured content, clear topical authority, strong technical foundations, and useful information for humans. The remaining 10% that's GEO-specific (structured data emphasis, direct answer formatting, citation-friendly content) is more of a refinement than a revolution.
Stuffing content with "AI-friendly" formatting
Some early GEO advice suggested restructuring all content with specific formatting patterns - bullet-point answers at the top of every page, TL;DR summaries, explicit question-and-answer formats. While some of this is useful, the mechanical application of "AI-friendly templates" across every page type produced awkward, unnatural content.
Category pages don't need TL;DR summaries. Product descriptions don't need FAQ sections stapled onto them. The format should serve the content and the user, not an imagined AI preference.
Obsessing over individual AI platform appearances
Teams spent significant time trying to track and optimize for specific appearances in ChatGPT, Perplexity, Google AI Overviews, and other platforms individually. The tracking is unreliable, the appearances are volatile, and optimizing for one platform at the expense of overall content quality was a losing trade.
The retailers who did best focused on making their content genuinely useful and well-structured, then measured overall organic performance rather than trying to game individual AI platforms.
Where this leaves us in 2026
GEO is real, but it's not a separate practice - it's an evolution of SEO that emphasizes structured, authoritative, directly useful content. For ecommerce retailers, the practical implications are:
Double down on structured data. If your product schema is incomplete or your site lacks proper markup, fix that before anything else. This is the highest-ROI GEO activity for most retailers.
Restructure content around questions, not just keywords. Category pages, buying guides, and product descriptions should answer the questions your customers actually ask. The Topic Sieve is how we identify what those questions are at scale.
Invest in internal linking quality. Not just volume of links - contextual relevance of links. A link from "waterproof hiking boots" to "trail running shoes" is worth more than a generic footer link.
Don't create a separate GEO strategy. Integrate generative search considerations into your existing SEO workflow. For a comprehensive overview of techniques, see our guide on generative engine optimization and our companion piece on answer engine optimization.
The retailers who treated 2025 as a learning year - testing, measuring, and refining their approach to generative search - are well-positioned for 2026. The ones who either ignored GEO entirely or went all-in on superficial "AI optimization" tactics have catching up to do.