Product discovery is moving from short search strings to full conversations. On a recent episode of Conversations with MarTech, Lucas Tieleman, CEO and co-founder of Opiversal, walked through how generative AI search is rewiring what shoppers type, what they expect, and which products actually surface.
From six words to 24-word prompts
Traditional Google product searches average about six words. Tieleman points out that LLM-driven queries now run roughly 24 words, often with brand preferences, family context, and use cases baked in. Shoppers no longer trim their questions into keywords; they describe the situation and expect the platform to understand it.
That shift exposes a widening gap between how consumers ask and how most retail catalogs are organized. A product page built for a six-word query has little room for conversational intent, and retailers that ignore the change risk becoming invisible in AI-generated answers.
Bot traffic is the new research assistant
Autonomous agents are not yet doing most checkouts, but they are heavily driving research traffic. LLMs crawl retail sites to answer user prompts, which Tieleman says is changing how marketers should read impressions versus sales. A jump in bot traffic can look alarming, but it may also be a discovery signal if the site is structured well enough to be cited.
For our readers, the practical implication is clear: your site now has two audiences—human shoppers and the AI systems that pre-filter choices for them.
What retailers should change now
- Feed highly structured data, conversational attributes, and detailed FAQs directly into AI and advertising platforms.
- Use lower content costs to build custom landing pages at scale for shifting social trends and natural-language queries.
- Treat product feeds as a living asset, not a one-time catalog export.
The trap of one right answer
Tieleman argues that the industry’s greatest lie is the obsession with finding a single right answer. Modern tooling lets marketers test multiple paths and fail efficiently instead of betting on one correct message.
“With the technology that you have now, you can be wrong a lot more efficiently.”
The better AI search strategy is to generate many relevant variations, measure what surfaces, and keep the feed current—not to perfect one static page.
Why feeds matter more than ever
As AI systems take on more of the research funnel, discoverability shifts from ranking a page toward feeding the right structured signals into models. That includes the attributes, FAQs, and conversational language shoppers actually use. Retailers that treat search as a content system, not a fixed catalog, will be better positioned as AI overviews and agentic research expand.
Source: MarTech
