Brands Face Three Common Errors Adapting to AI E-Commerce

For thirty years, online retail followed a consistent structure of a search box, a grid of results, and a page of specifications. That structure is changing as large language models synthesize answers instead of listing links and as shoppers delegate decisions to AI rather than browse catalogs.

Executives sometimes respond by producing large volumes of AI-generated content in an attempt to improve placement in answer engines. Forbes states that models trained on AI output degrade and that providers suppress pages resembling machine-generated filler.

The recommended approach is to supply novel, unique, and authentic content. Forbes notes that large language models already draw heavily from sources such as Reddit and LinkedIn during training. Traffic arriving from LLMs converted up to nine times better than ordinary channels in the author's research.

Brands have added chat interfaces modeled on general-purpose systems such as ChatGPT. Forbes reports that a standalone chat window often adds friction rather than service, citing the author's own 2024 launch of a consultative bot and Amazon's Rufus as examples.

The suggested alternative integrates conversation directly into product pages. 6 times better conversion in the author's A/B tests by surfacing relevant details and hiding others.

Brand teams continue to overwrite AI-generated rankings by pinning products or forcing positions the algorithm did not select. Forbes indicates that such overrides typically lower revenue and increase returns. The proposed solution relies on transformer models and longitudinal behavioral data to predict the next best product.

Fewer returns translate directly into operational savings, according to the article. The common element across the three fixes is the use of a brand's own transaction and interaction data. Forbes concludes that this proprietary record of intent and purchases cannot be replicated by general models and forms the basis for future branded AI systems.