Shopify describes richer catalog data guiding AI-mediated product discovery, while an IEEE Spectrum proposal asks scientific publishing to expose research in a format agents can read, reproduce and extend.
Artificial Intelligence··Morning
Catalog data complements the search channel
Shopify says AI-driven traffic and orders to merchants' stores tripled year over year in the second quarter. Company president Harley Finkelstein presented that growth to analysts as another channel alongside traditional search. Search sessions grew 1.3 times over the previous two years and still make up roughly a third of storefront sessions. Shopify distinguishes the two discovery routes by the data they use: Search engines rank popularity from a handful of keywords, while AI agents make multiple calls into its catalog, read richer structured data and match it to a buyer's specific intent. Traditional search continues as a human-facing path while the same product information becomes more detailed input for a machine-mediated query path. That boundary also matters to TechCrunch's comparison with online publishing: Falling click-through rates from AI summaries in publishing are not presented as the outcome for Shopify's own storefront traffic.[1]
An agent-oriented proposal for scientific format
The second report, from IEEE Spectrum, examines a proposal that moves the machine-readability question from product catalogs to scientific literature. Lead author Jiachen Liu completed a computer science doctorate at the University of Michigan in 2025 after preparing the proposal during that work; in May she became a cofounder of Agent Native Research Lab in Palo Alto. Liu's starting point was seeing that a coding agent released at the end of 2024 could take over work she did as a researcher, while still requiring substantial manually built infrastructure on top of the model. The proposal argues for a format suited to a setting in which agents can read, reproduce and extend scientific work, instead of research infrastructure being designed chiefly around human readers. Its status is limited to a proposal about which reader the paper and its surrounding infrastructure should accommodate; no adopted publishing standard is announced. The report also supplies a caution: The spread of AI tools in research is contested, and some findings associate their support for individual careers with fewer new ideas and topics.[2]
Two separate discovery systems
The two developments share neither a product nor a deployment. Their resemblance lies in the question of how information is arranged for a machine that performs discovery. In Shopify's account, structured catalog data lets an agent build multiple queries around a particular shopping intent instead of merely imitating one search result. In the scientific-publishing proposal, research output has two functions for an agent: finding the text, then reading, reproducing and extending the work. The distinction is as important as the overlap. The commerce case rests on measured traffic and orders in operating stores. The science case describes a design proposed for agent-oriented infrastructure; publisher adoption of the format is not reported. Nor does either report show the human-facing channel disappearing. Shopify says traditional search is growing, while IEEE Spectrum presents the proposal through an interview and a news article written for human readers. Read together, the reports show that agent access need not replace human access: One information domain can preserve its human-readable surface while also gaining a more explicit structure for machine queries.[1], [2]
