Keyword Search Is Dying. Is Your Catalog Ready for AI Agents? — PayPal
Source
youtube.com
Author
AI Engineer
Date
Why it matters
PayPal's experiments found that enriching merchant product data improves agent recommendations, but unstructured boilerplate dilutes the signal and increases hallucination. Useful if you are making catalogs or retrieval readable to agents.
Key takeaways · AI-distilled
PayPal's Nixon Dinh describes shopping moving from a search era of typed keywords, to an intent era where shoppers describe needs, to a delegation era where agents do the searching.
He argues catalogs syndicated for ads and keyword SEO are not readable by agents, and that hybrid keyword plus semantic search works better than either approach alone.
In PayPal's experiments, enriching merchant product data consistently improved AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → recommendations, and thin catalogs gained the most.
The limit he reports: over-enriching with unstructured boilerplate diluted the signal and made agents hallucinationWhen a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.Full definition →, so content quality mattered more than adding text.
Terms in this piece · Glossary
hybrid search — Combining keyword and semantic search in one ranked result, so exact terms and paraphrases both find what they should.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.
