Before an agent can buy, it has to find. This report maps how AI agents discover and shortlist products in 2026 and which signals tip a product onto the shortlist. Agents do not browse the way humans do. They issue a structured query, pull from a small set of authoritative sources, and shortlist three to five candidates before evaluating in depth. Our analysis of observed agent sessions shows that structured product data, source authority, and price clarity dominate the shortlist decision, while visual merchandising that wins human attention barely registers.
Discovery Source Mix
Where agents pull candidate products from differs by platform, but a few sources dominate every agent we track.
| Discovery source | Share of candidates surfaced | Avg shortlist inclusion |
|---|---|---|
| Structured product feeds | 34% | 61% |
| Retailer and marketplace pages | 27% | 48% |
| Aggregators and review sites | 19% | 42% |
| Brand-owned sites | 13% | 37% |
| Social and community sources | 7% | 21% |
Signals Agents Weight When Shortlisting
We modeled the relative weight each signal carries in the shortlist decision across the major agent platforms.
| Signal | Relative weight | Human equivalent importance |
|---|---|---|
| Structured data completeness | High | Low |
| Price clarity and currency | High | Medium |
| Source authority and consistency | High | Medium |
| Availability and stock signal | Medium | Low |
| Review volume and rating | Medium | High |
| Visual merchandising | Low | High |
Key Findings
- Structured feeds dominate discovery. Structured product feeds surface 34% of candidates and earn a 61% shortlist inclusion rate, the highest of any source.
- Agents shortlist narrow. Agents typically evaluate three to five candidates in depth, so missing the shortlist means losing the sale before evaluation even begins.
- Human and agent priorities diverge. Visual merchandising that wins humans carries low weight for agents, while structured data completeness that humans ignore is a top agent signal.
Methodology
Data is compiled from the Presenc AI monitoring platform plus public sources, with Presenc AI estimates used where authoritative figures are unavailable. Discovery and shortlist metrics are modeled from observed agent sessions and vendor-reported benchmarks. Projections use compound growth modeling. Findings are reviewed quarterly. Last update June 2026.
How Presenc AI Helps
You cannot be bought if you are never shortlisted. Presenc AI shows whether agents are surfacing your products, from which sources, and which signals are holding you back from the shortlist. Track which AI agents hit your site and review your agent-payable readiness so your products clear discovery and reach the small shortlist that agents actually buy from.