Picked
The agent recommended your product as the single thing to buy. The top of the shelf - and the only result most shoppers ever act on.
Feature · Agent Shelf
The shelf the agent reads is the only one that matters now.
Shoppers no longer browse - they tell an assistant "buy me the best one under $150" and the agent decides. Agent Shelf tracks whether AI agents pick your products, names the competitor that won, quotes the reason, and tells you the exact feed fix that puts you back in the cart.
The prompts
Buying decisions, not searches.
An agent turns a shopper's goal into a buying-intent prompt, evaluates a handful of products, and returns one pick. Agent Shelf runs these prompts across every model and scores where your product lands.
- › best trail running shoes under $150, ships by Friday
- › a quiet espresso machine for a small kitchen
- › is [your product] worth it vs [competitor]?
- › most durable carry-on for frequent flyers
- › best gift for a new parent under €60
Four verdicts. One 0-100 Pick Score.
For every buying-intent prompt, Agent Shelf reads the agent's answer and grades your product - then rolls the verdicts into a single Agent Pick Score you can watch move as you fix your feed.
Shortlisted
Named as a good option among two or three others. In the consideration set, but the agent steered elsewhere for the final call.
Rejected
Mentioned, then explicitly passed over - "but X is cheaper / better reviewed / in stock." We capture the stated reason so you can answer it.
Absent
Never surfaced at all. The agent didn't know you existed for this decision. The visibility floor - and the most common starting point.
What Agent Shelf tracks per product.
Not just whether you were mentioned - whether you were chosen, why the winner won, and whether the agent is even reading your data correctly.
Pick rate across agents
How often each model picks, shortlists, rejects or never surfaces your product - per agent, so you see where ChatGPT loves you and Gemini ignores you.
Who won, and the reason
The competitor product the agent recommended instead, plus the rationale it gave - "cheaper," "better reviews," "in stock." The exact objection to fix.
Fact accuracy
Whether the agent is quoting the right price, availability and specs. Agents repeat stale or wrong product data constantly; we flag the mismatch against your truth.
Agent-readiness of your data
When we pull a product URL, we grade how machine-readable it is - clean schema, partial, or scraped. A poor grade is the first reason agents skip you.
Agent Pick Score trend
A 0-100 score per product, snapshotted daily, so you can prove a feed change moved the number instead of guessing.
The cited shelf
Which retailers, reviews and feeds the agent pulled from to make the call - the sources you need to win to change the outcome.
The plays that put you back on the shelf.
The lever for agentic commerce is your data, not a blog post. Each fix is tied to a specific losing prompt and the agent's stated reason.
Complete Product & Offer schema
Ship valid schema.org Product / Offer / AggregateRating on every PDP. Agents read these verbatim - missing or partial schema is the most common reason you grade "scraped" and get skipped.
Fix price & availability accuracy
When the agent quotes the wrong price or "out of stock," it stops recommending you. Correct the feed and structured data so the agent's facts match yours.
Publish an llms.txt
Give agents a clean, machine-readable summary of your catalog and key facts so they don't reconstruct you from third-party guesses.
Win the cited sources
Agents lean on a small pool of retailers and review threads. Agent Shelf names the ones deciding your category so you know exactly where to earn placement.
Structure the deciding specs
The spec the winner had and you didn't is often the whole reason. Surface it as structured data and in copy so the agent can quote it.
Monitor the drift
Agent answers change as the web re-indexes. Daily Pick Scores flag the day a competitor's change knocked you off the shelf - while you can still respond.
See if the agents are buying you.
Add a product by URL, add a buying-intent prompt, and get your first Agent Pick Score across ChatGPT, Claude, Gemini and Perplexity. Free plan available.