The fastest way to lose a shopper's trust is to sell them something that isn't there.
You've probably lived it. You find the item, you pay, and then a message arrives saying it's no longer available. The order is cancelled and the afternoon you spent planning around it is gone.
On JENNi, search is built to avoid that. It only shows items that are in stock. That's true for people using the site, and for AI agents searching through our MCP server.
The real cost of a phantom listing
A listing for something that can't be fulfilled looks harmless. It isn't.
- The shopper pays, waits, then gets a cancellation.
- Someone has to unwind the order and the payment.
- The shopper remembers, and next time they look somewhere else.
That last cost is the biggest one. Trust is slow to build and quick to lose. One bad surprise is enough to make a shopper hesitate on every order after it.
For software buyers, the problem is sharper. An agent that recommends an item that doesn't exist looks wrong in front of the person it's helping. Good inventory accuracy in ecommerce isn't a nice extra. It decides whether anyone relies on the answer.
Show what's in stock, hide what isn't
Our rule is simple. If we can't say an item is in stock, search doesn't list it.
Behind this is the SKU Graph. Each product is linked to the stores that carry it, as store offers. Every offer holds a price, a stock status, where the information came from and when we last saw it. Search reads that and keeps out-of-stock offers out of the results.
We're careful about what that claim covers. We say search. A product opened directly by its exact ID can still be quoted, and that's by design. But when you or an agent go looking for something, the list you get back is made of things that are in stock.
We also don't claim our stock picture is perfect. Shelves change. What we can do is be honest about what we last saw, and keep the doubtful items out of the way.
Think of it as editing. A good shop doesn't put empty shelves on display. It shows what's there, so the person walking in can pick something and leave happy.
How it reaches agents
When an agent calls search_catalog with a query and a 5-digit US ZIP, it gets back items that are in stock. Local delivery options come first, then ship-to-home fills out the page. Pages hold 20 results.
That means an agent answering "can I get this near me?" starts from a list that's already filtered. It starts from results we last saw in stock, and checkout checks again before anyone pays.
Before anyone pays, get_buying_options returns availability, price, delivery fee, the delivery promise and the total before tax. Checkout then re-verifies current price and availability on the secure page. So there are checks at more than one point, not just one.
What shoppers see instead
A shorter list, and a more useful one.
Instead of a long page of maybes, you see things you can actually get. When you pick one, you see the option, the date and the total before you pay. Fewer mystery listings, and fewer "sorry, this one is gone" moments after the card is out.
We'd rather show fewer results that hold up than more results that fall apart.
It also changes how a search feels. When every result is something you can get, you stop second-guessing the list. You compare, choose and move on. That's the experience we're after: less doubt, more done.
What this means for agents and their users
An agent is only as trustworthy as the catalog behind it. If the catalog is full of items nobody can ship, every answer the agent gives carries a hidden risk. Filtering at search means the agent's first answer is already a good one.
It also saves time. Fewer dead ends means fewer follow-up questions, fewer retries and fewer wasted turns for the person on the other side.
For the industry
If you run inventory, the lesson is the same one we follow. A listing is a promise. Only make the ones you can keep, and be honest about how fresh your information is.
Shoppers can try it at shop.jennipro.com. Want your shelves visible to AI agents? Get in touch at jennipro.com.