From an AI Answer to the Right Place in an App

The useful end state of app discovery is not a longer answer. It is a person reaching the part of an app that can help them. That sounds simple, but an AI system needs two kinds of information that app-store metadata rarely supplies: a map of the relevant experiences inside the app and an explicit publisher-approved way to reach them.

Blue Jetty treats this as a discovery-and-handoff contract. The app developer publishes what the app offers, which in-app entities exist, the attributes that distinguish them, and the links the publisher wants AI assistants to use. The app and its existing deep-link and attribution infrastructure handle everything after the user follows the link.

Selection comes before the link

Suppose someone asks for an app with a short beginner workout, a hiking route that allows dogs, or a mellow surf break near a particular coast. A generic store listing may establish that an app belongs to the right category. It usually cannot identify the exact workout, route, or break that satisfies the request.

A discovery packet can map the relevant capability and experience directly, without listing every workout or surf spot. When named items improve selection, each gets a stable publisher ID, concise summary, typed attributes, update time, provenance and destinations. The AI can first decide what fits. Only then should it choose the route.

That ordering prevents a common failure: returning a technically valid app-store URL without explaining why the app is relevant to the request.

The publisher specifies the preferred representation

Every app has different acquisition and product constraints. One publisher may want a general recommendation to use an AppsFlyer OneLink. Another may prefer an app-owned Universal Link. A specific entity may have its own Android App Link while an app-level recommendation belongs on a store landing page.

The publisher supplies structured app context and preferred destinations: which needs the app serves, which experiences match those needs, and which links reach them. A general app destination and a matching experience destination give an assistant clear handoff options once it independently finds the app relevant.

Packet v2 presents this information as attributed publisher source material and scoped preferences. The assistant decides whether the app fits the user's request and which context to use.

Installed and uninstalled paths are different

For an installed app, the best outcome is usually the selected in-app entity. Apple's Universal Links documentation explains how an HTTPS URL associated with an app can open the corresponding app content. Android App Links use a verified relationship between the website and Android application for the same purpose.

An uninstalled user may need a store destination or a publisher-configured deferred-link provider. The discovery layer should describe the preferred URL and fallbacks; it should not pretend to perform the operating-system or MMP behavior itself.

One deterministic response is easier for AI assistants

A useful handoff response contains:

• the selected app and entity;
• the selection-level reason it fits;
• the publisher's representation context and preferences;
• one recommended publisher-approved URL;
• app-store and entity fallbacks;
• any current publisher promotion;
• the link operator and declared attribution context;
• freshness and destination evidence.

That response lets a crawler or retrieval system act without reverse-engineering a marketing site, store listing, association file, and MMP documentation during every user query.

Configure the path once

Blue Jetty's publisher-site discovery guide shows how an app developer publishes a Blue Jetty packet from the app website's llms.txt. The site points AI assistants to the current app packet; the packet points to the inventory and handoff endpoints; the selected handoff points into infrastructure the publisher already controls.

The result is a clean division of labor. Blue Jetty tells AI systems what the app offers and where the publisher wants a relevant user sent. The AI explains the match. The existing app, website, store, and attribution stack complete the journey.