How AI-Driven App Discovery Can Remain Attributable

AI assistants are becoming a new starting point for app discovery, but mobile attribution was not designed around a conversational recommendation. If an answer sends someone to an ordinary website or store URL, the publisher may see a web session, an organic install, or no durable acquisition signal at all.

The clean solution is not for an app-discovery service to become another measurement provider. It is to make the publisher's existing attribution-capable link legible to the AI system that is selecting the app.

Start with a publisher-approved link

An app developer may already operate Universal Links, Android App Links, AppsFlyer OneLink, Branch links, Adjust links, or Kochava SmartLinks. Those systems control what happens after a click: installed-app opening, store routing, deferred deep linking, campaign attribution, and reporting.

Blue Jetty stores the link the publisher has approved and returns it as part of the app's discovery packet. It does not shorten the URL, redirect the click, fingerprint the user, hold MMP credentials, or replace the publisher's SDK.

This matters because link ownership and link transport are separate facts. An AppsFlyer-operated HTTPS link may eventually open an app through Universal Links. A Branch link may preserve entity context through installation. An app-owned HTTPS link may be enough when the publisher's analytics recognizes referral parameters.

Preserve parameters instead of rewriting them

MMP links often contain publisher configuration that must remain intact. AppsFlyer's current attribution-link reference documents pid as media source and c as campaign on OneLink URLs. Branch's deep-link reference documents analytics tags including channel, feature, and campaign. Adjust documents recognized short-link parameters such as adj_campaign, adj_adgroup, and adj_creative.

A safe retrieval layer should never overwrite values the publisher already supplied. It may fill an absent, documented parameter only when the publisher explicitly selected that provider and campaign. It should return the base URL, resulting URL, applied parameters, consumer identity source, and a clear evidence state.

Referral identity has levels

There are at least three useful configurations:

1. A publisher-approved generic AI-discovery link identifies the channel broadly.
2. A known consumer requests a publisher-approved variant for that product.
3. An app-owned URL preserves a referral parameter added by the consumer.

OpenAI's publisher FAQ says ChatGPT search referral URLs include utm_source=chatgpt.com. That can make a web referral visible. Whether the signal survives a store installation and appears in an MMP report depends on the publisher's link and measurement configuration.

A query parameter or User-Agent can identify the requested consumer, but it is still declared identity. A public caller should not be able to manufacture “provider verified” status merely by typing consumer=chatgpt.

Measure stages separately

A reliable funnel distinguishes:

• crawler request to the app packet;
• app or entity selection;
• handoff response generated;
• link displayed and clicked;
• app opened or store reached;
• install attributed;
• deferred entity reached after first open;
• downstream product outcome.

Blue Jetty records packet/resource requests and handoff resolution requests or responses. These observations do not establish an assistant's internal app selection, whether a link appeared in its answer, or a human click. Your app and existing MMP measure eligible downstream activity when users follow your links. Keep retrieval and downstream reporting distinct; a missing MMP attribution does not mean the app context was unavailable or unused.

Make the configured path discoverable

The Blue Jetty publisher API lets an app declare structured app context and scoped publisher preferences, a preferred HTTPS handoff, store fallbacks, an MMP provider, and a campaign. AI assistants receive that configuration through public JSON or Markdown packets and entity-level handoff responses.

The business effect is straightforward: the AI system can recommend the app using the same acquisition infrastructure the publisher already trusts. The publisher gets a better chance of recognizing the source, and the user gets a path that preserves the experience that made the app relevant.