canonical: https://jentic.com/apis/amadeus.com/amadeus-flight-choice-prediction

# Amadeus Flight Choice Prediction

The Amadeus Flight Choice Prediction API scores a list of flight offers with the probability that a traveller will pick each option. It takes a flight-offers payload (typically the output of Flight Offers Search) and returns the same offers annotated with a probability value, so booking flows can re-rank or highlight the most likely choice. The API is intended to sit between search and pricing or order creation, helping travel sites surface fewer, better options instead of dumping a long list on the user.

## For AI agents

Score Amadeus flight offers with a per-offer probability that the traveller will choose them, so an agent can re-rank or pre-select the best option before pricing or booking.

## Scope

Does not search for flights, confirm prices, or create bookings - use only to score an existing list of Amadeus flight offers with choice probability.

## Capabilities

- Score each flight offer in a search response with a per-offer choice probability
- Re-rank a list of flight offers from cheapest-first to most-likely-to-be-chosen
- Filter long flight-offer result sets down to the top predicted choices before showing them to a user
- Pre-select a recommended flight offer in a booking funnel before pricing
- Feed prediction scores into A/B tests of itinerary ranking strategies

## Use cases

### Re-rank Flight Search Results

After calling Flight Offers Search, post the offers to /shopping/flight-offers/prediction to receive the same offers annotated with a choice probability. Travel sites use this to replace a default cheapest-first sort with a likelihood-of-purchase sort, which typically reduces the number of options the traveller has to compare and shortens time to booking.

Example prompt: Take the JSON response from a Flight Offers Search call, POST it to /shopping/flight-offers/prediction, and return the top 3 offers ordered by descending probability.

### Cut Down Long Itinerary Lists

Online travel agencies often receive 50-200 offers per route. By scoring the offers with the prediction endpoint and keeping only those above a probability threshold, a booking funnel can show 5-10 high-relevance options instead of an overwhelming list. The endpoint accepts the full search response unchanged, so integration is a single extra POST call.

Example prompt: Filter a 100-offer Flight Offers Search response down to offers with predicted probability greater than 0.05 and return them sorted by probability.

### Pre-Select a Recommended Flight

On a results page, highlight one offer as recommended by passing the search response through the prediction endpoint and selecting the offer with the highest probability. The API does not change the offer payload itself, only adds a probability field, so the same offer can be passed straight into Flight Offers Price or Flight Create Orders without further translation.

Example prompt: Pick the single offer with the highest probability from a Flight Offers Search response and pass it forward to /shopping/flight-offers/pricing for confirmation.

### Agent-Driven Itinerary Recommendation via Jentic

An AI agent that helps a user book a flight can call Amadeus Flight Offers Search, then call Flight Choice Prediction through Jentic to score the results, and present only the top-ranked option to the user with a short rationale. Jentic resolves the OAuth 2.0 client_credentials flow against Amadeus and returns a structured tool, so the agent does not need to manage tokens itself.

Example prompt: Call jentic.search('predict the best Amadeus flight offer'), load the returned operation, and execute it with a Flight Offers Search response as the body to retrieve the top-scored offer.

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| POST | /shopping/flight-offers/prediction | Score a list of flight offers with a per-offer choice probability |

## Key resources

- **Flight Offers Prediction** — Single endpoint that accepts a Flight Offers Search response and returns the same offers annotated with a per-offer choice probability.

## Why Jentic

- **Setup:** Wiring the Flight Choice Prediction API by hand means running the Amadeus OAuth 2.0 client_credentials exchange, refreshing the bearer token, and shaping the flight-offers payload into the POST /shopping/flight-offers/prediction body against the correct Amadeus host and version. Through Jentic you install once, import Flight Choice Prediction from the API Directory, store the Amadeus client credentials once, and your agent calls it.
- **Permission scoping:** The list of flight offers travels in the request body rather than as a resource id in the URL path, so scoping is by operation. You limit the agent to the operations it needs, such as scoring a set of offers with a choice probability, and it cannot reach search, pricing, or booking operations unless you add them.
- **Credential handling:** Your Amadeus client_id and client_secret are stored once, encrypted, by your own Jentic One instance, which runs the client_credentials exchange and injects a fresh bearer token at execution time. The raw secret never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'predict which flight offer the traveller will pick', and Jentic returns the /shopping/flight-offers/prediction operation with its input schema so the agent posts the offers and reads the probabilities without browsing the Amadeus docs.

## Related APIs

- **Flight Offers Price** — Confirm the final price of a flight offer before booking, after the choice has been predicted.
- **Flight Create Orders** — Book the flight offer that prediction selected as most likely.
- **Flight Delay Prediction** — Different prediction model focused on delay probability rather than booking choice.
- **Flight Price Analysis** — Adds historical price-quartile context alongside the choice probability score.

## FAQ

### What authentication does the Flight Choice Prediction API use?

The API requires an Amadeus Self-Service OAuth 2.0 access token obtained via the client_credentials flow against https://api.amadeus.com/v1/security/oauth2/token, then passed as a Bearer token on the prediction call. Through Jentic, the Amadeus client_id and client_secret are stored in the encrypted vault and the access token is fetched and refreshed automatically, so the agent only sees a scoped tool reference.

### Can I score flight offers I did not get from Amadeus with this API?

No. The /shopping/flight-offers/prediction endpoint expects the exact JSON shape returned by Amadeus Flight Offers Search, including its internal offer ids, segments, and travelerPricings. Pass the search response straight into the prediction call unmodified.

### What are the rate limits for the Flight Choice Prediction API?

Amadeus Self-Service applies tiered limits per API key. The Test environment is capped at 10 transactions per second per endpoint and a fixed monthly quota; Production limits depend on your Self-Service plan and are listed in the Amadeus developer portal. Bearer tokens themselves expire after roughly 30 minutes.

### How do I score a Flight Offers Search response through Jentic?

Run pip install jentic, then call jentic.search('predict the best Amadeus flight offer'), load the returned operation, and execute it with the body field set to the JSON returned by your earlier Flight Offers Search call. Jentic handles the OAuth token exchange and POSTs to /shopping/flight-offers/prediction on your behalf.

### Is the Flight Choice Prediction API free?

The Test environment is free with limited monthly quota. Production access is metered per transaction under an Amadeus Self-Service plan; pricing is published on the Amadeus for Developers portal and depends on monthly volume.

### Does the prediction endpoint change the flight offer payload?

No. It returns the same offers you sent in, with an added probability value per offer. You can pass any selected offer directly to Flight Offers Price or Flight Create Orders without translation.

### Can I limit what my agent is allowed to do with the Flight Choice Prediction API?

Yes. Because your Jentic One instance is self-hosted, you write the rules that decide which operations and credentials the agent may use. Since this API scopes by operation, you can allow only the flight-offers prediction call that scores a set of offers with a choice probability, and the agent cannot reach any search, pricing, or booking operation unless you add it. Your Amadeus client credentials stay with your instance, so the agent only ever sees the scoped tool you granted it.