The Flight Shopping API,
built for AI-scale
search.
Flight shopping designed for the search volumes AI travel is generating.
- → A fraction of the cost per query. An ML model predicts prices instead of fetching them live — trading a small amount of accuracy for dramatically lower cost per search.
- → Right-sized for exploratory traffic. Built for the browsing and comparing queries that dominate today's search volume — where booking-grade precision isn't needed.
- → Serve metasearch and AI agents profitably. Make the traffic from price comparison sites and AI travel agents work for your unit economics, not against them.
Flight shopping was built for humans. AI search breaks the look-to-book math.
Look-to-book — the ratio of searches a shopping system handles to actual bookings — used to be 1:50 in the days of human travel agents. With OTAs it climbed to 1:500. Metasearch pushed it past 1:10,000. AI travel agents and modern comparison tools take it past 1:200,000. At those volumes, computing every fare from scratch on a GDS or airline system stops paying for itself.
Phone, terminal, one query at a time.
Self-serve. The form moves online; the question stays the same.
Comparison sites fan one query out to many OTAs. Cost climbs.
No form. No constraints. One trip becomes hundreds of thousands of searches.
Use the calculator below to plug in your own traffic and see where you land.
An AI agent planning a flexible Greek-islands week fans out across origins, destinations, and dates in parallel — one trip request becomes hundreds of thousands of searches. Look-to-book was already creaking under human traffic, and the ratio grows exponentially. Calculate yours.
One trip request. 216,000 searches.
“Show me flights from Milan to Greek islands in June.”
Origins the agent considers (e.g. MXP, BGY, LIN).
Destinations in scope (Greek islands, the Med, etc.).
e.g. 24 possible 7-day trips that fit inside a 30-day month.
Trip requests on your channel that turn into bookings. e.g. 1 means 1%.
Look-to-book = (airports × destinations × dates) ÷ conversion
These are alternative routes the booking can flow through — costs are mutually exclusive, not additive.
Airline ADM excess fees apply when L2B exceeds the contractual baseline on direct / NDC. Defaults follow the AFKL 2024 ADM Policy §3.8: 1 : 1,000 baseline, €0.70 per 1,000 excess queries. GDS / aggregator rates apply when search routes through a third-party; defaults reflect typical self-service tiers. The two paths are mutually exclusive per booking — adjust to model your own carrier or contract.
Unlock revenue locked behind low-conversion queries.
Total metasearch queries reaching your channel per month.
Share of queries you currently respond to. The rest are dropped because cost-to-serve exceeds expected value.
Searches per booking. e.g. 1 : 50,000 means one booking for every 50,000 metasearch queries.
Average gross booking value including ancillaries (bags, seats, insurance).
"Additional GTV" is the gross travel value of bookings recovered if currently-dropped queries were filled: unanswered queries ÷ L2B × avg gross value (incl. ancillaries). 1 : 50,000 means one booking per 50,000 queries.
Estimate the answer. Don't recompute it.
Instead of computing every fare combination in real time, our machine-learning model estimates the cheapest options — without sending a query to a GDS or directly to an airline.
We've trained the model on years of historical fare data. Given a route and dates, it predicts the cheapest itineraries in milliseconds — without computing every combination.
Weekend-stay requirements, advance-purchase rules, route restrictions — most fare logic is stable from one day to the next. If an itinerary worked yesterday, it almost certainly works today.
We regularly query real prices and availability and feed the results back into the model. Predictions self-correct continuously as fares move.
Built for OTAs and airlines. Works on any channel — from metasearch, to AI agents.
We help both on the two channels where the look-to-book math has stopped working: metasearch — including the long-tail queries that have always been too expensive to bid on — and AI travel assistants, where one user can generate hundreds of searches in a single conversation.
Serve the searches AI agents and metasearch sites send you without your shopping infrastructure scaling with traffic.
Show fares on metasearch sites and to AI travel agents without paying GDS fees on every search. Pay full distribution costs only when a customer actually books.
Get flight search and booking links with a single API call. A generous free quota to build, prototype, and ship — no GDS contract, no enterprise sales cycle.
- 01 Generous free quota — enough to develop, test, and run a side project without a credit card.
- 02 Deeplinks to major booking providers — point users straight to checkout on the OTAs and airlines they already trust.
- 03 REST or MCP — drop into any backend, or plug straight into Claude, ChatGPT, and other agent runtimes.
Search with us. Book the way you already do.
zerolook is a flight-shopping layer, not a distributor. When a customer commits to a booking, send them through whatever channel makes sense — NDC, GDS, your existing consolidator, a direct connect. Distribution fees apply only with your existing partners.
Predicted itineraries, prices, and confidence scores. This is the only step you pay for.
Route the booking by route, fare class, market, or commercial deal — exactly as you do today. zerolook stays out of the booking path.
Distribution fees apply only here, and only with your existing partners. You pay nothing to zerolook.
Talk to us about adding zerolook to your shopping stack.
Sandbox key, integration plan, or partnership conversation — we respond within one business day.