Zerolook raises €1.6M pre-Seed to cut the cost of AI-scale flight search

Zerolook (Zug) has raised €1.6 million / CHF 1.5 million / $1.9 million in a pre-Seed round led by Playfair, with Vento Ventures, TrueSight Ventures, Alpha Venture, and angels from technology, travel, and financial services participating. The company was co-founded in 2025 by Simone Lini (CEO), a 2× founder and former Head of Partnerships for Google Travel and later Chief Commercial and Product Officer at Lastminute.com, and George Hadjiyiannis (CTO), former VP of Product at Kayak with an MIT Computer Science PhD and an engineering background spanning Google, NVIDIA, and eBay.

The look-to-book problem

Flight shopping has always generated more searches than bookings. The ratio — look-to-book — was 1:50 in the travel-agent era, 1:500 when OTAs moved purchasing online, and 1:10,000 when metasearch platforms began fanning a single query out to multiple OTAs simultaneously. AI travel agents have now pushed it past 1:200,000: one user asking "cheapest Greek island in June" generates queries across every island, every date, and every routing combination, in parallel.

At those volumes, computing every fare from scratch — which is how the current infrastructure works, querying a GDS or an airline system in real time for each combination — stops paying for itself. Airlines and travel agencies pay per query; they only earn revenue when someone books. The economics no longer close.

Zerolook's answer is to stop computing and start predicting. Its ML model, trained on historical fare data, returns estimated itineraries and prices in milliseconds without touching a GDS. A confidence score attached to each result lets sellers calibrate: exploratory traffic — the browsing phase that dominates AI-agent output — gets served cheaply; booking-intent queries can still route to live systems. "There is no technical reason that this has to be a problem holding back the entire industry. It is the result of legacy decisions made decades ago," said Hadjiyiannis. "We plan to modernise that infrastructure without requiring an overnight change in how the entire industry operates."

Playfair's third ProYarn-tracked bet in 2026

Playfair's lead position carries a pattern worth naming. The fund previously backed Astral Systems — Bristol's fusion-reactor company that closed a £23 million Series A earlier in 2026 — and Exclaim Robotics, Zürich's autonomous maintenance robot for AI data centres, which raised €4.29 million pre-Seed in August. Astral and Exclaim are deeptech bets on physical infrastructure; Zerolook is a B2B software wedge into a distribution market where a structural cost shift is measurably underway. What connects all three is less sector than timing: Playfair is backing founders solving a specific, bounded technical problem at the moment its cost implications have become unavoidable.

The founders' credentials are unusually direct for a pre-Seed round. Lini led Google Flights' OTA and metasearch partnerships across EMEA, built the Things to Do vertical from inception, then founded Waynaut — acquired by Lastminute.com — and served as its CCO/CPO. Hadjiyiannis was VP of Product at Kayak, one of the world's largest flight metasearch platforms. Between them they have been, at various points, on the airline side, the metasearch side, and the OTA side of the exact cost problem Zerolook is building against.

What the capital funds and what comes next

The pre-Seed round funds two things: the launch of Zerolook API v1 and the build-out of an ML team in Zurich. Signed letters of intent with "a group of online travel partners" are the only traction signal disclosed at this stage. That is an honest disclosure — letters of intent are expressions of intent, not revenue — and the next meaningful proof point is named customers generating live API traffic.

The model's accuracy claim rests on an assumption that fare logic is relatively stable day-to-day: advance-purchase rules, weekend-stay requirements, and route restrictions change infrequently, so a model trained on historical data should capture most of what matters for exploratory queries. That assumption holds for standard economy routes. It is less reliable for complex itineraries with tight fare classes, for routes where inventory changes rapidly, or for markets with non-standard carrier pricing behaviour. Whether Zerolook's "sample, compare, correct" feedback loop corrects fast enough across those edge cases is the technical question the next 18 months will answer.