From filters to dialogue: how chat search makes trip planning easier
In short: instead of dozens of checkboxes - a conversation in plain language. Voyager understands the conditions, sets priorities, analyzes reviews and shows not a "ton" of options, but top results with scoring and explanations.
How we are used to searching - and where the difficulties arise
Filters and catalogs work great when the criteria are simple: "date, city, 2 guests." But as soon as context appears - "need a warm pool, quiet in the evenings, walking distance to the tram, no overnight flights" - the usual approach forces you to jump between tabs, read reviews manually, and check details.
We are not "fighting" with filters - they are also present in Voyager. But for complex tasks it is more convenient to say it as it is, and the assistant will understand everything, clarify, and offer a short list of justified solutions.
How Voyager works
1) Conversation instead of checkboxes
Write naturally: "Hotel by the sea, warm pool, child 1 year, quiet in the evenings, up to 60,000 per night." The assistant translates intent into a set of conditions.
2) Hard and soft conditions
- Hard (mandatory): dates, budget, "no overnight flights," "airport IST."
- Soft (preferable): "quiet," "warm pool," "sea view," "to the tram ≤ 10 minutes."
If priorities conflict, it clarifies: "What is more important - quiet or sea view?" and builds a vector of preferences.
3) Review analysis and fact-checking details
The assistant summarizes reviews on key topics (noise, cleanliness, pool heating, deposit, transport), highlights risks: "Frequent complaints about noise on weekends," "Pool heating - from November." This saves time on "manually reading a thousand comments."
4) Scoring and explainability
Instead of 100+ cards - top 3-5 options with a matching score based on your priorities. Next to it - why this option is higher:
- ✓ Quiet (high)
- ✓ Heated pool (year-round)
- ~ To the tram 12-15 minutes on foot
- ✕ No rooms with direct sea view in budget
If a condition is relaxed - this is clearly shown.
5) Combined trips
The assistant connects air + train + bus/transfer, considers realistic transfer times and offers scenarios with different risk levels (low/medium/high) - along with advice like "better late check-out."
6) Privacy and control
You control what is saved. The dialogue history is under your control. We transparently explain which data we do not use.
Live scenarios (step by step)
Scenario 1. Family with a toddler, warm pool, 5-7 nights, quiet in the evenings
- You: "Hotel by the sea, warm pool, child 1 year, quiet in the evenings, 5-7 nights"
- Assistant: "What is more important - quiet or heated pool?"
- You: "Quiet"
- Result: 3 options with explanations and notes on reviews (for example, "pool is heated November-March," "rooms facing the courtyard - quieter"), scoring based on priorities.
Scenario 2. Weekend without overnight flights, only IST, direct flight
- Formulate the task in one sentence.
- The assistant excludes SAW, discards overnight flights, leaves only direct and warns: "Return flight is late; hotel with late check-out solves it."
- On the output - 2-4 routes with short explanations.
Scenario 3. Almaty → Cappadocia: plane + bus, daytime transfer in Istanbul
- The assistant offers 2 scenarios with different risk of delays, calculates time for transfer between the airport and bus station, and adds a checklist: documents, luggage, buffer time.
Why chat is more convenient for complex tasks
- Less noise. Instead of an endless stream of cards - top solutions with scoring.
- Clear compromises. It is visible which conditions are met and where a concession was made.
- Context is preserved. "Make quiet more important than view" - and the ranking is recalculated on the fly.
- Reviews - to the point. Not "thousands of words," but a summary of pros/cons.
Questions and answers
Can conditions be clarified during the process? Yes, this is a live dialogue. Change priorities and budget, add details.
Where does the data come from? We use partner and open sources + our own review analysis algorithms.
How many options do you show? Usually 3-5 - those that best match your priorities. More - upon request.
If I have a rare route? The assistant will gather a combined scenario: air + train/bus, with realistic transfers.