Skip to content

How Scout decides

Real places, ranked by fit, with the reasons shown.

What happens between the sentence you type and the places you get back — and the standards Scout holds itself to along the way.

1 · Reading the request

Your sentence becomes a brief.

Before it searches, Scout reads your request the way a good scout would: what you’re looking for, where, when, what it must have, what you’d prefer, and what you don’t want. You write it the way you’d tell a friend — no filters or keywords.

Looking for“a café to work from”
The kind of place, and what you’ll do there. The kind of place is the anchor: Scout looks for that type, not something that merely looks similar.
Location“Kyoto”
The city or area you name. Results stay there.
Timing“weekday mornings”
When you plan to go, when it matters — light, crowds and opening times change through the day.
Requirements“quiet, with outlets”
What a place must have to count as a match.
Preferences“good morning light”
What would make one match better than another.
Exclusions“not touristy, no chains”
What you don’t want. Excluded places are removed.

A follow-up like “quieter”, “cheaper” or “exclude rooftops” updates the same search instead of starting over, and keeps everything you already asked for.

2 · Requirements and preferences

Must-haves are held as must-haves.

The difference matters. A requirement narrows the search: a place that doesn’t meet it isn’t a match, however good it looks. A preference only changes the order among places that do. An exclusion removes places outright.

Limits you state are checked against what a listing shows. If you ask for a stay under a certain price, Scout compares it with the price the listing gives; when no price is listed, it says so rather than assuming.

3 · Evaluating places

Each place is read for your need.

Scout looks places up in public place and map data — place listings and images through SerpAPI, OpenStreetMap, Wikidata and Wikimedia Commons — so every result is a real, named place you can look up. Then it reads each one for what you said you need.

What “fits” depends on the request. For working, it’s quietness, seating, Wi-Fi, outlets and the best time to go. For a night out, atmosphere, noise and whether a group can book. For a shoot, light, access, permits and how crowds change through the day. The same café can be a good fit for one request and a poor one for another.

4 · Ranking and fit

Ranked by how well it fits — not by how popular it is.

  • Places that meet your requirements rank above places that don’t. A near-miss is never labelled a match.
  • Among real matches, your preferences decide the order.
  • Labels like “Best overall” or “Strong fit” appear only when a place clearly earns them on what you asked for.
  • Every result says why it fits, in terms of your request — not a generic description of the place.

5 · Evidence labels

Every claim says how sure it is.

Scout keeps three things apart: what a source establishes, what Scout infers from it, and what nobody has confirmed. Each read carries one of four labels.

Verified

A reliable source confirms it.

Supported

Evidence backs it, but no source settles it outright.

Inferred

Scout’s judgment from what it found — useful, not a fact.

Needs verification

Scout couldn’t confirm it. Check before you rely on it.

A judgment is never shown as a fact. A guessed permit status shown as a green light would be a promise Scout can’t keep — so it’s labelled for what it is.

6 · Fewer results

Why you sometimes get fewer places.

When only a few places genuinely fit, Scout returns those few rather than padding the list with near-misses. We would rather show three honest results than eight flattering ones. If the list is short, loosen a requirement or widen the area and search again.

7 · What Scout can’t confirm

Some things only the place can tell you.

  • Bookings, availability and live prices — ScoutHalo doesn’t take reservations or see a place’s calendar.
  • Permits and private access — Scout can say a permit is likely to matter; only the owner or the city can say yes.
  • Anything that changes day to day: opening hours on a holiday, an event that closes a terrace, today’s crowd.
  • Details no source mentions, like where the outlets are or whether a host welcomes a group of twelve.
  • Where a photo was taken. Visual Intelligence can suggest a possible place, but that is an unverified guess, never an identification.

These show up as “Needs verification” or in “Confirm before you go”, so you know what to check before you plan around a place.

How we keep it honest

Changes are tested against real requests.

We test changes to how Scout reads and ranks requests against fixed sets of example requests across everyday needs, travel and production, and we change one thing at a time so we can tell whether it helped. Scout is still early: if a result is wrong, tell us.

See example searches

What are you looking for?

Describe it in a sentence, and Scout finds real places that fit.