Editorial Policy

Our standards

We publish travel-safety information that we believe is useful to travelers planning a trip, on the day of arrival, or already in a destination. Everything we publish follows three rules:

1. Cite the source. Every live data point on an assessment page links back to the public source it came from — Open-Meteo for weather, the US State Department for advisories, GDACS for disaster alerts, the CDC for health notices, Google News for recent headlines, and the AI provider for its synthesized commentary. Every AI sentence cites the snippet ID it was built from.
2. Separate baseline from live. Our curated city baselines (crime, health, political-stability, transport) are clearly labeled as our editorial judgment. Live data is clearly labeled as such, with timestamps and source attribution.
3. Say what we don't know. Every city page has a "What is estimated" and a "What is unavailable" section. We do not invent data to fill space.

Who reviews what

Travel Risk Calculator is operated by an independent editorial team. The operator of record is BYO (contact: [email protected]). We are not affiliated with the US Department of State, the World Health Organization, the Centers for Disease Control and Prevention, any government, embassy, or police force.

Three review tiers apply to city pages:

• needs_review — the page exists in our database (often imported from SimpleMaps or auto-discovered from a user search) but has not been reviewed by a human. These pages carry a noindex meta tag and are excluded from our sitemap.
• reviewed — a human has confirmed the city exists, its name and coordinates are correct, and its baseline scoring is plausible. These pages are indexable.
• verified — a human has manually checked the city's threat scores against current public sources within the last 90 days. Verified pages appear with a "last reviewed" date.

Where AI fits

We use a large language model (currently gpt-oss:20b-cloud via Ollama Cloud) to synthesize the prose of the AI briefing on each assessment. The model:

• receives the live search snippets, the baseline scores, and the user-supplied trip details;
• is instructed in its system prompt to never invent facts, never recommend passport/visa/embassy steps for domestic trips, and always cite snippet IDs;
• has its output validated against the snippet IDs it cites — claims that reference snippets that don't exist are dropped before publication;
• does not write or edit the underlying threat scores, only the prose summary.

See our AI Content Policy for more detail.

How we choose what to publish

We do not publish a destination page for every entry in our 45,000-city database. Pages are promoted to indexable status only when they meet a quality threshold:

• the city is a national capital or has a population above 500,000 (initial cohort), or
• the city has been assessed at least 3 times by users with consistent AI grounding, or
• an editor has manually marked it as reviewed/verified.

This is intentional — we'd rather have 700 deep city pages than 45,000 thin ones.

What we will not publish

• Threat scores presented without the underlying category breakdown and confidence indicator.
• AI briefings whose claims fail snippet-ID grounding validation.
• Affiliate or sponsored content presented as editorial — paid placements (if any) are clearly labeled.
• Personally identifying information about users — assessments are anonymous unless saved to an account.
• Content that frames us as an emergency service or official authority. We are a planning tool.

Corrections

If we get something wrong, we want to know. See Corrections for how to report inaccurate data and our log of recent fixes.

Updates to this policy

This editorial policy was last updated on May 15, 2026. We will publish notice of material changes here.