Our methodology

Every visa entry keeps the URL and collection date of its sources. Two families of sources coexist, and a badge tells them apart on visa pages: official government websites (“Verified”) and the visa policies documented on Wikipedia (“To confirm”). Descriptive texts (country descriptions, route guides) are written by a language model; the section “What is generated automatically” says which ones.

• The visa policies documented on Wikipedia, one page per destination country: they are the main source of our short-stay regimes (visa-free, visa on arrival, e-visa, consular visa) for every nationality. These pages cite official texts, but remain a secondary source.
• Corrections from official websites: when Wikipedia is wrong or slow to follow a change, the regime is corrected by hand from the relevant government’s website (about fifteen destinations so far, including the United States, the United Kingdom, South Korea, Thailand and Cape Verde). Short-stay visa fees for about fifty countries, including the whole Schengen area, also come from official fee schedules.
• Cross-checks against official lists: for the most visited destinations (Canada, the United Kingdom, Thailand, Japan, the United States…), the lists of visa-exempt, visa-on-arrival and e-visa nationalities published by the government are kept in the code and compared with the catalogue after every update, nationality by nationality. A route is marked “Verified” only if the official list settles its nationality and says the same thing; at the first disagreement, the destination goes back to “To confirm”. The page of a verified route links to the official page that shows its regime, where there is one.
• The websites of the foreign ministries and immigration services of 186 destinations, for other visa types (work, study, family, retirement, business, transit): fees, processing times and official links, collected one by one from those sites in October 2026. A fee without a readable official schedule is left out rather than guessed.
• The official e-visa portals of about twenty countries (India, Turkey, Vietnam, Saudi Arabia, Egypt, US ESTA, Canadian eTA, UK ETA, etc.), whose fees, processing times and eligible nationalities are taken from the portal itself.
• Schengen rules (the EU list of visa-exempt nationalities, the common visa fee) for the 29 Schengen states.
• An independent dataset, passport-index, used solely as a counter-check: it never changes our entries.

Several checks run without human intervention, at collection time and then on published entries:

• At collection: visa policies are re-read every Sunday. Each page goes through consistency checks (at least 180 nationalities covered, no nationality in two regimes at once, Schengen list, known reference cases). A page that fails does not replace the live data, and a new entry that fails stays in quarantine until a human reviews it.
• Weekly counter-check: every Monday, all published short-stay regimes are compared with the passport-index dataset. An abnormal rate of differences triggers an alert.
• Divergences between sources: every day, regimes are compared from one source to another (for example Wikipedia and a country’s e-visa portal), and transit and business fees against the Schengen reference fee. A gap flags the entry for review, without removing it from the site.
• Confidence status: data taken from an official source (ministry, government portal, EU rules) or checked by hand is marked “Verified”. Data sourced from Wikipedia is marked “To confirm”.
• Log: collector writes and manual interventions are recorded (source, date, fields changed). The latest visible changes, such as a new visa or a correction, are published on our updates page.

See the updates log

The site has several thousand pages. To be honest about what to expect from them:

With AI

• Finding official fees: the fee schedules of ministries and portals are read by AI agents (Claude), required to cite the official page they used. No amount is kept without a readable official schedule, and a fee that varies by nationality without a public schedule is not displayed.
• Country descriptions (overview, safety): written by a language model from public sources, without page-by-page review. The cost of living on country pages is an estimate, presented as such.
• Route guides (“visa for X from Y”): written by a language model from the catalogue data (verdict, fee, processing time, documents). Before publication, automated checks reject a guide that contradicts the catalogue verdict, claims a visa on arrival or an e-visa the catalogue does not know, or promises a longer stay than the one allowed. Every week, an audit compares each published guide with the catalogue and withdraws those that understate the visa requirement; every day, a check flags amounts missing from the catalogue. These guides are not reviewed one by one: each shows its generation date.

By a human

• The choice of sources and the rules for reading pages.
• Regime and fee corrections from official websites, and the removal of fees that cannot be verified.
• Reviewing alerts and quarantined entries.
• Corrections following reader reports.
• Approving the methodology, about and legal pages.

No correction is applied blindly: alerts from automated checks, quarantined entries and fees collected by AI agents are reviewed by a human before entering the catalogue. However, we do not review every page: route guides and country descriptions are published without individual review. That is why every page reminds you to confirm with the embassy before booking.

Spotted outdated or inaccurate information? Tell us: every report is checked against the official source and corrected when needed. It is the most useful contribution you can make to the reliability of this site.

Contact us