Methodology

How I verify every fact.

This site exists to say Chile is bacán with numbers instead of adjectives. That is only worth something if the numbers survive being checked. Here is the whole method: where a fact has to come from, what blocks it and what happens when a bad one gets through anyway.

Including the uncomfortable part: below is the record of the mistakes I have published and corrected, with the wrong sentence exactly as it read.

01 Specific document
The source reaches the report, dataset, or paper behind the figure, never just an organization's front page.
02 Gate before publishing
One unsupported figure sends the whole article back. The checklist blocks; it does not advise.
03 Public correction
If a better source appears or a figure changes, it is corrected and the mistake stays visible.

How a fact travels before it ships

Four stages, in order, and none of them can skip the one before.

Before those four there is a binary gate. If a claim does not have the URL of a specific document backing it, it does not enter the process at all. It does not score low: it does not enter. A spectacular fact with no document is worth exactly the same as an invented one, because from the outside you cannot tell them apart.

Research

I look for the document from the body that produced the number. If a strong fact has no primary source, the search is redirected to the specific regulator or ministry for that sector, instead of settling for whatever a search engine returns.
Editorial decision

The angle gets chosen and facts get rejected, with the reason written down. Out go anything without a good enough source, anything that depends on political news cycles, and anything that would turn a private brand into the lead of an achievement that belongs to the country.
Writing

The piece is written with the approved facts and nothing else. This stage cannot add new numbers: if a better one shows up halfway, it goes back to stage 1 and gets verified from scratch.
Validation

A blocking checklist reviews the whole piece: sources, charts, accessibility, style and voice. It is not a suggestion. If a single item fails, the piece goes back whole to the previous stage and does not ship until it passes clean.

Those four stages are run by a team of AI models with different jobs, and I do not hide it: even this site's code was built that way. What matters is where control sits. Stage 4 is a filter and not an opinion, and it bounces the entire piece over one source that does not qualify. At the end I check it myself, with the document open beside me. AI accelerates. Judgment decides.

What counts as a source

In order of preference. If the first one exists, the second is not used.

The document from the body that produced the data. Chile's INE, Central Bank, Subtel, DGA, CONAF, UNESCO, NASA. Linked to the specific document, never to the organization's front page.
A peer-reviewed paper. When the fact is scientific rather than administrative, this is where it lives, and it has usually been published for years while you look for it in the wrong place.

And what does not count as a final source, no matter how high it ranks:

The press, specialized press included. It is a way to reach the document. If an outlet quotes a figure without linking where it came from, I go find it myself.
Wikipedia. A good starting point, a bad finish line.
A trade body or a brand talking about itself, or about the industry it represents.
An organization's front page, even when the organization is impeccable. If the link does not reach the document, it cannot be checked.

If a fact does not have that backing, it does not get published no matter how incredible it sounds. Especially if it sounds incredible.

The rules, applied

Three published articles where you can check the work yourself.

Trace the source to its origin 1,240, 1,247 and 1,215 kilometres of Carretera Austral Three different figures for the same road because they measure three different things: the declared scenic route, the total including Route X-91 and the kilometres already built. Each with its own document, instead of one round number.
If it cannot be verified, it is not published 99.33% of polling tables counted in under two hours It is exactly the kind of figure that sounds too good to be true, so it comes with the official report from the body that produced it, not a news story repeating it.
Mistakes are not hidden The Chilean passport: from 13th with 175 destinations to 14th with 174 A new edition of the ranking came out and Chile dropped one place. The article was updated with the worse figure; the older, better-looking one was not left in.

When I get it wrong

The full record, with the wrong sentence exactly as it was published. Without this, promising that I correct things would just be a nice line.

Seven cases. Six were live and got corrected; one was caught by the filter before it shipped. I found all of them auditing the site myself, none were reported by a reader, and that is exactly what I want to change: if you find one, write to me.

Was live Map card · Antofagasta · July 2026

I credited NASA with something NASA does not do there. And my first correction was wrong too.

It said “So much like another planet that NASA uses it.” Cited source: NASA / Sernatur. First correction, also wrong “The floor of an ancient salt lake between the Andes and the Domeyko range.” Cited source: National Monuments Council. It says today “13,200 hectares protected as a Nature Sanctuary since 1982, inside Los Flamencos National Reserve. The National Monuments Council lists it as geological and geomorphological heritage.”

The NASA project that does happen in Atacama is ARADS, and it happens in the hyper-arid Yungay area, not in the Valley of the Moon. I had two separate cards and lent one the other's credit. I corrected it, and the correction claimed the place was an ancient salt lake floor while crediting the National Monuments Council, which does not say that either. Only the third version rests on what the body actually states.

What I learned: correcting a borrowed claim does not guarantee the one replacing it is verified. Both times I made the same underlying mistake, which was writing first and looking for the backing afterwards.

Was live Map card · Maule · July 2026

I blamed the wrong plague, and leaned on a trade body instead of a paper.

It said “Wiped out in Europe by a plague, it survived hidden in Chile and is now the country's flagship grape.” Cited source: Wines of Chile. It says today “Nearly extinct in France after phylloxera, it survived in Chile mistaken for Merlot for 140 years, and is now the country's flagship grape.” Cited source: AJEV / G3.

The paper documenting the case does not say phylloxera wiped it out: it says it became nearly extinct in France because of poor fruit set and late ripening. The plague destroyed the vineyards, but what kept it from being replanted was agronomic. The cited source was also an industry association talking about its own product, which is exactly what the hierarchy above does not accept.

The sources that stayed are two peer-reviewed papers: Hinrichsen et al. (American Journal of Enology and Viticulture, 2001), which is the DNA evidence and is signed by the very ampelographer who made the discovery, and Minio et al. (G3, 2019), which documents the arrival in Chile in the 1850s.

Was live Map card · La Araucanía · July 2026

I stated something the archaeological evidence has left in dispute.

It said “Never conquered by the Incas.” It says today “They resisted the Spanish empire for over 300 years, until the crown recognized the Biobío border at the Parliament of Quilín (1641).”

Here a better URL was not enough, the text had to change. The original claim is disputed among specialists, so the card now rests on a documented fact with a date and a name, which is also more specific and more interesting than what it replaced.

Was live Map card · Atacama · July 2026

I cited a TV channel as a source.

Cited source History / NASA Cited source today NASA (ARADS)

The card's fact was correct and has not changed. What was wrong was the backing: History Channel is not an acceptable source under my own rules, and having it there made the card uncheckable even though it was true. It now points to the NASA program that does happen in the desert.

Was live Map card · Los Lagos · July 2026

The cited source did not support the number on the card.

Cited source Sernatur Cited source today DGA, MOP

The card states that Lake Llanquihue is Chile's second largest. That is a hydrographic fact produced by the water authority, not by the tourism agency. Citing Sernatur was not exactly false, it was worse: it was citing someone with no reason to know.

Was live Article · July 2026

A new edition of the ranking came out and Chile dropped. I swapped in the worse figure.

It said 13th in the world, with 175 visa-free destinations. It says today 14th in the world, with 174 visa-free destinations.

This is the most boring case on the list and that is exactly why I keep it: nobody would have noticed. A feel-good number that gets worse is precisely the one you do not feel like updating, and it is the test of whether the rule works or is decorative. It was also changed in the welcome email, which repeated the old figure.

Caught before publishing Instagram piece · July 2026

I nearly published that Gabriela Mistral was Latin America's first Nobel laureate.

It was going to say “Latin America's first Nobel was not García Márquez.” What shipped “Latin America's first Nobel Prize in Literature...” and only then “It was not García Márquez.”

The original version is false. The region's first Nobel was Carlos Saavedra Lamas, Argentine, Nobel Peace Prize in 1936, nine years before Mistral. She is Latin America's first in Literature, which is a different fact and still enormous. One word was enough to move the sentence from false to true, and that word came from the filter, not from me.

The uncomfortable part: the false sentence was a better hook than the true one. That is the kind of decision where the method earns its keep.

Found a figure that does not add up, a better source or an outdated fact? Write to me at hola@somosbacanes.cl. It gets corrected and written down here, with your finding. You do not need to be an expert on the subject: having the document is enough.

If you want to know who is behind all this, the story is in The story.