1. 1. Read the article

   We fetch the article, isolate the main content (tables stay tables) and break it into single, checkable statements – each with its original quote.

2. 2. Find sources

   For every statement we search for evidence, preferably on the official page of the provider named. Sources already linked in the article are checked first.

3. 3. Read the page ourselves

   We fetch every candidate page and look for the matching passage. Only a passage we really read can justify a change – never a search snippet.

4. 4. Judge and report

   The verdict comes from the article value and the evidence passage. Every reported change is verified a second time against the same source. Anything not properly evidenced goes to “To review” instead of becoming an update.

The score starts at 100 and subtracts: 3 points per change found, 5 points per dead or offline source, 1 point per open review.

From 80 an article counts as current, 50 to 79 means updates are due, below 50 it is urgent.

• Regional pricing & target country

  Many providers show prices depending on the visitor's country. That is why we fetch pricing pages from the project's target country (configurable per project). If that is not possible or the currency differs, we report “To review” instead of a change – a US price is the wrong truth for a German guide.

• Pages behind a login or bot protection

  Some providers block automated access (403) or require a login. We cannot evidence those statements; they appear as one bundled notice with a link to the provider's page.

• Content that only appears via JavaScript

  We read the delivered HTML but do not execute JavaScript. Prices that are loaded in the browser afterwards stay invisible to us – the fact then remains unevidenced.

• Multilingual websites of one provider

  If a provider offers the same page in several languages, we pick the version matching the project's language and target country. If no matching version exists, the source may be another language version.

• Scan duration

  The first scan of an article takes several minutes depending on its length – all facts are extracted and evidenced individually. Later scans are much faster because unchanged pages are detected and known sources reused.