On this page
An article can pass every automatic check and still contain a mistake. That is not a paradox: the checks read the form of the text, not its meaning. A report that is all green makes the work look finished, even when the most important part is still to do.
The full production method is described in how an AndroidLab article is built. This piece looks at a smaller but useful point: what the automatic check does before publication and where it stops.
What the guard checks
The editorial guard is a script that stops with an error before publication. In its current code it checks seven things:
- the title stays under 60 characters;
- the excerpt stays under 155 characters;
- there are no emoji in the title, the excerpt or the body;
- in Italian texts, some accent errors written with an apostrophe instead of the accent are blocked, but the check uses a fixed list of twelve words;
- the title is not repeated as a heading in the body;
- there are at most three tags;
- the source block exists and contains at least one URL.
The accent check is a regular expression, which means a list of word forms written in the code. To understand how these rules work, the place to start is the Python re module documentation.
What it does not check
The guard cannot tell whether a fact is true. It does not distinguish a primary source from an article that repeats a press release. It does not verify whether a date is right, whether a download link leads to the announced version, or whether the Italian and English versions say the same thing. It does not judge whether the piece is actually useful to anyone.
The test
On 25 October 2026 we passed a few test sentences through the guard, without publishing anything. A wrong form on the list was blocked. A wrong form not on the list passed. A correct form, with the right accent, passed without issues.
The result is not spectacular, but it is useful. Italian words with a final accent are far more than twelve, so a fixed list covers only the cases already foreseen. The automatic check is a safety net for the most repetitive errors, not a guarantee about the text. That is why reading the full Italian text remains a human step, even when the guard is green. For correct forms, the Treccani dictionary is a reliable reference, for example its entry for perché.
What it changes for readers
Readers of AndroidLab do not see the guard report. They see the text. If a sentence is wrong, the mistake belongs to the editorial team, even if the automatic check let it through. The check is worth having when it reduces repetitive errors and frees time for what needs judgment: sources, figures, angle and usefulness.
In brief
- The guard checks lengths, emoji, listed accent errors, a repeated title, the number of tags and the presence of sources.
- It does not check truth, source freshness, source hierarchy, translation or download links.
- A wrong form outside the list passes without warning.
- A green report does not replace a full read.
For a broader human checklist, see also the Italian article AI Lab: checklist anti-brodo per articoli generati con AI.