Skip to content
M

Editorial policy: automated checks and corrections

mind.io’s writing standards, the editorial rules its automated checks enforce, and how errors are fixed and recorded.

Editorial rules checked automatically

These rules are confirmed by automated checks that live in the site’s code. The unit tests are run when changes are made, and the search and licence checks run on every build. When a text breaks a rule, the check fails until the text is fixed. Each check covers its own set of pages; open a rule to see where it applies, what it checks and why.

Wording and tone

  • No verdicts on types or pairings
    What is checked
    A check fails if the Enneagram, attachment and love language pairing pages, the Enneagram wing and Holland code pages, the type explanations (other than MBTI), the friend compatibility card, or the zodiac animal compatibility, sun sign and blood type pages use words such as “destined”, “always”, “guaranteed”, “soulmates” or “worst match” (or their Korean, Japanese and French counterparts). Among them, the FAQs of the pairing and type pages must include several answers saying the outcome is not fixed and depends on the people and their effort.
    Why
    Types and pairings describe tendencies; they cannot decide how a relationship turns out. Verdict words make a result weigh more than it should.
  • Saju and fortune pages do not frighten
    What is checked
    A check fails if a day pillar, day master or saju term page, the zodiac animal compatibility pages, a sun sign page or a 2027 fortune page uses words of ill omen such as “doomed”, “unlucky”, “curse”, “will get sick” or “talisman”. Blood type pages may not link types to health, illness or diet either.
    Why
    Traditional and just-for-fun readings must not turn into words that feed anxiety or predict someone’s health.
  • No diagnostic labels in personality writing
    What is checked
    A check fails if the Enneagram, attachment and love language pairing pages, the Enneagram wing and Holland code pages, the type explanations (other than MBTI) or the Enneagram results use diagnostic or clinical terms such as “depression”, “panic attack”, “disorder”, “narcissist” or “psychopath”.
    Why
    Personality types are not a diagnostic tool, and a description that reads like a diagnosis is misleading.
  • No gender roles
    What is checked
    A check fails if the Enneagram, attachment and love language pairing pages, the Enneagram wing and Holland code pages, the type explanations (other than MBTI), or the day pillar, day master, saju term, zodiac animal compatibility, sun sign and blood type pages assume a gender or a gender role, with words such as “he”, “she”, “husband” or “wife” and their counterparts. The Teto and Egen label pages use gendered words only where they discuss the stereotype, say on every page that the labels can apply to people of any gender, and describe pairings by style, never by gender. Among the French texts, the love language and self-esteem results and the one-line MBTI descriptions are checked for not assuming the reader’s gender.
    Why
    Personality style does not follow gender, and gender stereotypes box readers into a narrow frame.
  • FAQs answer in the first sentence
    What is checked
    In some page families (the hubs, saju pages such as zodiac animal compatibility, blood types, Enneagram pairings and the Teto and Egen labels), the first sentence of an FAQ answer has to answer the question on its own. A check fails if it only points elsewhere, as in “see the table above”, or is too short.
    Why
    Search results often show an answer on its own, so the sentence right after the question has to be the answer.

Every page its own text

  • No sentences copied between pages
    What is checked
    A check fails if an Enneagram, attachment or love language pairing page, an Enneagram wing or Holland code page, a type explanation, a day pillar page or a Teto and Egen page shares an identical run of 22 to 30 Korean characters or more (7 to 9 words or 60 characters in English and French, depending on the family) with another page of its family. A sentence that only swaps the type numbers or names counts as the same sentence. MBTI pairing pages may not repeat a sentence from another pairing page. Stock phrases such as “chemistry” or “synergy” are capped at a few pages.
    Why
    Each pairing and type should get an explanation of its own, not a copy that only adds pages.
  • Guides are in-depth and in one voice
    What is checked
    A guide needs six to eight sections and at least 2,300 Korean characters, and its search description must be 80 to 150 Korean characters long. A check fails if the Korean text slips from the site’s friendly polite style into the formal one.
    Why
    A guide should be worth reading to the end, not a short summary, and the whole site should read in one voice.

Facts and numbers

  • Published numbers match the scoring code
    What is checked
    The score bands, cut-offs and item structure on the methodology page, and the question counts on the home and test pages, are recomputed from the actual scoring code and question lists. A check fails if any of them differs.
    Why
    If scoring changes and the explanation does not, the site ends up stating something false about itself.
  • Facts in the text are tied to the calculations
    What is checked
    Facts quoted in the guides, such as the example saju chart and its hidden stems, the samjae year table and the pairings often cited for each MBTI type, are compared with what the site’s calculation code and type data return. The Enneagram stress and growth directions stated on the pairing pages are checked against the type data too.
    Why
    When a calculation rule is fixed, no article should keep quoting the old value.
  • Help line numbers are managed in one place
    What is checked
    The crisis line numbers quoted in the break-up and attachment guides, and the domestic violence line numbers in the type pages’ safety note and the guides it points to, must match the site’s list of numbers. When a number on the list changes, a check fails until every text quoting it is updated. The burnout guides must name a local crisis line and say that a self-check cannot diagnose.
    Why
    If a number changes in one place and an old one stays elsewhere, someone could get the wrong number at the worst possible moment.
  • A changed guide prompts a new date
    What is checked
    If any text a reader sees in a guide changes, in any language (its title, description or body), a check fails and asks for the guide’s modified date to be updated. Changing only where a link points does not count as a change.
    Why
    Readers should be able to tell from the date how recent the content is.

Translation and naming

  • All four languages, nothing left out
    What is checked
    Every interface string must exist in the Korean, English, Japanese and French files, and a check fails if a Japanese or French string is still in English. Words that are meant to stay the same, such as brand names, are kept on a separate exception list. Korean left inside English, Japanese or French text, or a common French word missing its accent (é, è and so on) in the interface, also fails.
    Why
    Filling a missing translation with another language leaves a page its readers cannot read.
  • One concept, one name
    What is checked
    Guides, type pages, pairing pages and the friend compatibility card must call attachment styles, love languages and Big Five traits by the same names as the interface. In Korean, for example, Big Five neuroticism is always called “정서 민감성” (emotional sensitivity).
    Why
    If every page used a different name, the same result would seem to be telling different stories.

Privacy and accessibility

  • Depression and anxiety check-in results stay out of analytics
    What is checked
    Checks confirm that the depression (PHQ-9) and anxiety (GAD-7) self-checks send no start or completion events, and that their result pages send nothing to Google Analytics, not even a page view. The page where a check-in starts (with its questions) is recorded as a page view like any other page, but its address and title carry no answers or scores. The checks also confirm that ads are not loaded when these pages, results included, are opened directly.
    Why
    Depression and anxiety scores are sensitive health information and must not mix with visit statistics or advertising.
  • “Not sent” is said only of what is not sent
    What is checked
    Every sentence in the site copy, in all four languages, that says information is not sent to a server (such as “never sent to a server” or “nothing is sent”) must be a reviewed sentence or be about something the anonymous type ranking never sends: answers and scores, the mood check-ins such as depression and anxiety, Saju, Daily Fortune, or face and palm photos. A new or reworded sentence fails the check until it is reviewed.
    Why
    Joining the anonymous type ranking sends a result type to the server, so saying that nothing is sent is only honest about what truly stays put.
  • Text and answer choices keep a readable contrast
    What is checked
    Contrast is computed from the colour values: the common small gray text colour and the result fact labels need at least 4.5:1 against their background, and form field borders and answer dots at least 3:1. Field borders are checked in both the light and the dark theme. Only these colours are checked; it is not a scan of every page.
    Why
    Everyone should be able to read the text and pick an answer, whatever their eyesight or screen brightness (the WCAG accessibility guidelines).

Search details and licences

  • Korean titles stay within 45 characters
    What is checked
    A check fails if the title of a Korean page listed in search is longer than 45 characters, “ | mind.io” included.
    Why
    Search results cut Korean titles at about 40 characters, so the words people search for go first and the title stays short.
  • Pages offered to search engines are checked at build time
    What is checked
    When the site is built, every page in the sitemap is checked for prerendered content and heading (H1), a single canonical address, links between its language versions, a description and a share card image. The build stops if anything is missing.
    Why
    Search engines that do not run JavaScript should see the same content, and the language versions must point to each other correctly.
  • Licences of borrowed code are published
    What is checked
    The build fails if the open-source packages shipped in the site and the published notices file (/third-party-notices.txt) do not match, or if a licence outside the allowed list slips in. A check also confirms that the MIT notice of the kor-lunar table used for lunar dates appears identically in the code and in the notices file.
    Why
    Code written by others comes with conditions, and anyone should be able to see where it comes from.

How errors are corrected

Errors are found through the automated checks above and when texts are reviewed again. The site does not publish a contact channel yet; once there is one, it will be listed on the contact page. When an error is confirmed, this is how it is fixed.

  1. Fixes that change results are added, with their month, to “Changes that affected results” on the methodology page. For example, the Big Five emotional sensitivity items had been scored the wrong way round since the test launched in January 2026; this was fixed and logged there in October 2026.
  2. Saved results are shown with the corrected calculation when they are reopened, wherever they can be recalculated: Big Five emotional sensitivity scores and saju charts are corrected this way. Where a result cannot be recalculated, it is kept as it was, and the change log explains how to read older results.
  3. A check that would catch the same error again is added with the fix. The Big Five fix added a check of each item’s scoring direction, and the October 2026 accessibility pass added the contrast calculations above.
  4. When a guide’s text changes, the check described above asks for its modified date to be updated. When the privacy policy changes, its text and effective date are updated together.

The Baseline Writing Standard

Our default is to choose a useful sentence over a flashy one. We would rather write something clear, grounded, and a little modest than squeeze a result into a line that sounds clever but overstates the point.

That matters even more in self-understanding content. Instead of turning a score into a dramatic identity statement, we try to leave space for the reader’s own judgment, context, and lived experience.

Why We Push Back On Overconfident Wording

A tiny change in tone can make a self-understanding result feel much more final than it really is. That is why we are cautious with phrasing that sounds like “this is who you are” or “this always happens to you.”

Confidence in writing is not the same thing as accuracy. If a careful sentence is a better match for the evidence, we would rather sound measured than make the result feel more authoritative than it deserves.

How We Think About Translation And Localization

A translation that preserves the words but loses the tone can make the same page feel like a completely different product. So we care about local readability, not just literal accuracy.

That does not mean every language has to mirror the same sentence shape. The goal is to keep the same standards and boundaries while letting each language sound more natural, more welcoming, and easier to understand.