Lowest: harmful pages

What the document says

Section 4.0 is the longest in the document — and that’s no accident. The text itself admits it: identifying Lowest and Low pages is the hardest part of the rating job, so the guidance is more extensive. Lowest pages are the untrustworthy, deceptive ones, harmful to people or society, organized in three big overlapping families: harmful, untrustworthy and spammy. This page covers the first family; the next one covers the other two.

Before the categories, the document lists triggers that alone justify Lowest: a hacked or spammed page; gibberish; MC made with so little effort, originality or skill that the page fails its purpose; an extremely misleading, shocking or exaggerated title; MC deliberately obstructed by ads; a total absence of information about who’s responsible on a YMYL page; a very negative reputation; a lack of E-E-A-T that prevents the page from achieving its purpose.

Harmful to self or other individuals (4.2)

Lowest for content that encourages, depicts, incites or directly causes physical, mental, emotional or financial harm. The vulnerability criterion matters: the page doesn’t have to be harmful to everyone — a reasonable possibility of harm to the most vulnerable is enough. And there’s a clear boundary on the other side: reporting, discussing or informing about harmful things with a beneficial purpose is not harm. A violent action movie, news about violence, a page explaining scams to raise awareness, a portrayal of a dangerous activity that discourages imitation — none of that is Lowest. Health advice that contradicts expert consensus and could keep someone from a life-saving treatment — that is.

Harmful to specified groups (4.3)

Lowest for pages that promote, condone or incite hatred against groups defined by age, caste, disability, ethnicity, gender identity, immigration status, nationality, race, religion, sex, sexual orientation, veteran status, victims of violent events — or any characteristic tied to systemic discrimination. Two safeguards: the tone must be serious or mean-spirited (comedy and artistic expression don’t count unless clearly malicious), and criticism of ideas is not an attack on people — criticizing a religious doctrine is not harm to the group that follows it. Historical documentaries, newspaper articles about hate organizations, the dictionary definition of a slur: all out of scope.

Harmfully misleading information (4.4)

Three types require Lowest: harmful, clearly inaccurate information (refutable by widely accepted facts — a false election date, a false report of a leader’s death); harmful claims contradicting well-established expert consensus (“lemons cure cancer”); and theories grounded in no reasonable evidence, especially ones that erode trust in public institutions. A crucial detail: intent doesn’t matter — it applies both to deliberate liars and to people who sincerely believe the inaccurate thing they’re sharing. And the scale has well-drawn limits on the other side: fiction, satire, astrology, folklore, opinions in reviews, legitimately debatable questions with no single established answer, and trivial errors are not harmfully misleading. One-sided, polarizing or controversial content isn’t either — unless it crosses the line into harmful inaccuracy.

In practice

The harm scale doesn’t respect authority. The document says it twice: harmfully misleading information must be rated Lowest “regardless of the source,” including seemingly “expert,” “authoritative” or “official” sites. For YMYL SEO the lesson cuts both ways: domain authority is no shield for factually loose content, and the reverse holds too — a rigorously accurate small site isn’t doomed by the big ones.

The title is a Lowest trigger on its own. “Extremely misleading, shocking or exaggerated” sits in the table of direct justifications. The clickbait headline your social team loves can cost the whole page’s evaluation. The honest test: does someone arriving through the title find what it promised?

Health and finance are the obvious minefields — but financial harm is the underrated one. “Harm” explicitly includes financial harm. Content about investing, debt, side income and crypto that promises guaranteed results or contradicts basic financial consensus (“lottery tickets as retirement savings,” in the document’s example) sits in the same category as medical misinformation.

The content-about-harm vs. harmful-content distinction protects journalism and education — use it. You can cover scams, violence, eating disorders and drugs with an informative, educational purpose at no risk, as long as the framing is unmistakably to warn, inform or discourage. The framing is what separates the page that helps from the page that sinks.