Spoke
Your Money or Your Life
Topics where bad information costs somebody their money, health, housing or legal position get held to a higher bar. If you publish in one, the review process isn't overhead. It's the product.
The YMYL categories and the higher standard applied to them are set out in Google's published Search Quality Rater Guidelines.
Most content, if it's wrong, wastes somebody's time.
Some content, if it's wrong, costs them money, health, housing or a legal right. Google's guidelines draw that line explicitly and tell raters to apply a higher standard on the second side of it.
What makes content YMYL?
Content is YMYL when being wrong about it could seriously damage somebody's health, finances, safety or legal position. Money, medicine, law, housing, employment and civic information are the usual categories, and the test is the potential consequence rather than the label on the subject.
For a law firm, essentially everything substantive you publish is YMYL. That isn't a reason to publish less. It's a reason to publish through a different process.
What a trustworthy YMYL page has
- Somebody named with a credential that fits this subject. A credential that impresses in general isn't a credential in the thing being claimed.
- Somebody who checked it who didn't write it, and who knows the subject.
- Citations to primary sources — the statute, the regulation, the platform's own documentation. Not another blog post citing another blog post.
- A visible last-checked date. On a fast-moving legal subject, an undated page is an unreliable page however good it was when written.
- Explicit limits. Which state, which period, what it doesn't cover.
- A disclaimer that means something, where a reader sees it before relying on anything.
Where AI fits
AI is a legitimate tool for this work — outlines, first drafts, summarizing source material, headings, metadata. The failure mode isn't using it. The failure mode is shipping output nobody competent has read.
Two things about generated text make that acute here. It's fluent whether or not it's right, so errors read exactly like correct statements. And it reproduces the consensus of its training data, which on a state-specific legal question is frequently the consensus of marketing copy written by people with the same incentive you have.
The gate that works is a human who knows the subject, reading for correctness rather than tone, and signing off by name. The signing is the part that makes it real.
Staleness is a form of wrong
On ordinary content, out of date is untidy. On YMYL content, out of date is wrong — a rule that changed, a threshold that moved, a procedure nobody uses anymore.
So budget for re-checking instead of waiting for somebody to notice. We give every page a next-check date and our build stops when one goes by. Blunt mechanism. It also works in a quarter when nobody has time, which is the only quarter that matters.
As a process
- Work out which of your content is YMYL. For most professional firms the honest answer is nearly all of it.
- Name the reviewer for each subject area, in advance, in writing.
- Make the sign-off a gate, not a courtesy. Nothing publishes without it.
- Publish an editorial policy saying how you use AI and where the human check sits. Link it from every page.
- Schedule the re-review and give it a budget line.
More in E-E-A-T & YMYL
What E-E-A-T is
It comes from the instructions Google gives the people it pays to grade search results. Knowing where it comes from explains both why it matters and why most advice about raising it is confused.
The Experience signal
The first E is the newest and the most neglected. It's also the only one you can't buy, study for, or fake — which is exactly why it's worth building content around.
Reputation & reviews
Reviews are among the strongest trust signals available and among the easiest to acquire illegally. The FTC rule on consumer reviews is in force, and several widespread marketing practices fall inside it.
Human vs AI content
The most argued-about question in search right now, split into the part Google has put in writing, the part people assert with no evidence, and the part nobody outside the AI companies can answer.
AI content workflow
How to use AI for the parts it's good at, keep it away from the parts it's bad at, and put a human check in the one place that matters.
Author bios
A byline says somebody stands behind this. That only means something if they're real, the credential fits the subject, and the same identity shows up everywhere else.
Backlinks
A link is somebody deciding your page was worth sending a reader to. Everything useful about links follows from that, and so does everything wrong with trying to manufacture them.
Mentions & citations
Getting named with no link attached is now a common kind of visibility. What it proves is clear. What any AI engine does with it is not, and this page keeps the two apart.
User engagement
These are the most over-interpreted numbers in search marketing. Worth watching, and they don't mean what your dashboard implies they mean.