Spoke
Content depth and search intent
Depth is not length. It is whether somebody who arrived with a question can leave without needing another page — and intent is the thing most content gets wrong before a word is written.
That there is no ideal page length is stated directly in Google's guidance on generative AI optimization, linked below. The intent framework is practitioner convention.
Two failures account for most underperforming content, and neither is a technical problem. The page answers a different question than the one asked, or it answers the right question halfway.
How long should a page be?
There is no ideal length, and Google states this explicitly in its guidance. The right length is whatever covers the question properly — which for a definition is a paragraph and for a procedure might be two thousand words.
Word-count targets persist because they are easy to measure and easy to sell. They correlate with depth only incidentally, and padding a thin page to hit one produces a longer thin page.
Intent comes first
Before length, before structure, before anything: what did the person actually want?
The useful split is coarse:
- Know — they want a fact or an explanation.
- Do — they want to accomplish something.
- Choose — they are comparing and want to decide.
- Go — they want a specific place or page.
The common failure is answering a choose query with a know page. Somebody comparing two options gets an explanation of what the options are, which is accurate, complete, and useless to them. No amount of additional writing fixes it, because the problem is that the page is about something else.
Check intent by looking at what currently gets served for the query. If every result is a comparison table and you have written an essay, the question is not whether your essay is good.
Depth, properly defined
A page has depth when a reader with that question can leave without needing another source.
Practically, that usually means: the direct answer, the condition under which it differs, the thing that goes wrong, what it costs or takes, and what to do next. Most thin content has the first and none of the rest.
It also means covering the adjacent questions a reader will have because of your answer. This is not padding — Google documents that its AI features may fan a question out into several related queries, so adjacent coverage is additional retrieval surface as well as a better page.
What depth is not
Not length. A long page that circles one point is thin content with a word count.
Not comprehensiveness about the topic. You are covering a question, not everything about a subject. Everything about a subject is a cluster, which is a different structure with its own page per question.
Not keyword coverage. Listing every related term is the old version of this mistake wearing newer clothes.
The honest test
Give the page to somebody who actually had the question and watch whether they go and search for something else afterward. What they search for next is the part you left out.
That is a slower feedback loop than a content score, and it measures the thing the content score is trying to approximate.
More in On-page & content
Keyword placement
The most superstition-heavy topic in the field. Here's the small amount that holds up, and the large amount that doesn't.
Headings & structure
Most people pick heading levels by how big they render. They're a document outline, and the difference shows up for exactly the readers you care about.
Internal linking
Isolated pages chasing isolated keywords lost their value slowly and then all at once. What replaced them is depth, and the mechanism is your internal links.
Titles & descriptions
These are the only part of your page most people will ever see. A decade of advice about character counts has crowded out the one question that matters.
Readability & formatting
Readability tools count syllables and sentence lengths and hand you a grade level. Writing to that number produces prose that scores well and reads badly.