The verdict — deep content, mis-routed links
Two different stories live in this module. Your content is genuinely deep — a median of 1,203 words, zero thin pages, and some of the most substantial writing in AI. But your internal links don't follow your best content: the pages that define Anthropic — the Responsible Scaling Policy, the Anthropic Institute, Core Views on AI Safety — are link-starved, reachable from just 1–3 other pages while product pages get full-nav treatment. Great content the site barely points to can't rank or get cited.
Content depth is maxed; the score is held down by link distribution — your important pages aren't fed.
The four drivers — transparent
what makes the scoreInternal-link distribution — a two-tier site
Every page averages 24.4 inbound internal links — but that average hides a split. Pages in the global navigation (homepage, research, products) are linked from essentially every page; pages reachable only through body content get a handful. There's no orphan problem (0 unreachable pages) — the problem is which pages get the equity.
The split
Nav-linked "hubs" pull ~30 inbound each; the starved tail sits at 1–3. Internal links are how you tell Google & the AI crawlers what matters — right now the site votes for products, not for its defining ideas.
Top hubs — where the links concentrate
outbound internal linksLink quality — are links near the top & semantically relevant?
real body links · nav/footer strippedNot all internal links are equal. A link near the top of the page, with a descriptive anchor that names the target's topic, passes far more value (to Google and to AI extractors) than a generic "read more" buried in a footer. Right now only 28% of your body links are near-top and 56% use generic anchors — more than half your internal-link equity is placed where it counts least. Rewriting anchors to be descriptive and moving key links up is a pure-formatting win.
Starved high-value content — the core finding
These are the pages that matter most to Anthropic's mission and rank lowest in internal-link support. Each is substantial, evergreen content — and each is reachable from only one or two other pages. This is the highest-leverage, zero-cost fix in the whole SEO suite: point your hubs at your best ideas.
Content depth — a real strength
Where many sites are thin, Anthropic is the opposite. Zero pages under 300 words, a median of 1,203, and a long tail of deep, authoritative pieces — including a 30,524-word policy document. The irony: the deepest content is often the least-linked.
Word-count distribution
pages by depthDeepest pages
words · inbound linksWatch the inbound column: the most substantial pages frequently carry the fewest internal links — the mismatch §07 plots.
Heading structure — the machine-readable outline
Headings are how crawlers (and AI extractors) parse a page's structure. Most pages are clean, but 10 have a structural issue — a missing or duplicate H1, or a skipped level (jumping H1 → H3) that breaks the outline.
Pages with heading issues
H1 count · level skips| Page | H1 | Skips | Total headings | Issue |
|---|
Readability — how dense is the writing?
Reading-grade level across the site averages 15 — solidly college-level. That fits a technical, policy-heavy audience, but the densest pages (legal, research) climb higher and could lose a general reader. Not a defect — a calibration point.
Reading-grade distribution
Flesch–Kincaid grade · higher = denserRead
A few very-high grades are legal/policy pages (the RSP, usage policy). For citation-readiness, a plain-language summary block at the top of those pages helps both readers and AI extractors.
Content vs links — the mismatch, plotted
This is the whole module in one picture. Each dot is a page: across = how much content it has, up = how many internal links point to it. A healthy site trends diagonally (more content → more links). Anthropic's most valuable pages sit bottom-right — deep content, almost no links. Those are the dots to drag upward.
Labelled red dots are your starved high-value pages — substantial content (right) with 1–2 inbound links (bottom). The fix in §09 moves them up.
The blog — a content engine the site under-links
We crawled all 230 /news/ posts (the page sample saw only 4). Anthropic publishes ~10 a month, every month, for two years — a serious content operation. The gap isn't output, it's connection: this firehose of content is barely linked together, and the links that do exist are mostly generic and low on the page.
The content engine
Publishing cadence
posts per monthBest-linked posts
near-top + semantic inboundSubstantial but starved
600+ words · ≤1 contextual link inAll 230 blog posts
sort any column · click a post for detail| Post | SEO | Published | Words | Links in | Near-top | Semantic |
|---|
The fix is §03 at blog scale: when you publish, link the new post from 2–3 related older posts with a descriptive anchor near the top, and replace "read more" blocks with real post titles. That turns 230 isolated posts into a navigable topical web that compounds authority across the whole blog.
Per-page — content & link signals, all pages
Every crawled page with its content depth, internal-link position, structure and schema. Sort by any column; click a row for the full breakdown.
| Page | SEO | Words | Inbound | Outbound | Headings | Grade | Schema | Link tier |
|---|
The fix — a deployable internal-link plan
No new content required. Just route existing link equity to the pages that earn it. Highest-leverage first; the projected lift takes Content & Architecture from 70 toward ~85.
Methodology & honest limits
—
Not collected here (needs an edge-capturing full crawl — a full-engagement add):