The Tech Sector's Empty Link Graph: Built for Browsers, Not Agents

Ready on the Surface

Across ten Fortune 500 technology sites crawled in Q3 2026, totaling 9,038 pages and 16.2 million tokens of public content, the sector looks like the one most ready for machines. Schema markup is near-universal, JavaScript dependence is low, almost every page has a title and a description. These are companies that build software for a living, and their surfaces show it.

Underneath runs the same fault line the finance and healthcare cohorts revealed, and on the worst tech sites it runs deeper than either. On the nine sites where it can be measured, the share of content that nothing links to averages 32%, and it does not cluster around that mean. It ranges from 0.3% to 80%: a forty-fold spread in link density underneath, between sites that publish a similar number of pages. One site in the cohort is a near-empty tree, 312 pages joined by just 415 editorial links, where four pages in five are unreachable. Another is a hyperlinked mesh of nearly 56,000 links across 575 pages, where almost nothing is unreachable at all. They are in the same business. Their topologies are opposites.

The sector that writes the crawlers, in other words, has not wired its own content for them. The polish is real and it is on the surface. The structure beneath is as uneven as any cohort we have measured.

Why Nine Sites, Not Ten

Link-invisibility is a question about a site’s full inventory: of every page it publishes, how many does nothing link to. Answering it needs that inventory, which is the XML sitemap. Read the sitemap, compare the published set against the link graph, and the unreachable pages fall out. Without it the question is unanswerable, because an agent following links and a crawler reading the sitemap converge on the same visible pages.

One of the ten, a networking-hardware giant, returned no usable sitemap: its inventory could not be listed in full, so its invisible tier cannot be counted. We record it as not-measurable, not zero. Reachability and invisibility are therefore reported for the nine measurable sites; topology, fragility, content, and the scorecard use all ten, because those do not depend on the sitemap.

The Three-Tier Reachability Model

Every page falls into one of three reachability tiers, judged against the full link graph (navigation included) for reach and the content link graph for editorial authority:

1. Reachable via editorial links. An in-content link path leads to the page. These are the pages a link-following agent finds on its own.
2. Navigation-only. Reachable, but only through chrome: header, footer, or menu links that repeat site-wide. An agent that filters boilerplate, as most do, loses these.
3. Link-invisible. Zero inbound links of any kind. The page exists as a URL and sits in the sitemap, but nothing on the site advertises it to a link-following crawler.

Averaged across the nine measurable technology sites:

• Navigation-only (10.2%)

View data

Segment	Value
Reachable via editorial links	57.7%
Navigation-only	10.2%
Link-invisible	32.1%

Just under three fifths of published content, on average, is reachable the way an unaided agent reads a site. Here is how the link-invisible tier distributes across the nine:

View data

Site	Link-invisible page rate (% of site's pages)
T10	0%
T6	0.3%
T4	10.5%
T1	25.8%
T8	31.4%
T5	42.1%
T2	47.9%
T7	51.6%
T3	79.7%
Sample mean	32.1%

The spread is the story. Two sites bury essentially nothing, and one of them (T3) buries 80%: an enterprise-software property whose live navigation has stopped pointing at the bulk of its own catalog and documentation. The invisible pages are not degraded crawls. They carry real body content and were discovered through the sitemap. The pattern is the same one finance showed: a large archive published, then orphaned as the site moved on, surviving only in the sitemap. For an AI agent, that archive is dark.

The Navigation-Only Trap

The invisible tier is the loud failure. The navigation-only tier is the quiet one, and one site in this cohort shows it at a scale we have not seen. Site T10, a semiconductor brand, records 0% link-invisible: by the headline measure, perfect. But 74% of its pages are reachable only through navigation chrome, not a single editorial link. The site routes its entire catalog through the header and footer menu, and almost nothing links page-to-page in the body.

To a human in a browser, T10 looks immaculate; every page is one menu click away. To an AI agent that strips repeated navigation to find the actual content, as most do, three quarters of the site evaporates. The “0% invisible” headline is true and almost meaningless. T10 has simply moved its unreachable content from the invisible tier into the navigation-only tier, where a different class of agent loses it. Reachability is not one number; it is which kind of link does the reaching.

Machine-Readability: Reaching a Page Is Not Reading It

Reaching a page is one thing; parsing it is another. Two crawl signals decide whether an agent can read what it reaches.

The first is structured data: the JSON-LD and schema.org markup that states a page’s meaning in machine terms. Technology leads every cohort here, averaging 80% coverage, but the distribution is bimodal, from under 20% to a perfect 100%.

View data

Site	Pages carrying JSON-LD / schema.org structured data (%)
T2	19.8%
T6	34.3%
T9	60.8%
T1	90.6%
T3	96.3%
T4	96.8%
T5	98%
T8	99.6%
T10	99.9%
T7	100%
Sample mean	79.6%

But coverage hides composition. Sorting every schema.org annotation across the cohort by what it actually describes tells a sharper story than the headline:

Navigation	41.8%	WebPage, BreadcrumbList, ListItem
Identity	35.7%	Organization, ContactPoint, Person
Article / media	11.6%	Article, TechArticle, VideoObject
Answer (FAQ / Q&A)	6.1%	FAQPage, Question, Answer
Product / other	4.8%	Product, SoftwareApplication

Share of all schema.org annotations across the 10 sites, by type.

More than three quarters of all structured data is navigation and identity scaffolding: who the company is, where to click. Only 6.1% is answer-shaped, the FAQ and question-and-answer markup an AI assistant can lift straight into a reply. Two sites markup 100% of their pages and carry almost none of it. The exception is telling: one SaaS site (T8) puts nearly 30% of its schema into answer-shaped markup, an order of magnitude above the cohort, because its business depends on being quotable inside an AI assistant. The rest describe who they are to a machine, not what they can answer.

The second signal is render dependency: pages whose content appears only after a browser runs their JavaScript. Technology, despite being the sector most fluent in front-end frameworks, keeps this low, averaging 6.8%, far below finance’s 20%.

View data

Site	Pages whose content requires browser rendering (%)
T4	0%
T6	0%
T8	0%
T10	1%
T1	1.2%
T7	2.1%
T3	2.3%
T9	16.8%
T5	20.7%
T2	23.7%
Sample mean	6.8%

This is the one machine-readability metric technology clearly wins. Seven of ten sites render under 3% of their content client-side; even the worst, at 24%, is below the finance average. The sector that popularized the JavaScript single-page app serves its marketing content as static HTML. Whatever is wrong with these sites, it is not that an agent cannot read the pages it reaches. It is that it cannot reach them.

Fragility: The Whole Site Hangs on a Few Pages

The second pattern, concentration, holds across all ten sites, and at the extreme it is sharper than anything finance showed. Betweenness centrality counts how often a page sits on the shortest path between two others: a high-betweenness page is a bridge that traffic, link equity, and an agent’s traversal all funnel through. When that concentrates in a tiny fraction of pages, the site is fragile.

View data

Site	Share of total betweenness carried by the top 1% of pages (%)
T2	25.5%
T4	28.2%
T5	31.8%
T6	32.3%
T8	35.3%
T1	38.4%
T9	45.5%
T3	64.3%
T7	83.9%
T10	84.3%
Sample mean	47%

On the average site, the top 1% of pages carry 47% of all betweenness. The two most concentrated sites push 84% through 1% of their pages, where betweenness pools in corporate hubs (investor relations, governance, the about section) rather than in the product content. The entire structure routes through a handful of pages that have little to do with what the company sells. For a link-following agent, this compounds the reachability problem: its picture of “what this company does” is mediated almost entirely by a few corporate pages, and whatever sits in the invisible tail never enters the frame.

Health Metrics: Orphans Without Dead-Ends

The orphan rate, pages with no inbound editorial links, follows the now-familiar shape: a long climb from near-zero to severe.

View data

Site	Orphan rate (% of site's pages)
T6	0.3%
T9	2.8%
T4	15.3%
T8	20%
T2	24.5%
T1	25.7%
T3	35.9%
T5	43.6%
T7	55.1%
T10	74%
Sample mean	29.7%

Half the cohort sits below 25%, then the rate climbs steeply to 55% and 74% on the two sites that also carry the heaviest betweenness concentration. The same sites that funnel everything through a few corporate hubs are the sites that orphan the most product content: two faces of one habit, never linking page to page in the body. The cohort-wide dead-end rate, by contrast, is a low 10%: these sites are not bad at passing visitors onward from the pages they do expose. Their failure is upstream, at the point of linking in.

The Industry Scorecard

The five-lens analysis assigns each site a green, amber, or red score per lens. Across all ten:

Site	Skeleton ⓘ Size, density, and average path length. How big and how connected the site is at the body level.	Circulation ⓘ PageRank distribution and structural bottlenecks. How importance flows between pages, and which hubs hold it all together.	Organs ⓘ Community detection. Whether the site's topical clusters cleanly separate, or whether one mega-cluster dominates everything.	Health ⓘ Islands, orphans, and dead-ends. Where content is structurally dying: unreachable, unlinked, or terminating.	Nervous Sys. ⓘ Click depth, bridges, and cross-community linking. Whether the site is a well-designed building or a pile of disconnected rooms.
T1
T2
T3
T4
T5
T6
T7
T8
T9
T10

• Green: healthy
• Amber: moderate concern
• Red: critical

The pattern holds for a third sector running. Skeletons are mostly fine; Health is the worst lens by far. Seven of ten sites score red on Health, the lens that captures orphaning, and only two earn green. Circulation splits three ways, with three sites red on authority concentration. The consistent finding across healthcare, finance, and now technology is the same: F500 sites are well-built at the surface and broken at the level of how editorial signal travels between pages. Technology’s surface is the most polished of the three. Its Health column is no better.

The Density Spectrum

The technology cohort has no single shape. Link density spans more than forty-fold between the sparsest and densest site. Raw edge counts run from 415 to nearly 59,000, but volume is not density: the site with the most edges is larger and less tightly wired than the true mesh.

Pole	Edges	Behavior
Dense mesh	~56,000 on ~575 pages	Hyperlinked, shallow, near-zero invisibility
Sparse tree	415–4,400	Hierarchical or barely-linked, deep invisible tails

The densest site (T6) wires nearly everything to everything: 575 pages, roughly 56,000 links, and an invisibility rate of 0.3%. Connect every page to dozens of others and nothing gets orphaned, almost by accident. At the other pole sits T3, 312 pages joined by 415 links, where the editorial graph barely exists and 80% of content is unreachable. But mesh density is not the same as resilience: T6 still earns a red on community structure, because wiring everything to everything erases the topical boundaries an agent uses to understand a site. One pole hides its content; the other dissolves its meaning. Neither is the goal.

Content Quality at Scale

Across all ten sites combined: 9,038 pages, 16.2 million tokens of body text, and the highest per-page word count of any cohort to date.

Content metric	Sample mean (n=10)	Notes
Pages per site	904	Median 1,000, range 575–1,001
Avg word count per page	1,099	Highest of the three cohorts
Avg token count per page	~1,843	Character-based estimate (length / 4)
Pages with thin content (under 200 words)	14%	Roughly half of finance’s rate
Internal-link share of all links	76% internal	Sites keep most links on-domain
Title tag coverage	99.8%	Near-universal
Meta description coverage	88.3%	Healthy across the cohort

Technology writes more per page and pads less: thin content runs at 14%, roughly half the finance rate. The pages themselves are substantial, well-titled, and well-described. This is what makes the invisibility finding sharper here than anywhere else. These are not weak pages an agent can safely skip. They are long, structured, schema-rich documents that nothing links to. The sector with the best content has buried a third of it.

What Every Technology Site Shares

Where healthcare organized around the newsroom and finance around jurisdiction, the technology cohort organizes around the product line, and increasingly around AI. The recurring top-level sections are products, analytics, communications, and legal, but the section growing fastest is artificial-intelligence: on the two sites where it appears as a distinct path, it already absorbs 16% of all pages, the single largest content section on those sites.

This is the structural fingerprint of a sector mid-pivot. The AI sections are new, large, and, on the sites that have them, among the most invisible: freshly published, heavily linked from a launch announcement, and not yet woven into the body of the site. The companies racing hardest to be cited as AI authorities are publishing the content that proves it into the part of their own graph an AI agent is least able to reach.

What This Means for AI Search Readiness

For an AI agent that uses internal link structure to discover and rank content, three implications follow.

1. Polish is not reach. Technology has the best surface signals of any cohort: near-universal schema, the lowest render dependency, the longest and least-padded pages. None of it helps if nothing links to the page. The sector optimized the things a checklist measures and skipped the one thing it does not: whether the structure actually points at the content.

2. There are two ways to be unreachable, and a clean headline can hide either. One site buries 80% of its content in the invisible tier; another scores 0% invisible but routes 74% of its pages through navigation chrome an agent discards. Both are unreadable to a link-following agent. A single “invisibility” number, read alone, would have called the second site perfect.

3. The new content is the most exposed. The AI and product sections these companies are investing in hardest are the freshest and least-linked. Being quotable by an assistant is now a structural property, not a content one, and the markup that makes a page answerable (the FAQ and Q&A schema that is just 6% of the cohort’s structured data) is exactly what most of them have not written.

A note on prestige: the sites that hide the most content and concentrate the most betweenness are among the best-known names in the cohort. Engineering reputation does not buy a navigable topology. If anything, the largest catalogs are the hardest to keep wired.

The fix is structural, and every piece of it is measurable: re-link invisible archives into the editorial graph, move content off navigation-only reachability, spread betweenness off the corporate hubs, cross-link the product and AI sections to each other, and add answer-shaped markup to the pages meant to be quoted. This is the entire premise of the Digital MRI service.

Methodology

Ten anonymized Fortune 500 technology sites, each run through the same pipeline: an HTTP-first adaptive crawl, main-content extraction (navigation, header, and footer links filtered out), dual-graph construction, and the five-lens topology analysis (Skeleton, Circulation, Organs, Health, Nervous System). This is a curated cohort, not a random sample, so the aggregate figures are exploratory benchmarks for F500 technology topology rather than point estimates for the sector. Reachability and invisibility use only the nine sites whose crawl captured a usable sitemap; topology, fragility, content, and the scorecard use all ten. The cohort skews toward hardware, semiconductors, and enterprise software; consumer-internet platforms are under-represented because many serve their marketing surface behind anti-bot layers or as login-walled application shells that expose little crawlable structure.

The dual-graph model. Each page is classified against two graphs from the same crawl: the full graph (navigation included) for reach, and the content graph (in-content editorial links only) for authority. A page is reachable via editorial links if it has an inbound content link, navigation-only if reachable in the full graph but not the content graph, and link-invisible if nothing links to it in either, surviving only in the sitemap. This matches the model used in IR-002 (Financial Services, Q2 2026); the two reports’ per-site rates are not directly comparable, but the shared finding holds across three sectors now: a measurable share of F500 content sits outside any link path an unaided agent can walk.