How the web really builds links

8,814 websites, each carrying 20 metrics across three families: the site's organic footprint, its own outbound structure, and the inbound authority pointing at it.

Every figure is computed per-site and reported as a median, so a handful of giant sites can't distort the picture, and no per-site link counts are quoted in the analysis — the study speaks in proportions and ratios by design. The full methodology (how skew is handled and how the data was cleaned) is set out in the appendix.

Data sources. The dataset was compiled from ButterSwipe and Ahrefs — ButterSwipe for the site list, industry classification and organic-traffic/keyword data; Ahrefs for backlinks, referring domains and internal/external link metrics. See a representative slice of the raw inputs in the appendix.

Sort the 8,814 sites into five traffic tiers and a clean gradient appears: the bigger the site, the larger the share of its profile that is earned inbound authority, and the more deeply each referring domain links to it.

Small sites are almost entirely internal architecture with a thin backlink base. As traffic climbs, two things happen together: the backlink share of the profile expands, and — more tellingly — the number of links each referring domain contributes multiplies. A big site isn't just linked by more domains; each of those domains links to it again and again, a sign of ongoing coverage and relationships rather than one-off mentions.

The six industries split cleanly into two camps. Finance and Tech live on earned authority — backlinks are ~42% of their profile. D2C, Marketplace and Education are architecture-led, 70–75% internal links.

These are not stylistic differences; they reflect how each industry actually earns its traffic. Tech and Finance publish authority-magnet content — research, tools, documentation — that attracts links from many domains, so their profiles tilt toward earned authority. Marketplaces and D2C brands, by contrast, run enormous internal page inventories (products, categories, filters) held together by internal links, and rank through breadth of coverage more than off-site endorsement.

How much a site leans on earned backlinks depends heavily on where it competes. Mature markets — South Korea, Canada and the US — run backlink-heavy profiles; large emerging markets like India, Türkiye and Brazil are overwhelmingly internal-link-led. Reliance on earned links swings from roughly 2% to 52% of the profile.

The pattern tracks market maturity and competitiveness. In markets with a deep, established web — where many strong sites compete for the same queries — earning links is both harder and more decisive, so profiles tilt toward backlinks. In fast-growing markets, sites can still win on content breadth and internal structure alone, so their profiles stay internal-heavy. For anyone expanding internationally, that gap is a roadmap: the off-site bar is far lower in emerging markets.

Correlate the on-site and off-site link signals against organic traffic and a clear order emerges: the diversity of referring domains leads, raw backlink volume follows, and on-site link counts trail. (Keyword metrics are set aside here — they're partly a restatement of traffic itself.)

The ranking is a useful corrective to link-count vanity. Ten links from ten different domains carry more signal than a hundred from one, and a large internal-link count — while essential for crawlability — barely moves the traffic needle on its own. The second chart reframes the keyword question the way it matters for this report: instead of asking where a site's keywords rank, it asks how a site's link mix shifts as its keyword footprint grows.

AI Overviews and LLM answers don’t cite at random — they pull from sites that already rank in the top results and carry real off-site authority. Using those signals, here’s which industries are best positioned to be surfaced by AI.

There is no public “AI rank” to download, so this is a modelled readiness index (0–100), not measured AI placement. It combines the four signals shown to predict AI-Overview / LLM citation — existing top-3 rankings (35%), referring-domain authority (30%), topical breadth (20%) and link-follow quality (15%) — into one score per site. The pattern it reveals is the point: authority, not raw ranking share, is what separates the AI-ready.

The six categories fan out into hundreds of sub-categories. This treemap sizes every niche (with ≥8 sites) by how many sites it holds, grouped and coloured by its parent industry.

Eight findings, each paired with the move it implies. The through-line: build the internal system first, earn diverse domains second, and match the emphasis to your industry and market.

A practical audit distilled from the data — work top to bottom. Benchmarks in bold come straight from the 8,800-site medians. Tick items off as you go.

Method. Ratios are per-site shares of Internal links, outbound External links (all-time) and inbound Backlinks (all), summarised as medians over 8,788 clean sites. “Links per referring domain” = backlinks ÷ referring domains, median per group. Correlations are Spearman rank correlations against organic traffic. Traffic tiers: <50K, 50K–250K, 250K–1M, 1M–5M, 5M+. Keyword tiers: <2.5K, 2.5K–10K, 10K–50K, 50K–250K, 250K+. Follow-rate = followed ÷ all, median per group. Figures are proportions and ratios only; absolute link counts are intentionally not published. Every chart above is mirrored by a data table so the findings remain fully readable by search engines and assistive technology.