LR is one number, 0–100. These are the rules that produce it.
Every LR on Theylink comes from the same formula over the same index, recomputed every night. This page is the complete rulebook — what counts, what doesn’t, what moves the number, and what it is not. No hidden weights.
A live domain from the index. The two terms of the formula on the right are recomputed for it every night; the blend is fixed at 50/50 in v1.
What counts as a referring domain
The formula, v1
# per domain, nightly at 03:00 UTC
ref_score = log(1 + referring_domains)
/ log(1 + max_referring_domains) × 100
rank_score = pagerank(follow links between domains,
dead links excluded, damping 0.85,
30 iterations) → log-scaled to 0–100
LR = round(0.5 × ref_score + 0.5 × rank_score)
# bounds
referring_domains = 0 → LR = 0
otherwise → 1 ≤ LR ≤ 100
most-referenced domain in the index → LR = 100Half the score is how many domains link to you, on a log scale so 1,000 vs 10,000 matters as much as 10 vs 100. The other half is an authority pass: a link from a domain that is itself well linked counts more than a link from an orphan. Only follow links carry authority.
What moves your number, and when
Three clocks. All of them are visible on the page next to the number.
Snapshot · 11 Sept 2026 · 4,562 domains
| Domain | Referring domains | LR v1 |
|---|---|---|
| instagram.com | 91 | 88 |
| tiktok.com | 33 | 57 |
Real numbers from the index, checked 25 hours ago. A famous domain with a modest LR is not a mistake: only the sites we have crawled so far count. This is what “relative to the index” means, and why the size is printed everywhere.
How a gap row is scored
The comparison list is sorted by how likely you are to get the link, not by how impressive the domain is. LR is one of four factors.
score = lr_weight(page.lr) × relevance(page.topic, your.topic) × ease(page.type) × freshness(page.last_checked) lr_weight 0.25 → 1.0 0.25 + 0.75 × LR/100 — a high-LR source is worth up to 4× a zero relevance 0.5 → 1.5 topic similarity, 1.0 when unknown ease see table how the link is usually earned freshness 0.55 → 1.0 1.0 under 30 days · 0.85 under 90 · 0.7 under 180 · 0.55 older · 0.6 never checked
| Page type | Ease | How it is detected | What the fix usually is |
|---|---|---|---|
| Directory | ×1.5 | URL patterns, a submit form, category grids | Self-serve: fill in the form |
| Listicle | ×1.3 | “12 best…”, “Top 10…”, numbered roundups | One email with a reason you belong |
| Review | ×1.0 | Deep dive on one product | Offer a license and a comparison angle |
| Blog mention | ×0.8 | Passing link in an article | Low effort to ask, low odds |
| Forum / UGC | ×0.5 | Reddit, HN, Q&A; often nofollow or ugc | Included, weighted down |
| Other | ×0.7 | Rules undecided, model unsure | Find the contact; the “report” link fixes the label |
Classification is rule-based for most pages; the rest goes to a language model once, at crawl time, and is cached. Mislabels happen; the “report” link on a row corrects it for everyone.