Use this when
backlinks, link profile, referring domains, anchor text, toxic links, backlink audit.
Process
Detect what's actually available before analyzing: a connected DataForSEO/Moz/Bing source if one exists, otherwise fall back to Common Crawl's domain-level graph (PageRank and presence signals, always available, no account needed) plus a verification pass on any specific links the user names.
Produce seven sections, each source-labeled: profile overview (referring domains, follow ratio, domain diversity, trend); anchor text distribution against healthy benchmarks (branded 30-50%, URL/naked 15-25%, generic 10-20%, exact-match keyword 3-10% with over 15% flagged as an over-optimization risk, partial match 5-15%); referring domain quality (TLD and country distribution, authority-tier spread); toxic link indicators (PBN patterns, unnatural anchor concentration, mass directory submissions, link farms, footer/sidebar paid-link patterns across every page of a domain); top pages by backlinks (link magnets and zero-backlink internal-linking opportunities); competitor gap analysis when a comparison domain is given; new/lost backlinks (explicitly only available with a premium source — say so rather than guessing at velocity from a point-in-time snapshot).
The hard rule for the whole playbook: never present a numeric score built on data you don't actually have. Count how many of the 7 scoring factors (referring domains, domain quality, anchor naturalness, toxic ratio, link velocity, follow ratio, geographic relevance) have real data behind them. Four or more factors scored: produce a numeric 0-100 score with weights redistributed across the available factors. Fewer than four: do not produce any numeric score, not even an "approximate" one — report "Backlink Health Score: INSUFFICIENT DATA (X/7 factors scored)" with each available factor's individual finding and its source, and recommend what a free Moz signup would unlock.
Before presenting anything, self-check: does every metric carry a source label; is every "not found" result distinguished from "not checked" versus "below threshold"; is the referring-domain count in the summary actually consistent with the detailed list; is any claim presented without a source behind it? Fix any failure before showing the report — never present inferred data as fact, and never let a Common-Crawl-only analysis carry a numeric score of any kind.