A backlink profile audit is a diagnosis, not a tactic. The point is not to chase a number, but to understand what the existing link profile actually looks like to Google’s link systems and decide whether the shape is healthy or whether something needs attention.
Most audit tools throw dozens of metrics at the screen, many of them derivative or tracking movement that changes nothing. The signals worth measuring fall into a smaller set, each answering a specific question about the profile’s structure, sources, and direction. What follows is the practical version: the signals that matter, and how to read each without the metric panic these tools induce.
Audit before you build #
Link building without an audit is building on an unknown foundation: a site might be a step from trouble and never know it, or have a clean profile that underperforms for lack of authority diversity, or be losing links faster than it gains them, and all three look identical from the outside until the data is pulled. The audit also frames what to build next, and its output is not a score but a list of decisions: what to keep doing, what to stop, and what to start.
Two kinds of data feed it. Google Search Console’s Links report shows a sample of the links Google has actually crawled and attributed to your site, from its own index. Third-party tools (Ahrefs, Semrush, Moz, Majestic) crawl the web independently and usually report larger counts because they include links Google may have already found and quietly discounted. Use both: Search Console tells you what Google sees, the third-party export tells you what exists on the crawled web, and the overlap is the profile you can most trust.
Signal 1: referring domains, not link count #
The single most useful number is the count of unique referring domains. A site with 10,000 links from 50 domains is generally weaker than one with 500 links from 400 domains, because each new referring domain is a separate editorial decision while each extra link from a domain already linking adds diminishing value. Many different sources build a distributed pattern that is hard to fake at scale. Compare referring domains against the linking-page count; the wider the gap, the more concentrated the profile. A sitewide link from a partner, sitting in every footer, inflates the total but represents one decision, so strip sitewide links first. The referring-domain count of pages ranking in the vertical sets the floor, and the gap defines the work.
Signal 2: authority distribution #
Next is how authority spreads across the linking domains, estimated by third-party scores: Domain Rating in Ahrefs, Domain Authority in Moz, Trust Flow in Majestic. These are not Google metrics. Google’s PageRank has not been publicly disclosed for years, and these scores are independent approximations that correlate roughly with linking power, so read them as directional. A grown profile shows a broad base: a few high-authority links, a solid mid-authority layer, and a wide base of legitimate lower-authority links from small blogs and regional publications. Shapes that look engineered include a profile of almost entirely very-high-authority links with nothing beneath, its opposite built only of very-low-authority sources, and a hollow middle that suggests top-tier placements chased without the mid-tier coverage where sustainable authority lives. Sort by authority score; a missing layer is the finding.
Signal 3: anchor text distribution #
Anchor text is the visible clickable phrase of a link, and its spread reveals whether acquisition looks natural or over-optimized. A natural profile is dominated by branded and URL anchors, because most editorial links reference a site by name or address, while exact-match commercial anchors are the smallest slice. The shares below are rules of thumb from audit practice, not published Google thresholds:
| Anchor type | Example | Typical natural share |
|---|---|---|
| Branded | "Acme Corp", "Acme" | Dominant |
| URL | "acme.com", "www.acme.com" | Substantial |
| Generic | "click here", "read more" | Small |
| Partial match | "marketing software platform" | Small |
| Exact-match commercial | "best marketing software" | Smallest slice |
Google does not publish an anchor-percentage cutoff, so treat any specific figure with caution. Its guidance describes the concern: keyword-rich anchors built deliberately across many sites are a hallmark of link schemes, and when a single commercial keyword comes to dominate, the pattern reads as engineered and its systems discount it. Sort a tool’s anchor-text report by frequency; a brand name on top with URL variations and contextual phrases below it looks healthy. If a commercial keyword leads with a significant share, the fix is slow: disavow does not change anchor text, so rebalancing means building enough branded and URL anchors to push the exact-match share down over months.
Signal 4: topical relevance of linking domains #
A link’s value depends partly on whether the linking site is topically related to the destination. A link from a marketing blog to a marketing-software company carries more weight than one from a recipe site at similar authority, because a relevant source suggests someone in the field found the destination worth citing. Most tools do not categorize topics, so grouping is manual: scan the top 50 to 100 referring domains and mark each relevant, adjacent, or unrelated. Google’s spam updates have repeatedly targeted patterns where high-authority but irrelevant sources link to a destination, since that frequently reflects paid placements or repurposed expired domains. If unrelated sources dominate the high-authority tier, the strategy that built them needs reconsideration.
Signal 5: link velocity over time #
Link velocity is the rate at which a site gains and loses links, revealing whether acquisition is consistent and organic-looking or whether spikes suggest campaigns or manipulation. Healthy velocity looks like a steady gain curve with occasional bumps when content earns coverage, a low loss rate, and a gentle upward trend. Unhealthy velocity looks like a flat line broken by a sudden spike of thousands of links in a week, a spike followed by rapid decay as low-quality links fall off the indexed web, or a slow decline where losses outpace gains unnoticed because no single loss registers. Read the slope first, the variance second, and the recent quarter against the trailing year to see whether the trend is accelerating, holding, or reversing. The same chart for competitors shows whether the site is keeping pace.
Signal 6: link attribute distribution #
The mix of link attributes shows how links were acquired and how Google treats them. A plain editorial link carries no rel qualifier; the three qualifiers are nofollow, sponsored, and ugc (user-generated content). One correction, because the older framing persists: since 2020 Google treats all three as hints rather than strict directives, and may still consider a qualified link for crawling and indexing, so none guarantees a link is ignored entirely. In practice qualified links generally do not pass ranking signals, but the idea that nofollow is an absolute block is outdated. An all-dofollow profile tends to look engineered, since organic acquisition picks up qualified links from forums, comment sections, social platforms, and publishers that qualify external links by policy. Sponsored attribution is not a negative signal but the correct disclosure for paid placements. Track the ratio over time rather than the absolute count, since a sudden shift sometimes marks a new campaign.
Signal 7: toxic links and spam indicators #
This category gets the most attention in audit tools and the least justified action in practice. Be clear about the metric: a Spam Score in Moz or a Toxicity Score in Semrush is a third-party heuristic, not a Google signal. Google’s John Mueller has said Google does not use these scores; they are correlation models built from features common to previously penalized sites, useful for triage but not a verdict, so a high score correlates with risk without proving a penalty. Other indicators include unusual TLD patterns, deindexed or zero-traffic domains, and shared footprints of link networks. None means action is automatically required: a few spam links accumulate on most sites without consequence, and Google’s link spam systems devalue obvious spam rather than penalize a destination for not removing it. The pattern matters more than any single link. Disavow belongs to a narrow set of cases: Google’s own guidance is that most sites never need the tool, because its systems already ignore the vast majority of spammy links automatically. Reserve it for a confirmed manual action for unnatural links in Search Console, or documented large-scale manipulation you cannot get removed at the source; otherwise a disavow file tends to do more harm than good, since it is easy to accidentally exclude legitimate links.
Signal 8: comparison against direct competitors #
The first seven signals describe a profile in isolation. The eighth places it in context, because ranking is relative: a profile that looks healthy alone can still be uncompetitive where the top sites carry more referring domains, deeper authority, and more relevant coverage. Pick three to five competitors ranking in the top ten for the main target queries, pull the same signals for each, and lay them side by side:
| Metric | Audited site | Comp 1 | Comp 2 | Comp 3 | Gap |
|---|---|---|---|---|---|
| Referring domains | |||||
| Authority spread (DR / DA) | |||||
| Top anchor share | |||||
| Topical relevance | |||||
| Monthly velocity | |||||
| Dofollow ratio |
The gaps become the priority list. Twice the referring domains means the work is volume; matched volume but stronger relevance means the work is source quality; a match on everything but faster movement means the work is sustained acquisition. Most tools also offer a link-intersect or backlink-gap feature that surfaces domains linking to competitors but not to the audited site, and that list becomes the outreach pipeline.
When to repeat, and what the audit makes visible #
Backlink profiles change continuously, so the audit is not a one-time exercise. An active campaign warrants monthly velocity checks and quarterly full audits; a stable site with light activity needs a quarterly check and an annual deep review. Certain events call for an immediate off-cycle audit regardless of schedule: a sudden ranking drop, a manual-action notice in Search Console, a confirmed major algorithm update, a negative-SEO attack, or a merger, migration, or major restructure. The value compounds when the audit produces a trend line rather than a snapshot: comparing the current quarter against the previous four shows whether the profile is strengthening, weakening, or holding static.
Together these eight signals form a footprint that Google’s systems read as one composite pattern, invisible to most site owners because the inputs sit across multiple tools. The audit assembles them, and once the footprint is visible the diagnosis tends to be straightforward: a profile of exact-match anchors and one-off domains looks like a link program that did not take time to look natural, while a profile of editorial coverage looks like coverage. The work that follows is changing that footprint one acquisition decision at a time, in the direction the diagnosis pointed.