A backlink audit establishes whether a link profile is helping, doing nothing, or actively suppressing a site. Most audits assume the third. In practice the second is far more common, and the distinction determines whether any action is warranted at all.
An honest answer, including when the answer is that nothing needs doing.
A backlink audit is an examination of every site linking to yours, classifying each as earned, purchased, automated or hostile, and establishing whether the profile is helping, neutral or harmful. It is warranted when traffic and link volume are badly mismatched, when a manual action has been received, or when rankings fell sharply without a site change explaining it.
The mismatch is the most reliable signal. A site with two thousand referring domains producing four hundred monthly visits is behaving very differently from one with three hundred and seventy domains producing two thousand six hundred. Both figures are real and both were measured in the same market.
Measured, not assumed: across seven companies analysed in one industrial market in July 2026, five showed evidence of purchased or spam backlinks — blog networks, cheap top-level domains, an SEO vendor case study and directory schemes. Not one of the seven appeared in the Google AI Overview returned for their own category, while three companies outside that set did.
Less often than the industry selling toxic link removal suggests. Search engines have stated for years that spam links are largely discounted algorithmically rather than penalised, and disavowal is genuinely necessary mainly where a manual action exists.
That does not make the question unimportant. It makes the diagnosis important, because the corrective work differs entirely. A site suppressed by a hostile link campaign needs disavowal. A site with a poor profile and thin content needs pages, and disavowing links would achieve nothing while consuming a month.
From our own measurement: in July 2026 we examined a company holding 2,000 referring domains and receiving 413 monthly visits, with the profile dominated by cheap top-level domains and around 700 domains using five identical exact-match anchors. In the same market a competitor with 376 referring domains received 2,600 visits. Separately, a market leader with an explicit spam vendor advertisement embedded in 200 of its backlinks showed the best traffic-to-link ratio measured and a mild decline of under 7 percent. The presence of bad links and the presence of harm are different findings.
Disavowal carries its own risk. Removing a genuinely earned link from the file is not quickly reversible, and a badly built disavow file can cost more than the links it removes. That is the reason to establish whether a problem exists before acting, rather than treating a long list of unfamiliar domains as self-evidently harmful.
A backlink gap analysis identifies sites linking to several of your competitors but not to you. Those sources are the most realistic acquisition targets available, because they have already demonstrated willingness to link to businesses like yours.
The direction of travel is documented outside our own work. Semrush tracked Google AI Overview appearance rising from roughly 6.5 percent of queries in January 2025 to a peak near 25 percent that July across a ten-million-keyword sample. BrightEdge, measuring nine commercial industries, recorded 48 percent of queries triggering an AI Overview by February 2026. Seer Interactive measured organic click-through on those queries falling from 1.76 percent to 0.61 percent. Prevalence estimates vary by methodology and should be read as a range rather than a figure.
Three pieces, in order: an analysis establishing where you stand, a build specification stating exactly what to construct, and optionally the pages themselves. Most companies start with the first and decide from there.
Your position in search and AI answers, who is beating you, what share of your market you capture, and the reasons behind all three.
Every page mapped. Target term, URL, headings, schema, internal links, priority order. Written so your developer can work straight from it.
Optional, quoted separately. Most specifications die on delivery because nobody writes the pages. We write them if you want us to.
Send your domain. We will examine the profile, identify anything purchased or hostile, and tell you plainly whether it is causing harm or simply sitting there.
No call required and no obligation. Read it, check it yourself, and decide. If the findings are worth discussing after that, we can talk then.