A backlink gap analysis finds the domains linking to your competitors but not to you. Every SEO tool will run one in about four seconds. The problem is that the raw output is close to useless, and acting on it unfiltered is how agencies end up buying links from sites nobody reads.
This is the full process, including the filtering that turns four thousand rows into a shortlist you can actually work.
What a gap analysis is for
It answers one narrow question: which sites have already demonstrated a willingness to link to a company like yours? That is a much better starting point than a cold prospecting list, because the editorial precedent already exists. Somebody at that publication has already decided your category is worth covering.
A gap analysis does not tell you which links are worth having. It tells you which links exist. Those are completely different questions, and conflating them is the single most common way link building budgets get wasted.
Step one: pick the right competitors
Most people pick their business competitors. That is usually wrong. You want search competitors, which is a different set.
- Take your five highest-intent commercial keywords.
- Note which domains rank in the top ten for each, ignoring marketplaces and directories.
- The domains that appear across three or more of those keyword sets are your real search competitors.
- Include one aspirational domain that is clearly above your weight class. You will not earn most of its links, but its profile shows you what the ceiling looks like.
Three to five competitors is the sweet spot. More than that and the intersection filters become meaningless.
Step two: run the intersection, not the union
Every tool lets you choose how many competitors a domain must link to before it appears in the report. Default settings usually show anything linking to at least one competitor, which produces enormous, low-signal lists.
Set it to at least two, ideally three. A site that has linked to two of your competitors independently is demonstrating a pattern rather than an accident, and pattern-linkers are dramatically more likely to link to you too.
Step three: filter, ruthlessly
This is the step that separates a useful analysis from an expensive one. Run every surviving domain through these gates and reject anything that fails a single one.
| Filter | Threshold | Why |
|---|---|---|
| Real organic traffic | 5,000+ monthly visits | The single best proxy for whether a site has actual readers rather than just metrics |
| Domain Rating | 50+ | A floor, not a target. It excludes the obviously weak and qualifies nothing on its own |
| Traffic curve shape | Steady or growing | A flat line that suddenly spikes is the signature of a rebuilt expired domain |
| Outbound link profile | No casino, crypto or pharma | Sites that sell links to those verticals sell links to everyone |
| Topical relevance | Same or adjacent category | Relevance is what makes the link make sense to a reader, which is the only durable definition |
| Link placement type | In-content, editorial | Footer links, author bios and sidebar sponsorships pass a fraction of the value |
Expect to reject somewhere between eighty and ninety-five per cent of the raw list. If you are rejecting less than that, your filters are too loose.
The one pattern to memorise
High Domain Rating combined with almost no organic traffic is the most reliable warning sign in link building. A rebuilt expired domain can carry DR 60 and receive nine visitors a month. It looks like a win in a spreadsheet and does nothing, or worse than nothing, in reality.
Run a prospect through our actual criteria
The Link Prospect Grader applies the exact six filters above and tells you pass or fail with the reason, including flagging the high-DR, low-traffic pattern automatically. No email required.
Grade a prospectStep four: sort by earnability, not by metrics
A filtered list is still not a plan, because it is sorted by how good the links are rather than by how likely you are to get them. Re-sort into three buckets:
- Replicable. Resource pages, tool directories, glossaries and curated lists. Someone maintains these deliberately and adding you is a small editorial decision. Highest hit rate, lowest ceiling.
- Earnable. Publications that cover your category and have linked to competitors in editorial context. These need a reason: original data, a genuine expert quote, or a story. Medium hit rate, high value.
- Aspirational. Tier-one publications. These need a real digital PR programme and usually a piece of original research. Low hit rate, transformative when they land.
Work all three simultaneously in roughly a 60/30/10 split of effort. Working only the first bucket caps your ceiling. Working only the third means six months with nothing to show.
Step five: find the actual link, not just the domain
For every prospect, open the specific page that links to your competitor and read it. You are looking for one of three situations:
- A list you could join. The easiest possible pitch, because the editorial format already accommodates you.
- A claim you could improve. If they cited a statistic from 2022 and you have published a better one from this year, you have a genuine reason to be in touch.
- A gap in their coverage. They covered the category but missed a use case you own. That is a story pitch, not a link request.
If you cannot articulate which of those three applies, you do not have a pitch. You have a link request, and link requests get ignored.
What to do with the output
The finished artefact should be a spreadsheet with one row per prospect containing: the domain, the specific URL linking to a competitor, which competitors it links to, DR, organic traffic, the bucket, the angle, and a status column. Nothing more. Anything you cannot act on is decoration.
A realistic monthly working list is thirty to fifty prospects. From that, expect somewhere between three and eight placements depending on the quality of the angle and the seniority of the person doing the outreach. Anyone promising you materially more than that from a list of that size is either buying links or counting things that are not links.
The AI-era addendum
One thing has changed. Because answer engines weight third-party mentions heavily, an unlinked brand mention on a relevant, well-read site now has real value even when there is no link attached. It is still worth asking for the link. It is no longer worth walking away when the answer is no.
Add a column to your spreadsheet for mention-only outcomes and count them. They will not show up in your backlink profile and they will show up in your AI share of voice.
Related reading: AEO vs SEO explains why third-party mentions carry the weight they do, and the free audit will tell you whether your own site can hold the authority you are about to build.