See who links to them. Then ask for the same link.

• The profile: referring domains, backlinks, the share of referring domains that pass authority, links pointing at pages of yours that are broken, and the index’s own authority rank for the domain.
• A week-by-week count of referring domains gained and lost over the past quarter, so a profile that is growing and one that is merely churning do not look alike.
• The referring domains themselves — the index’s own top 100, which it ranks — each with its authority, how many links it sends, when the index first saw it and the spam score the index gives it.

The referring domains are stored for the domain you analysed and for up to 5 of the rivals your project tracks, which is the same number an analysis run compares. Nothing caps how many rivals a project may hold, so that is a bound on what one press costs rather than on what you may record.

The share that passes authority is referring domains divided by referring domains, and the figure says so where it is printed. The provider also reports a count of individual links by attribute, and mixing the two would divide one population by another into a percentage that looks entirely plausible and means nothing.

With a project, Backlink gap lists the stored prospects. You can also name domains and compare them from the answers Domain overview has fetched for each, with no project involved.

Both sides of the comparison are stored sets: the backlinks index’s own top 100 referrers of each domain, which it ranks — not all the domains linking to it. A domain among a rival’s and not among yours is a prospect, and the table names every rival it was found on. Where one domain links to several of them the row keeps the first rival’s figures rather than adding them up: how many links a domain sends is a fact about that rival’s relationship with it, and a sum across two rivals would be a figure neither answer reported.

So a domain that does link to you, from outside your own top 100, can appear here as a prospect. It is missing from the set it was compared against, not from your profile — check a row before you pitch it.

With a project, the comparison is always the same set — the domain you are looking at is the subject, and the rest of the tracked set is what it is compared against — so every member’s referrers are stored whether or not you have opened that member yet. For a domain outside that set the tab shows what your workspace has fetched for it in Domain overview, or the press to fetch it.

A domain the index gave no first-sighting date for is never counted as newly gained, and a domain it gave no spam score is neither clean nor flagged. An answer the provider did not give is printed as an absence rather than defaulted into a band.

There is a count of referring domains at or above a set point on DataForSEO’s spam score. The score is theirs and the line is ours — we picked where it falls, and the screen names the number so you can disagree with it. The figure also states what it is a count of: the referring domains stored for that domain, which are the index’s own top ones, rather than every domain linking to it. A count that read as a share of the whole profile would be a far more alarming number than the one actually measured. No change is drawn beside it either, because the stored set is replaced whole on every Analyze backlinks and there is no earlier count to compare it with.

There is no toxicity grade and no disavow file. This product reports the provider’s own spam score under the provider’s own name, and never issues a verdict of its own about how trustworthy somebody else’s site is; nor will it generate a file asking Google to ignore links pointing at you. Both are decisions rather than things not built yet.

The authority rank the provider reports arrives on more than one scale, and every stored measurement records which scale it was taken on. Where two measurements of one domain came from different scales the change is simply not drawn — a figure that moved because the ruler changed is not a change in anybody’s backlink profile.