Pull the same domain in two backlink tools and you will get two numbers, often several multiples apart. This is not a bug in either tool, and it does not mean one of them is lying. It means they are measuring different things with different instruments, and the single number each one shows you compresses all of that away.
The practical risk is real: teams switch vendors and report a "link loss," agencies get judged against a competitor's dashboard, and audits treat one tool's count as ground truth. None of that survives contact with how the data is actually made.
Where backlink data comes from
Google does not sell or share its link graph. The links report in Search Console shows you a sample of links to your own verified properties, with no metrics and no view of competitors. So every commercial backlink index — Ahrefs, Semrush, Majestic, Moz, and the rest — is built by running an independent crawler over the web and storing the links it happens to find.
That word independent does most of the explanatory work. Each vendor decides which URLs to crawl, how often to revisit them, how deep to go into large sites, whether to render JavaScript, and how to spend a finite crawl budget across an effectively infinite web. Two crawlers making different choices will discover overlapping but different sets of pages, and therefore overlapping but different sets of links. The disagreement starts before a single counting rule is applied.
Index size and freshness pull against each other
A backlink index has two qualities in tension. A bigger index has seen more of the web, so it finds more of your links — including links on pages that were deleted years ago. A fresher index revisits pages often, so it knows which links still exist — but revisiting known pages costs crawl capacity that could have discovered new ones.
Vendors resolve this tension differently, and some expose the choice directly: Majestic publishes separate Fresh and Historic indexes, with very different counts for the same domain by design. Other tools blend recency into one number, flag links as "lost" after a recrawl, or quietly age links out. Ask, for any count you are looking at: over what time window, and verified how recently? A tool reporting every link it has ever seen and a tool reporting links confirmed live this quarter are answering different questions, and both answers can be correct.
The tools do not agree on what a link is
Even on the same crawled page, counting rules diverge. The differences that matter most:
- Duplicate links on a page. Same target linked from the header, body, and footer — one link or three?
- Sitewide links. A footer link on a 100,000-page site can be one relationship or 100,000 backlinks depending on deduplication policy.
- Redirects and canonicals. Whether a link to a URL that redirects to you counts, and whether links are consolidated to canonical URLs, varies by vendor.
- Nofollow, UGC, and sponsored. Included, excluded, or included-but-flagged.
- Subdomain boundaries. Whether blogspot-style subdomains count as separate referring domains changes domain counts substantially.
Referring domains is more robust to these choices than raw backlink counts, which is one reason practitioners prefer it. But it is only more robust, not immune — the subdomain and redirect rules still bite.
Scores are not comparable across tools
Domain Rating, Authority Score, Domain Authority, and Trust Flow are proprietary metrics computed on each vendor's own graph with each vendor's own algorithm. A DR of 60 and a DA of 60 are not the same claim about the world; they are not even claims on the same scale type. Comparing your DR to a competitor's DA — which happens in real pitch decks — is a category error. Within one tool, compared over time or across competitors, these scores are useful relative signals. Across tools, they are noise.
How to compare vendors honestly
You cannot benchmark against the true link graph, because nobody outside Google has it. You can benchmark against the best partial ground truth available: your own site.
- Export Search Console links for a property you control. It is a sample, but it is a Google-sourced sample.
- Check recall. What share of the referring domains Search Console shows does each tool know about? This tests coverage where you can actually verify it.
- Check precision. Sample 50 "live" links from each tool and open them. What fraction still exists and still links to you? This tests freshness claims directly.
- Test discovery speed. When you earn a new link you know about, note how many days each tool takes to report it.
- Weight your niche. A tool with the biggest global index may still be thin in your language or industry. Run the recall test on a competitor set, not just your own domain.
Then pick one tool and stay in it for trend measurement. Consistency of instrument matters more than the instrument's absolute calibration — a stable index measuring growth is more useful than oscillating between two indexes that disagree.
What to do
Stop treating any backlink count as a census; treat it as a survey with a documented methodology. Use referring domains rather than raw link counts for goals and reporting. Never compare metrics across vendors, and never let anyone else do it in a document you sign off on. When link counts move sharply, check the vendor's index update notes before diagnosing anything about your site — index rebuilds routinely move numbers with no real-world change behind them. And once a year, rerun the recall-and-precision test above; vendors change their crawlers, and last year's winner is not guaranteed to be this year's.