The backlink audit: sort by first seen, then by relevance

The newest links to a site are often junk from link sellers and casino domains, with high spam scores. Underneath them are the links that count. How to tell the two apart in an hour, and why relevance beats every third-party score.

Two things, and they pull in different directions. The first is to find out which links are doing anything for the site, so you can get more like them. The second is to find out whether anything pointed at the site is a liability. Use the audit to identify links worth reading and target pages worth repairing.

Sort the backlinks by the date they were first seen and the pattern usually shows up in the first screen. The newest rows on many small sites come from link sellers, PBN adverts and casino domains, with high spam scores and anchor text that is an advert for buying links. Nobody at the site asked for them. Below them sit the links someone chose to make: a blog citing a resource guide, a related site linking a services page, a recap of an interview. Low spam scores, anchor text that describes the page.

The junk arrives on its own and announces itself. The links that matter look like someone chose to make them.

The score is not the audit

Every backlink tool carries an authority number for the linking domain. SEO Scaler shows DataForSEO's domain rank alongside a spam score. They are useful for sorting. They are not the verdict.

What decides whether a link is worth anything is whether it is relevant, not a third-party metric. A spam domain can carry a higher domain rank than every legitimate link beneath it. Sort by rank and the worst link on the page can float to the top.

So the audit has two passes. The first uses the scores to sort. The second reads the rows.

Pass one: sort and bucket

Start with one page of the profile, using one row per referring domain. Turn off the spam filter so you can see the rows you are sorting. Then make four buckets.

Obvious junk. High spam score, or anchor text that is an advert, or a source domain whose name has nothing to do with anything. Expect a long tail of /all/2585/6.html-shaped URLs on throwaway domains. You will recognise the pattern after ten rows.

Broken targets. The backlink overview reports broken backlinks and broken pages. Open the target URLs to check what is wrong. Restore a missing page or redirect it to a relevant replacement where one exists. This profile request returns live backlinks, so it is not a list of lost links.

Nofollow and syndication. Podcast directories, aggregators, social profiles. Real, harmless, and not what you build a strategy on. Count them so they do not inflate the total in your head.

Everything else. This is the bucket you audit properly.

For each row in the last bucket, the question is not "how strong is this domain" but "would a person on that page have a reason to click through." Check whether the linking site is a real business with real traffic, and whether the page is about the same thing yours is. For local businesses, local publications tend to be the links that do the most.

A link from a town newspaper to a plumber is worth more than a link from a high-authority marketing blog to the same plumber, and no score will tell you that. The row will, if you open it.

Read the anchor and the surrounding text. Read what the linking page is about. Note whether your link is the only outbound link or one of two hundred. Twenty rows takes twenty minutes and tells you more than any composite.

Run this with the SEO Scaler MCP

Connect the SEO Scaler MCP. The agent shows the credit cost before each call runs, and nothing spends until you approve. The link-prospecting skill packages the outreach side of this.

Using the SEO Scaler MCP on project [name], audit the backlinks for
[domain]:

1. Call get_backlinks_overview with scope "domain". Report backlinks,
   referring domains, rank, spam score, broken backlinks and broken
   pages.

2. Call get_backlinks_profile with scope "domain", mode
   one_per_domain, hideSpam false, sorted by firstSeen descending,
   page 1, 100 rows. Bucket every row for review: suspected junk
   (spam score over 30 or advert-style anchor text), broken targets,
   nofollow, or worth reading. Scores are sorting aids, not verdicts.

3. For the "worth reading" bucket, list source domain, source page
   title, anchor text and target page. Flag any target page that
   returns an error.

4. Start a new get_backlinks_profile request with scope "domain",
   mode as_is, page 1, 100 rows, filters maxSpamScore 10 and
   minDomainRank 20. In this sample, list domains linking to more
   than one distinct target page. Say that other links may be absent
   from the sample.

5. Give me two lists: broken target pages to investigate for repair
   or a relevant redirect, and ten linking pages to read in full.

Step 3 is the one that matters and the one an agent cannot finish for you. It can list the rows. Reading them is the audit.

The audit is a reading exercise

A backlink profile is a list of decisions other people made about your site. The scores help you sort the list. They do not tell you what anyone decided or why. An hour with the rows does, and it ends with two things you can act on this week: the broken pages to investigate, and a short list of sites that already chose to link to you, which is where more links usually come from.

Start with a page of backlink rows, one per referring domain, sorted by first seen, with spam filtering off. Sort them into suspected junk, broken targets, nofollow, and worth reading. Check broken pages before restoring them or choosing a relevant redirect. Read the remaining rows and decide whether a person on the source page would have a reason to click through.

A link from a page that exists only to sell or host links: link seller listings, PBN adverts, casino and pharmacy domains, and pages with hundreds of unrelated outbound links. They tend to have high spam scores and anchor text that reads like an advert. Google generally ignores them.

It is a sorting aid, not a verdict. A spam domain can carry a higher rank than a relevant small site. Use the score to order the list, then judge each link on whether the linking page is about the same thing as yours and whether a real site published it.

The overview reports broken backlinks and broken target pages. Profile rows include broken status, dofollow or nofollow, domain rank, spam score and first-seen date. The profile request used here returns live backlinks, so it cannot list lost links.

Run this strategy in SEO Scaler

Run the MCP prompt in this guide with SEO Scaler. SEO Scaler is free to try, and asks before every paid call.

Start with SEO ScalerBack to Link Building