Backlink analysis: how to audit your backlink profile

Backlink analysis is the part of an SEO audit most likely to be done by a tool and least likely to be understood by the person reading the output. A score goes up or down, a column says “toxic”, and somebody files a disavow they did not need. This guide explains what backlink analysis measures, what the authority scores actually are, how to check backlinks without paying, how to run a competitor comparison, when a disavow is justified according to Google, and what the evidence says about links in the age of AI answers. One number sets the tone: Research shows that in Ahrefs’ study of 75,000 brands, backlinks correlated with visibility in Google’s AI Overviews at just 0.218, while branded web mentions correlated at 0.664.

What is backlink analysis?

Backlink analysis is the review of every link pointing at a website from other sites: how many there are, how many distinct sites they come from, what pages they point to, what anchor text they use, how the profile compares with competitors, and whether any part of it looks manufactured. The output is a short list of decisions: which pages deserve more links, where competitors get theirs, and whether anything needs cleaning up.

It answers five questions in order:

1. How big is the profile, really? Referring domains, not raw backlinks, because one site linking a thousand times is one vote.
2. Where does it point? Which pages attract links, and whether the pages you want to rank are among them.
3. What does the anchor text say? Brand names and URLs are normal; a wall of exact-match commercial phrases is not.
4. How does it compare? Which sites link to competitors but not to you.
5. Is any of it a risk? Patterns that match Google’s spam policies, not a vendor’s “toxicity” column.

The same analysis is step 5 of the full SEO audit process; this guide is the long version.

What is a backlink, and which ones count?

A backlink is a hyperlink from a page on one site to a page on another. Search engines read it as a reference. A link from a page that is itself well linked carries more weight, a link from a site that has never linked to you before counts for more than a repeat, and the words in the anchor tell the engine what the linked page is about.

Three attributes change how a link is treated:

Attribute	What it tells search engines	Typical use
none (a plain link)	The linking site vouches for the page	Editorial links, citations, references
rel="nofollow"	Do not treat this as an endorsement	Links the site does not vouch for
rel="sponsored"	This link was paid for or is an advertisement	Affiliate links, sponsorships, paid placements
rel="ugc"	This link was added by a user, not the publisher	Comments, forum posts

Google says paid links are acceptable only when qualified with rel="nofollow" or rel="sponsored", and that since its 2019 change all three attributes are treated as hints about which links to consider rather than as hard instructions. In practice, a backlink audit counts plain links as the signal and tracks the others as context. The nofollow and referring domain glossary entries cover the mechanics.

What do DA, DR and Authority Score actually measure?

They measure link strength as estimated by the company that sells the metric. None of them is a Google number, none of them sees Google’s link graph, and they are not comparable with each other. Each is useful for ranking a list of sites by rough link strength and useless for predicting whether a specific page will rank for a specific query.

Metric	Owner	Scale	What the owner says it is based on	What it is not
Domain Authority (DA)	Moz	0 to 100, logarithmic	A machine-learning model trained on Moz’s link index to predict how likely a domain is to rank compared with others (vendor claim)	A Google ranking factor; Moz’s own documentation says so
Domain Rating (DR)	Ahrefs	0 to 100, logarithmic	The strength of a site’s backlink profile, from the number of referring domains, their DR and how many sites each links to; nofollow links do not pass DR, and “the exact formula of Domain Rating is not disclosed” (Ahrefs’ DR definition)	A measure of traffic, content quality or spam
Authority Score (AS)	Semrush	0 to 100	Semrush describes it as a compound of link strength, organic traffic and spam signals (vendor claim)	A Google number

Three cautions apply to all of them.

• Google says it does not use them. Google’s John Mueller has answered the question in public for years with variations of “search engines don’t use it”. The scores correlate with rankings because both depend on links, not because one feeds the other.
• They move for reasons unrelated to you. Index refreshes, formula changes and the vendor discovering a batch of links to a competitor can change your score while nothing changed on your site.
• Logarithmic scales flatten the top and exaggerate the bottom. Moving from 20 to 30 is routine; moving from 70 to 80 can take years.

Use them the way they were designed to be used: to sort a list of prospects or competitors. Our domain authority checker reports a score with that context attached rather than a bare number, and the domain authority glossary entry explains the differences in one place.

How do you check backlinks for free?

Start with the two search engines, because their data is first-party and free, then use a third-party index to see what they will not show you.

Google Search Console, Links report. It lists top linked pages, top linking sites and top linking text for your verified property. Google says the report is a sample rather than a complete list, that tables are capped at 1,000 rows, that the export offers up to 100,000 rows split between “Latest links” and “More sample links”, and that it does not indicate whether a link is nofollow. It is still the only view of links Google itself has recorded. For a walkthrough of this and the other reports, see the Google Search Console SEO audit.

Bing Webmaster Tools, Backlinks report. Bing shows referring domains, anchor text and linking pages for your site, describes the data as a representative sample, and lets you add two other sites for a side-by-side comparison without verifying them. That comparison is the cheapest competitor backlink analysis available, and Bing’s index also feeds Copilot.

Third-party indexes. Every backlink tool crawls the web itself and will show links the two consoles do not. Free tiers vary in what they withhold; the table reflects the limits as published on each vendor’s own pages at the time of writing (vendor claims) and will drift, so confirm before relying on a number.

Source	Free without an account	What is withheld	Covers competitors
Google Search Console	Full report for your own property	Nofollow flags; links beyond the sample	No
Bing Webmaster Tools	Full report for your own property plus two comparison sites	Links beyond the sample	Yes, two at a time
Our backlink checker	Totals, authority score, top 10 referring domains and an anchor summary for any domain	Top 100 rows, CSV export and new or lost links are sent with the full report	Yes
Typical commercial checkers (Ahrefs, Semrush, Moz and others)	A handful of rows or a daily quota	The full list, history, filters and exports	Yes, on paid plans

Our view: for a site with fewer than a few thousand referring domains, the two consoles plus a free checker give you enough to make every decision in this guide. Paid tools earn their fee when you need history, alerts or a link list you can filter by hundreds of attributes.

How do you run a competitor backlink and link-gap analysis?

A link-gap analysis lists the sites that link to your competitors but not to you, then sorts them by how plausible it is that they would link to you too. It takes an hour with free tools and produces the only kind of link prospect list worth working: sites that have already shown they link to pages like yours.

1. Pick three competitors that rank where you want to. Use the pages that outrank you for your main queries, not the biggest brand in the industry. A site ten times your size is not a useful comparison.
2. Export their referring domains. Bing Webmaster Tools gives you two at a time; a backlink checker gives you the top referring domains for each.
3. Subtract your own. Any domain linking to two or more competitors and not to you goes on the list.
4. Classify each domain. Editorial (a journalist, a blogger, a resource page), structural (a directory, a partner page, a supplier list), or manufactured (a guest-post farm, a paid placement). Drop the third category; Google’s spam policies describe exactly those patterns.
5. Find the page that earned the link. A competitor’s calculator, dataset or definitive guide is what got linked. Your prospect list is really a content brief.
6. Score plausibility. Does the linking page still exist, is it still maintained, and does it link to more than one site in your space? Three yeses is a prospect.

Then look at the anchor text and the target pages across all three competitors. If 60 percent of their links land on one guide, that guide is the asset to beat. If their anchors are overwhelmingly brand names, so should yours be. The website audit crawl of a competitor’s site shows which pages they are pushing internally, which is often the same set.

Which backlinks are toxic, and should you disavow them?

“Toxic backlinks” is a vendor label, not a Google category. Google says most sites do not need the disavow tool, that it can harm your site if used incorrectly, and that it should be used only when you have a considerable number of spammy, artificial or low-quality links that have caused, or are likely to cause, a manual action. Random low-quality links you did not build are not a reason to disavow.

That position is stated plainly in Google’s disavow documentation, which also says to remove as many of the offending links as you can before resorting to the file. The tool exists for one scenario: you, or an agency you paid, built links that violate Google’s spam policies, and you now need to tell Google to ignore them. The policies define link spam as links created primarily to manipulate rankings, with examples that include buying or selling links for ranking purposes, excessive link exchanges, automated link programs, requiring a link in a terms of service, low-quality directory and bookmark links, keyword-rich widget links, and forum comments with optimised links.

A backlink audit therefore has a narrow job:

Pattern in your profile	What it suggests	Action
Hundreds of links from one network of sites with identical templates	A paid link network, possibly bought by a previous agency	Document it; disavow only if you also have or expect a manual action
Anchor text dominated by exact-match commercial phrases	Manufactured links	Same as above
Thousands of links from pages in languages and topics unrelated to yours, appearing suddenly	Scraper sites or a negative-SEO attempt	Google says its systems generally handle these; monitor, do not disavow
Links from hacked pages, adult or gambling sites you never contacted	Noise	Ignore
A vendor “toxicity” or spam score column	A model’s guess, not Google’s	Treat as a prompt to look, not as an instruction

If you do need to disavow, Google’s format rules are strict: one URL or domain:example.com entry per line, UTF-8 or ASCII, a .txt file, up to 100,000 lines and 2 MB, uploaded for a URL-prefix property rather than a Domain property. A new upload replaces the previous list, and Google says it can take a few weeks to incorporate. The toxic backlinks glossary entry summarises the decision in a paragraph.

Our view: the disavow tool is where more rankings have been lost through nervousness than gained through cleanup. If you did not build the links and you have no manual action, close the tab.

Do backlinks still matter in the AI era?

Yes for the indexes AI engines retrieve from, and less than mentions for whether a brand appears in the answer, on the evidence available so far. Links remain a core input to Google and Bing, and ChatGPT, Perplexity, Copilot and Google’s AI features all retrieve pages from an index before writing. But the strongest published correlation with AI visibility is being talked about, not being linked to.

The evidence to date:

• Research shows that in Ahrefs’ analysis of 75,000 brands, the factor with the strongest Spearman correlation to AI Overview visibility was branded web mentions at 0.664, followed by branded anchors at 0.527 and branded search volume at 0.392. Domain Rating scored 0.326, referring domains 0.295 and backlinks 0.218. The author stresses that correlation is not causation and that AI Overviews are still changing (Ahrefs’ AI Overview brand visibility study). A later Ahrefs update across ChatGPT, AI Mode and AI Overviews reported branded web mentions still correlating strongly, with YouTube mentions highest of all; treat those as the same kind of evidence.
• Google says, in its AI optimisation guide, that seeking artificial or inauthentic mentions across the web is not helpful because its ranking and spam systems focus on quality, and that you do not need special files or markup to appear in AI features (Google’s AI optimisation guide).
• No AI operator has published how it weights links. OpenAI, Anthropic and Perplexity document which crawler feeds their search index and how to block it; none describes a ranking formula.

What that means for a backlink analysis in 2026:

1. Keep counting referring domains, because they still move the Google and Bing rankings that put you in the retrieval pool.
2. Add unlinked brand mentions to the analysis. A review, a comparison article or a forum thread that names you without linking is now a measurable asset. Search for your brand in quotes, minus your own domain, and log the sources.
3. Prefer links and mentions from pages that AI engines actually cite. Run your main queries in ChatGPT search, Perplexity and Google’s AI Mode and note which domains are cited; those are the publications where a mention does double duty.
4. Stop paying for authority scores. A DR of 60 on a site no engine cites is worth less than a mention on a site that every engine cites.

The AI visibility guide covers how to measure citations platform by platform, and how to get cited by ChatGPT turns the mention evidence into an outreach plan. The AI citation and E-E-A-T entries explain why operators keep describing the same signals in different words.

A backlink action plan

This is the sequence we use, with the time each step takes for a typical small business or publisher site.

1. Baseline (20 minutes). Export Search Console links and the Bing backlinks report. Record referring domains, top linked pages and the top ten anchors. Run the backlink checker on your domain for the third-party view and the authority score, and save the date.
2. Health check (20 minutes). Scan the anchor text and the largest linking sites for the spam patterns in the table above. For almost every site, the conclusion is “nothing to do”, and that conclusion is worth writing down so the next person does not reopen it.
3. Target review (15 minutes). Compare the pages that attract links with the pages you want to rank. If they differ, the fix is usually internal: link from the linked pages to the money pages with descriptive anchors.
4. Competitor gap (45 minutes). The six-step link-gap process above with three competitors. Output: a list of twenty plausible sites and the content that earned each link.
5. Mention sweep (20 minutes). Search for your brand name, product names and founders without links. Log unlinked mentions on sites worth a link; a polite request converts a share of them.
6. Decide what to build (30 minutes). From steps 4 and 5, pick one asset to create or improve this quarter: the calculator, dataset, benchmark or definitive explainer that competitors’ links point at. Links follow assets, not outreach templates.
7. Repeat quarterly. Compare the new export with the baseline. Referring domains up, anchors still natural, no manual action: the profile is healthy.

What to do next

Run the free audit on the page you most want to rank; the report includes a backlink summary alongside the technical, on-page and AI-visibility checks so you can see whether links are the constraint or just the easiest thing to blame. Then do the baseline and health check above; forty minutes settles the toxicity question for most sites. If you want to understand how the AI engines decide what to cite before you spend a quarter on link building, read what generative engine optimisation actually involves.

Frequently asked questions

What is a backlink?

A backlink is a link from a page on one website to a page on another. Search engines treat it as a reference: a sign that someone else found the page worth pointing to. Links from many different sites carry more weight than many links from one site, and links marked nofollow, sponsored or ugc are treated as hints rather than endorsements.

How do I check backlinks for free?

Use the Links report in Google Search Console for your own site and the Backlinks report in Bing Webmaster Tools, which also lets you compare against two other sites. For any domain, free tiers of backlink checkers show totals, top referring domains and anchor text; the full link list usually sits behind a limit or a login.

Is domain authority a Google ranking factor?

No. Domain Authority is Moz's metric, Domain Rating is Ahrefs' metric and Authority Score is Semrush's. Each is a third-party estimate of link strength on a 0 to 100 scale, and Google's John Mueller has said repeatedly that Google does not use them. They are useful for comparing sites, not for predicting a specific ranking.

What are toxic backlinks?

Toxic backlinks is a label SEO tools give to links their algorithms consider spammy, such as links from link farms, hacked pages or irrelevant directories. Google's own documentation does not use the term. Google says its systems generally ignore links it considers manipulative, and that the disavow tool is only for sites with a manual action or a considerable number of spammy links likely to cause one.

Should I disavow bad links?

Usually not. Google says most sites do not need the disavow tool and that using it incorrectly can harm your performance. Disavow only if you have a manual action for unnatural links, or you bought or built spammy links at scale and expect one. Random low-quality links you did not create are not a reason.

Do backlinks still matter for AI search?

Less than they do for classic rankings, on the evidence so far. Research shows (an Ahrefs study of 75,000 brands) that branded web mentions correlated with AI Overview visibility at 0.664 while backlinks correlated at 0.218. That is correlation, not causation, and links still drive the Google and Bing indexes that AI engines retrieve from.