Backlink analysis: reading your link profile and your competitors'

You run SEO or marketing for a B2B company and you have inherited a link profile, or you need to defend it before investing. A backlink analysis exists to support decisions: which links count, which to ignore, where your domain stands against its competitors and which pages lack support. It covers the first stage of the journey described in our netlinking guide: the diagnosis, before choosing publishers and running the campaign.

What the analysis should deliver: an actionable diagnosis

A useful analysis produces a decision, not just an export. Aim for an output that answers four questions:

• Strengths: which pages attract credible, relevant links, and why?
• Weaknesses: which strategic pages receive no external support?
• Risks: which signals suggest artificial or devalued links?
• Priorities: what should you recover, fix or strengthen first, with reasonable effort?

Formalise each finding as a three-part note: the signal (what you see), the hypothesis (what it means), the action (what you do and how you measure it). It is gaps that trigger action, not a single data point.

Clarify your scope: domain, subdomains and pages to strengthen

Before reading a single score, define the scope:

• the main domain and the subdomains included or excluded;
• the business-critical directories: solution pages, comparisons, contact, pillar guides;
• the pages that naturally earn citations: studies, data, glossaries.

Then separate the pages that receive links from those that should receive them: the gap gives you your first list of priorities.

Build a reliable inventory

The inventory starts from two sources. Google Search Console shows external links, referring sites and anchor texts: it is the reference on Google's side, without being exhaustive. A link index completes the picture with the metrics and history of each referring domain. Export both on a fixed date so you compare snapshots of the same scope.

The scarcity of links justifies this work: 94–95% of web pages have no backlinks at all (Backlinko, 2026). The inventory must tell you whether the right pages are the ones benefiting.

Avoid false positives

Before concluding that a link is lost or worthless, check what lies behind the raw data:

Finally, normalise the URLs, deduplicate by referring domain and attach each link to a target page: the dataset becomes segmentable by domain, source page, target page and attribute.

Reading the profile: dofollow referring domains, TF/CF ratio and topics

A profile is judged on the diversity and trust of the domains that link to you, not on volume. The Trust Flow, Citation Flow and Topical Trust Flow metrics are defined in the guide dedicated to backlinks; here, you read them to make decisions:

Third-party authority scores such as Domain Authority help you sort a list, not decide: they tell you neither whether the link is dofollow nor whether the domain covers your topic.

Read a distribution rather than an average

An average Trust Flow hides what matters. Read the distribution from three angles:

• The median reflects the typical level of your links, without being pulled up by two very strong media outlets.
• The extremes point to the trusted domains to protect and the weak areas to watch.
• Concentration measures your dependency: if a handful of domains carries most of the weight, losing one partner is enough to widen the gap.

The typical warning case combines a very low Trust Flow with a high Citation Flow: many links, little trust. This is not proof of toxicity, but a priority for manual review: context, indexing of the source page, anchors. Conversely, aligned topics, a high Trust Flow and a link in the body text signal a durable link.

Interpret nofollow links without a binary conclusion

Since March 2020, Google has treated nofollow, sponsored and ugc as hints, not directives. A share of nofollow is normal, and these links can bring traffic. A publisher that only places nofollow links does not count towards your dofollow referring domains, but it can count for GEO if AI engines cite it.

Anchors: the four-type profile and over-optimisation

Anchor analysis checks that your profile remains credible. Classify each anchor text into one of the four types:

• Brand: the name of the company or product;
• Partially optimised: a natural variant containing part of the keyword;
• Exact-match: the exact keyword of the target page;
• Generic: "read more", "here" or the naked URL.

No universal ratio applies. A natural profile is carried by brand and generic anchors. The more commercial the target page, the rarer exact-match anchors should be.

Patterns that trigger a review

Risk shows in repetition and dynamics, not in a single anchor. Trigger a review when you see:

• the same exact-match anchor across many domains, in a short period;
• identical commercial anchors on domains that look alike;
• an anchor unrelated to the paragraph that carries it;
• links concentrated on a single page, the homepage or a commercial page.

If you are unsure, document the signals and set a checkpoint before correcting anything.

Toxic links and disavow: rarely

Penguin has been part of Google's core algorithm and has run in real time since 2016. The spam policies target link schemes: undisclosed paid links, large-scale exchanges, PBNs, automated links. A weak link you did not solicit is documented and ignored. Detection relies on cumulative signals:

• repeated topical mismatch between sources and target pages;
• pages with no real content, saturated with outbound links;
• network footprints: same templates, same link blocks, same anchors;
• unjustified sitewide links;
• acquisition spikes with no explainable event (launch, press coverage, published study).

Prioritise risk

Prioritise by exposure (how many domains, pointing to which pages), by likelihood (clear patterns or an isolated case) and by stakes (business pages affected, concurrent ranking drop).

When the disavow tool is necessary

The Search Console disavow tool is reserved for two situations: a manual action notified by Google, or link schemes you created yourself, such as old undisclosed paid links. In all other cases, ignore. When it is necessary, the sequence remains cautious:

• document source URLs, anchors, target pages and dates of appearance;
• request removal when the site is legitimate and reachable;
• monitor links removed or changed;
• disavow what remains, at domain level, without disavowing broadly as a precaution.

You versus your competitors

Compare yourself with the domains that rank on your priority keywords, and read their pages that perform rather than just their homepage. The fictitious example below shows the expected reading:

In this example, your gap is measured against A and C, not against B, whose 420 domains reflect directories and networks. At URL level, the link target is set on the median number of dofollow referring domains among the top 10 organic results for its keyword, a method that belongs to link strategy.

Read referring domain gaps

The highest-yield pool lies with sites that already link to several competitors, but not to you: they have shown that they cite players in your sector. In this list, prioritise:

• recurring domains, present on two or more competitors;
• source pages close to your topics;
• easy opportunities: unlinked mentions, resources to update, URLs to correct.

Benchmarking is for discovering sources, not replicating a profile: a different angle is worth more than a copy.

Recover, fix, strengthen: the actions that follow

The analysis leads to three families of actions, ranked by increasing effort:

• Recover: lost quality links, unlinked brand mentions, URLs to have updated at the source.
• Fix: links to redirects or 404 pages, inconsistent canonicals.
• Strengthen: strategic pages without external support, strong links landing on a weak page to enrich, internal linking to redirect towards business pages.

Strengthening assumes a target page that is ready: a link does not make up for poorly served intent. Start with the pages that are already visible.

Testing a link: does it count?

A link present in a publisher's CMS may pass on nothing. Validate four prerequisites at publication, then regularly:

Test in batches, over a fixed reading window

A single link gets lost in the noise of the results. Instead, assess a batch of consistent links, published within a close window, with:

• a baseline before the first link: profile metrics, rankings, impressions;
• a fixed reading window after publication, long enough to show a trend;
• recorded variables: attribute, placement, anchor, target page, topic;
• a log of parallel events: redesign, seasonality, Google updates, cannibalisation.

Compare the page with itself, before its first link and today, speaking of correlation, never causation. Reading AI citations by question, over a 28-day window, belongs to backlink history.

Cadence and alerts: monthly check, quarterly review

Keep a single cadence: a monthly check of key indicators and a quarterly review of the full profile (TF distribution, anchors, competitor comparison). Between two checks, monitoring should alert you to:

• new referring domains, to check topic, anchor and attribute;
• lost links and deleted source pages;
• attribute changes, from dofollow to nofollow or sponsored;
• anchor drift, especially exact-match anchors pointing to commercial pages.

Connect alerts to pages

Connect each event to the pages concerned: rankings and impressions in Search Console, conversions in your analytics tool. Losing twenty weak links may have no effect; losing a key domain is handled first. Gains, losses and survival cohorts are covered in our guide to backlink history. For a one-off, complete review, the link building audit applies this analysis to a wider scope.

Depending on your situation

Depending on what the analysis has revealed, here is where to go next:

• You have identified your under-supported pages: set their link target and priority order with the URL-by-URL method.
• You need to explain the metrics to your team: start from the backlink definition and the metrics table.
• Your links are changing attribute: check the role of each rel value in our article on dofollow and nofollow links.
• You are looking for a tool to automate this monitoring: compare the expected features in our selection of netlinking tools.

To position your domain against its competitors without a spreadsheet, the Incremys Offsite media module displays the Trust Flow, Citation Flow and dofollow referring domains of each side by side. At the time of writing, its reporting also tracks the status of your links (live, lost, unverified), your anchor profile and your rankings.

FAQ on backlink analysis

How do you analyse your backlinks?

Analysing your backlinks means inventorying the links you receive, then reading them to make decisions. Cross-check Search Console with a link index, rule out false positives, then read dofollow referring domains, the TF/CF ratio, topics, anchors and attributes. Finish with the competitor comparison and a list of actions.

Which backlinks are toxic?

A backlink becomes a concern when it combines several signals: unrelated topic, page saturated with outbound links, network footprint, repeated exact-match anchor, unjustified sitewide link. An isolated weak link that you did not solicit is documented and ignored. The real risk comes from link schemes, especially those you created yourself.

Should you disavow links?

Rarely. The Search Console disavow tool is reserved for manual actions notified by Google and for link schemes you created yourself, such as old undisclosed paid links. In all other cases, you document and ignore, without disavowing broadly as a precaution.

How do you compare your link profile with your competitors'?

Compare your domain with the sites that rank on your priority keywords, on three values: Trust Flow, Citation Flow and dofollow referring domains. Read each one's TF/CF ratio before volume: a competitor with high volume and no trust is not a model. Finally, spot the domains that link to several competitors but not to you.

How do you test whether a backlink counts?

A backlink counts if the source page is indexed and the link appears in the rendered code, with the planned attribute and anchor. Check these points at publication, then regularly. To measure an effect, test a batch of consistent links over a fixed reading window, comparing each page with itself before and after.

How often should you analyse your backlinks?

A monthly check of key indicators and a quarterly review of the full profile are enough for most B2B sites. The monthly check covers new domains, lost links, attribute changes and anchors. The quarterly review revisits the Trust Flow distribution and the comparison with competitors.

What is the difference between a backlink analysis and a link building audit?

Backlink analysis is a method for continuously reading your link profile. A link building audit is a one-off, delivered review that applies this analysis to a wider scope and results in a complete action plan.

How do you interpret an unbalanced TF/CF distribution?

A distribution dominated by domains with a low Trust Flow and a high Citation Flow signals unreliable volume: pages saturated with links, directories, networks. This is not proof of toxicity, but a priority for manual review. Check the context, the indexing of source pages and the anchors before taking any action.