To conduct a useful backlink analysis, define the site and page scope, export Google Search Console data and any needed third-party crawler data separately, record each source and observation date, compare like with like, and investigate link context before approving an action. If a guide has fewer links than a competing page, first compare relevant referring domains to that specific page and check whether your guide is crawlable, useful, and aligned with the same search intent. A raw count is not a diagnosis.
This workflow helps you find patterns, recovery opportunities, and links that merit closer review. It cannot prove that backlinks caused a ranking change or identify a universal number of links a page needs.
What backlink analysis can and cannot tell you
Backlink analysis is the review of links pointing to a site or page, including their source pages, target URLs, context, attributes, and changes over time. A backlink is an observed link from one page to another. A referring domain is a linking domain: ten links from the same site count as ten backlinks but one referring domain.
Google uses link analysis, including PageRank, among many systems and signals used to understand pages and rank results. Links can also help Google discover pages. Backlink totals alone do not explain performance. Content usefulness, search intent, competition, and technical accessibility also matter. Google’s ranking systems overview and guidance on crawlable links and anchor text provide the relevant context.
A backlink count is an observation, not a diagnosis: record its scope, source, date, and comparison before deciding what it means.
Set the audit scope before opening a tool
Decide what you are measuring before exporting anything. An audit may cover a verified Search Console property, a domain, subdomain, URL prefix, directory, or one page. These scopes are not interchangeable. Record the exact scope and use the same one when comparing competitors.
- Choose competitors by search intent. Select pages or sites competing for the target queries, not simply the largest or most familiar brands. Record the query or reason each competitor belongs in the comparison.
- Set the period and observation date. Note the reporting period, export date, country or database setting where relevant, and filters applied. A tool’s first-seen date is an observation by that tool, not necessarily the date a publisher created the link.
- Check the page itself. For an underperforming URL, review crawlability, indexability, content usefulness, and search intent alongside its backlink profile. Google’s crawling guidance explains why link discovery and crawling are related but distinct.
A practical run definition contains the site or property, scope type and value, competitor list and rationale, reporting period, tool and database setting, filters, and export timestamp. If scope and period cannot be made reasonably consistent, label the comparison as directional rather than presenting it as a like-for-like benchmark.
Collect link data and keep its provenance
Use Google Search Console for a first-party report about a verified property. Its External links report includes top linked pages, top linking sites, and link text, and Google documents export to CSV or Google Sheets. It is not a competitor backlink database or a guaranteed inventory of every link Google knows about. See the Search Console External links report documentation.
Use a third-party crawler when you need its own backlink index, competitor research, or additional link-level fields. Semrush documents link status, dates, anchor text, attributes, filters, and CSV or Excel export in its Backlinks report guide. Ahrefs documents the distinction between backlinks and referring domains and provides backlink and referring-domain reports. Features, coverage, and access vary by tool and plan.
Different crawlers maintain different indexes and apply different discovery and normalization methods. Keep each export labeled by tool, property or scope, filters, filename or storage reference, and observation timestamp. Do not combine vendor totals into one supposedly authoritative count. Proprietary scores such as Authority Score or Toxicity Score are vendor measures, not Google scores or direct ranking factors.
Normalize records before comparing metrics
Keep three data grains separate. One backlink observation row represents one observed source-page-to-target-URL relationship from one source tool in one export run. One referring-domain row represents a normalized domain for that run and can summarize multiple backlink observations. One run record describes the export itself, including its scope and filters. Aggregates such as total backlinks or referring domains should be calculated from those records, not stored as if they were individual links.
Retain exported values alongside normalized ones so an analyst can trace a finding back to the source. Normalize source and target URLs, referring domains, link attributes, status, and timestamps consistently. Mark each field as observed, calculated, or inferred, and preserve a vendor’s definition when a metric is proprietary.
The following is an illustrative record format, not a vendor export schema. A stable vendor link ID is preferable when available. Otherwise, use the run ID, normalized source and target URLs, and a link-context identifier where a page can contain more than one link to the target.
{
"run_id": "illustrative-run-2026-10-12-01",
"source_tool": "third-party-crawler",
"observed_at": "2026-10-12T14:30:00Z",
"source_url": "https://publisher.example/resources",
"target_url": "https://example.com/guide",
"referring_domain": "publisher.example",
"link_context_id": "illustrative-placement-01",
"anchor_text": "link analysis guide",
"link_attribute": "unknown",
"status": "active"
}Before importing, validate the property and scope, export timestamp, column headers, URL parsing, allowed attribute values, and source-row reference. Preserve the untouched file. For a custom database, a proposed observation key is the source tool, run ID, normalized source URL, normalized target URL, and context identifier. A domain summary can use the source tool, run ID, and normalized referring domain. These are implementation-design examples, not published vendor requirements.
Do not rely on a lookup-then-create import as the only duplicate control. Two concurrent workers can both find no existing row and insert duplicates. Enforce the observation key with a database uniqueness constraint and use a transactional upsert or insert-on-conflict operation. Keep separate runs instead of silently overwriting earlier observations; that preserves changes in status, anchor text, and attributes over time.
Read volume, target pages, anchors, and context
Review backlink count, referring-domain count, and top linked pages together. A high backlink total can reflect repeated links from one site, including a sitewide placement, rather than broad support from distinct domains. Then ask which of your own pages attract links and whether they serve current content and acquisition priorities.
Inspect anchor text alongside the source page, link placement, topic, attributes, and surrounding pattern. Google recommends descriptive and relevant anchor text, but no single wording or percentage proves a violation. Follow, nofollow, sponsored, and UGC are distinct link qualifications, not a simple good-or-bad scale. A link may also have referral or audience value independent of how a search engine treats it.
An unusual country-code domain, shared IP or subnet, repeated anchor, or vendor flag can help prioritize inspection. None establishes common ownership, manipulation, negative SEO, or ranking harm on its own. Google defines link spam in terms of links created primarily to manipulate rankings in its spam policies. Context and evidence matter.
Use deterministic checks to flag missing URLs, repeated domains, unknown attributes, or unusual patterns. A person must inspect the source page, placement, intent, and site-owner involvement before classifying a link as editorial, paid, hacked, automated, or user-generated.
AI can optionally summarize a supplied page’s context or suggest a provisional category. Keep the original URL and fields attached, record the model and prompt version if AI is used, and require a human reviewer to decide whether the interpretation is accurate. Do not let a score or generated label trigger an accusation or action.
Investigate lost links and broken target URLs
A crawler marking a link as lost is a prompt to verify, not proof that a publisher removed it. Open the source page and confirm whether it still exists and links to the target. Then check the target’s HTTP response, redirect, canonical, indexability, historical purpose, and internal discovery paths.
- Restore the old page if its content remains valid and useful.
- Redirect to a relevant replacement if the old URL has genuinely moved or a suitable equivalent exists. Google’s redirect and canonicalization guidance does not mean every deleted URL should point to the homepage.
- Ask the publisher to correct the link when the source still links to an obsolete URL and a correction is appropriate.
- Make no change if there is no relevant replacement and restoring the content is not justified.
Record the evidence, decision, owner, and recheck date. A lost link may mean lost referral visits or a potential link signal. It does not, by itself, prove Google cannot crawl the site; check internal links, sitemaps, redirects, and other discovery paths before drawing that conclusion. Require a site owner’s approval before changing a redirect.
Compare competitors without treating crawler trends as ground truth
Compare the specific pages competing for the same search intent, not only whole-domain totals. In a third-party tool, look for relevant referring domains that link to several competing pages but not the target page. Review each prospect’s topic, editorial context, and realistic fit before adding it to an outreach or content queue.
Use one vendor, comparable URL scopes, the same database or country settings where available, consistent filters, and the same reporting period. New and lost link reports are trends observed by that crawler; different crawl frequency, index coverage, and date detection can distort apparent acquisition rates. Use the pattern as a lead for research, not as a precise record of when competitors published links.
A useful opportunity record contains the referring domain, competitor page, evidence URL, relevance rationale, proposed content or outreach action, owner, and review status. Prioritize relevance and a useful destination page over copying a competitor’s raw link count.
Choose tools for the decision you need to make
- Google Search Console: Start with the External links report for a verified property and export it to CSV or Google Sheets. Use it for Google’s reported view of your property, not competitor research.
- Ahrefs: Consider Site Explorer when you need its backlink and referring-domain reports or competitor research. Check the current plan and access information before standardizing on it; pricing and limits can change.
- Semrush: Its documented Backlinks reports include link-level fields, filters, and exports, and its tools support competitor and backlink-gap research. Confirm current account and plan access in the relevant documentation before relying on a particular report.
- SEOptimer: Its current backlink research page describes domain and link-level research. Check the displayed limits and plan details directly. Do not assume an older free-request limit still applies.
Select a tool by the question, scope controls, export fields, historical data, and repeatability of its reports. Do not choose it because a proprietary authority or toxicity score is presented as if it were a Google metric. Google’s guidance on evaluating third-party SEO advice is a useful standard for checking vendor claims.
Turn findings into a reviewed action plan
Separate routine monitoring, content and outreach opportunities, broken-link recovery, and possible link-spam policy issues into distinct queues. For each finding, save the evidence, recommended action, owner, approval status, and recheck date. This makes it clear whether the next step is to improve a page, contact a publisher, investigate a pattern, or do nothing.
Google says most sites do not need the disavow tool and warns that incorrect use can harm a site. Consider it primarily when there is a considerable number of artificial or spammy links and a manual action exists or is likely. For a manual action involving unnatural inbound links, follow Google’s specific guidance: try to remove violating links, document the effort, and disavow links that could not be removed. The general disavow guidance explains the tool’s cautions and conditions.
- Preserve and verify the source URL, target URL, anchor, observation date, and surrounding link pattern.
- Document the policy concern and the site owner’s involvement; do not treat an unfamiliar domain or vendor score as proof.
- Where Google’s manual-action guidance applies, record removal attempts and the links that could not be removed.
- Confirm the correct Search Console property and proposed file, then record the reviewer, decision, and recheck date.
Do not automate removal requests or disavow submissions from a heuristic or AI label. If a pattern may involve hacked pages, coordinated manipulation, or contractual obligations, escalate it to the appropriate SEO, security, or legal owner.
A repeatable audit is successful when another analyst can reproduce the comparison from the saved scope, source files, filters, and dates, and when each action has evidence, an owner, and a recorded outcome. If you need support designing a repeatable operations workflow, see ConsultEvo’s workflow and systems consulting overview.