Links report: complete guide

Review sampled external links, anchor text, top-linked pages, and internal link targets.

GSC UI required Intermediate

The Links report is Google's sampled view of links it has found to and within the selected Search Console property. It can answer questions such as which canonical pages receive many external or internal links, which root domains link most often, and which external anchor text appears most often.

It is not a complete backlink index, live-link checker, link-quality score, orphan-page detector, or substitute for a site crawl. The most important rule is to use it for patterns and investigation leads, not exact link inventory or automatic SEO verdicts.

This guide reflects Google's documentation on July 18, 2026.

What the report contains

The landing page separates external and internal links.

Area	What it groups	A useful question
External: Top linked pages	Canonical pages in the property	Are the pages earning recognition the ones we expect?
External: Top linking sites	Linking sites, grouped to root domain	Which domains account for many discovered links?
External: Top linking text	External link text, grouped by case-insensitive exact match	How do other sites commonly describe this site?
Internal: Top linked pages	Canonical pages in the property	Which pages are prominent in Google's sampled internal-link view?

Click More to expand a table. A row can lead to a narrower drilldown:

From an externally linked page, open the sites linking to that page.
From a linking site, open the pages it links to on your property.
Continue to the source pages found for a site-to-target combination.
From an internally linked page, open sampled source pages within the property.

These paths are useful for validating examples. They do not turn the sample into a complete list.

Start with the correct property

In this report, an external link is a link from outside the current property. Property scope therefore changes the classification.

For example, a link from blog.example.com to www.example.com/product is internal to a Domain property covering example.com, but it can be external to a URL-prefix property limited to https://www.example.com/. Record the property type and exact scope before comparing exports or investigating surprising domains.

Google displays linking sites by root domain: protocol, subdomain, and directory are removed. Links from www.example.org, news.example.org, and http://example.org/path/ can all appear under example.org. Top-level domains remain distinct, so example.com and example.com.de are not combined.

This root-domain presentation has two consequences:

A domain row can blend several independently operated or technically different subdomains.
A domain identical to your own can appear in a narrower property because its linking subdomain was stripped from the display.

Drill down to source URLs before interpreting ownership or intent.

The data is sampled, grouped, and historical

Google documents four behaviors that prevent the visible counts from being treated as an exact inventory.

It is a sample

The report is not a comprehensive list of every internal or external link Google knows. Google may omit URLs for reasons including non-indexed or deduplicated pages. Absence means “not present in this sample,” not “Google has never seen this link.”

Interface tables stop at 1,000 rows

Each table is limited to 1,000 rows and can be truncated on large sites. A page outside the table is not necessarily unlinked, and the last visible row is not a reliable boundary between linked and unlinked pages.

Targets are grouped by canonical URL

Pages on the property are grouped by canonical URL. Parameters and anchors may be omitted during grouping. A backlink that points to a tracking URL, redirecting legacy URL, or duplicate may therefore be reported under a different canonical target.

Google's detailed source table can show a Target URL (if different). A different target can reflect a redirect from the linked URL or Google's canonical selection. Verify the actual link, redirect chain, and indexed canonical before asking a publisher to change anything.

Duplicate links are combined

Google combines links from the same source URL to the same target URL after removing parameters and performing other normalization. Link-text duplicates follow link-text rules instead: case-insensitive exact matches are combined.

Counts are therefore normalized report counts, not raw <a> element counts from HTML. Repeated navigation links, parameters, duplicate sources, and canonical consolidation can all make a crawler's raw count differ.

Old links can remain

The report includes links Google found over time. A displayed source link may have been removed, and its source page may no longer exist. Search Console does not provide a dependable live-status or last-seen field for each reported link.

Always visit or crawl important source examples before treating them as current.

Analyze externally linked pages

Open External links → Top linked pages → More and classify the visible targets by page purpose or template.

Look for:

original research, tools, guides, and other intended link-worthy assets;
the homepage and major brand or product hubs;
old URLs whose signals should consolidate to a maintained destination;
parameter or protocol variants that suggest canonical inconsistency;
expired campaigns, discontinued products, or 404 targets still earning links;
unexpected targets created by hacked or user-generated content.

A useful review table is:

Target	Expected role	Current response	User canonical	Google canonical	Action
`/research/`	Linkable resource	`200`	Self	Self	Maintain and link internally
`/old-study`	Retired asset	`404`	—	Unknown	Restore or redirect if a close replacement exists
`?utm_source=...`	Tracking variant	`200`	Clean URL	Clean URL	Confirm links consolidate as intended

High counts do not prove ranking value, quality, or authority. Low counts do not prove a page is weak. Use the table to find targets worth validating, then bring in relevance, live status, editorial context, traffic, and business value.

Analyze top linking sites

Use Top linking sites to recognize broad patterns:

publishers, partners, associations, suppliers, clients, and community sites you expect;
a legitimate domain producing many sitewide links from navigation or templates;
old campaign or syndication relationships;
unfamiliar domains worth a small source-URL sample;
concentration in one source that creates reporting noise or business dependency.

The count describes reported links, not unique editorial endorsements. One domain can generate thousands of template links while another contributes one highly relevant citation. Do not rank prospects or risk from count alone.

For any important or suspicious row:

Open the linked targets for the domain.
Open a sample of source pages.
Confirm whether the pages and links are live.
Record context, placement, target, redirect behavior, and apparent intent.
Use other evidence before deciding whether outreach, remediation, or no action is appropriate.

Interpret top linking text

The link-text table covers external links only. Google groups text using case-insensitive exact matching: capitalization variants combine, but even a small wording difference creates a separate row. Links without text, including image links with no usable text in this report, appear as (empty) in the interface and as empty rows in exported data.

Use the table to check whether the sample broadly reflects:

the brand and common brand variants;
page, product, person, or publication names;
descriptive language aligned with the destination;
old names after a rebrand;
irrelevant or spam-like phrases that warrant source investigation.

This table cannot map every anchor phrase to every destination from its landing view, measure anchor diversity reliably, or show whether text was deliberately optimized. It also cannot show nofollow, ugc, or sponsored attributes. Never infer a penalty from an odd phrase alone.

Google recommends descriptive, reasonably concise, natural anchor text. For links you control, especially internal links, improve clarity for users rather than forcing keyword variations.

Interpret internal links

The internal table shows which canonical pages receive many links from pages within the property sample. It can reveal useful contrasts:

important conversion or informational pages with unexpectedly low representation;
utility, tag, faceted, or pagination pages dominating the sample;
retired or redirected URLs still used in navigation or templates;
a hub-and-detail structure that appears aligned—or misaligned—with site priorities.

Do not convert the ranking into a quota. Google says there is no magical ideal number of links on a page. Context, crawlability, information architecture, and the usefulness of each connection matter more than reaching a universal count.

Every page you care about should have a crawlable link from at least one other page. For links to be reliably crawlable, Google recommends an <a> element with an href that resolves to a real web address. Important navigation implemented only through click handlers may not be reliably extracted.

Why this cannot find orphan pages reliably

An orphan page is normally a page in a known URL inventory with no crawlable incoming internal link. Search Console provides a sampled list of observed links; it does not provide the complete URL inventory or a guaranteed complete internal-link graph.

Therefore:

absence from the report is not proof of zero links;
absence from a crawler is not proof the URL does not exist;
a sitemap URL can still lack internal links;
a URL can be linked only after rendering or only in certain states.

Use a complete site crawl plus XML sitemaps, analytics or log URLs, CMS/database exports, and other known inventories. Then calculate incoming crawlable links across the combined set. Use URL Inspection on representative anomalies to compare Google's indexed canonical and rendered view.

Export the data correctly

The export limit depends on where the export starts.

Export location	Available data	Documented maximum
Links report landing page	Latest links	100,000 rows
Links report landing page	More sample links	100,000 rows
A page containing one expanded table	That table	1,000 rows

Latest links contains recently discovered links sorted by discovery date. It is not a “links still live” export. More sample links is another sample drawn from Google's known links and can expose examples omitted when the latest-links output reaches its limit.

Exports are available as CSV or Google Sheets. Sorting a single table in the interface does not change the exported order. Preserve the export date, property, source slice, and any drilldown path with the file; otherwise two exports can look comparable when they are not.

For recurring review, compare sets of source-target pairs rather than treating count changes as a precise gained/lost-link metric. Sampling, canonicalization, normalization, recrawling, and property changes can all alter an export without a publisher intentionally adding or removing a link.

Why other backlink tools disagree

Search Console and third-party crawlers operate different systems. Differences can result from:

different crawl coverage, discovery sources, and revisit schedules;
Google's sampling and 1,000-row interface tables;
canonical target grouping and URL normalization;
root-domain grouping in the interface;
treatment of redirects, parameters, duplicate links, and non-indexed sources;
historical links that are no longer live;
each provider's own definitions and filters.

Neither a higher nor lower total establishes which tool is “correct.” Define the question first. Use Search Console for Google's sampled property view, a link crawler for a provider's discovered graph and link attributes, and direct validation for high-impact decisions.

The report is not a link audit

Search Console does not provide, per reported link:

a guarantee that it is currently live;
nofollow, ugc, or sponsored attributes;
placement, surrounding text, or editorial context;
a quality, authority, risk, or relevance score;
referral traffic or conversions;
a complete first-seen/last-seen history;
every source, target, or internal link;
competitor backlink data.

It also has no supported Search Console API endpoint for downloading the Links report. The Search Analytics API and URL Inspection API do not expose backlink tables. UI export is required; broader automation needs an external crawler or link-data provider.

Do not disavow from the report alone

An unfamiliar domain, strange anchor phrase, or large link count is not by itself a reason to disavow. Google describes disavowal as an advanced feature that can harm Search performance if used incorrectly and says most sites do not need it.

Google's documented threshold is narrow: consider it when there is a considerable number of spammy, artificial, or low-quality links and those links caused, or are likely to cause, a manual action. First try to remove links you or a hired party created in violation of spam policies.

Before any disavow change:

check the Manual Actions report;
investigate how the links were acquired;
validate current source examples;
preserve the existing disavow file, because a new upload replaces it;
obtain experienced review of the evidence and property scope.

Disavowed links can continue to appear in the Links report, so the report is not a confirmation that a disavow “worked.”

A repeatable review workflow

Record the exact property and export date.
Export Latest links and More sample links from the landing page.
Review top external targets by page type and current URL state.
Review source domains, then validate examples behind unusual rows.
Check link text for brand accuracy, rebrand residue, and investigation leads.
Review internal targets against a business-priority URL list.
Crawl the site for complete internal-link and orphan analysis.
Inspect important redirect or canonical anomalies with URL Inspection.
Assign only evidence-backed actions: repair, redirect, update internal links, outreach, investigate, or monitor.
Preserve the evidence and repeat on a defined cadence.

Common mistakes

Treating the first 1,000 rows as the full link graph.
Calling a missing URL an orphan without crawling a complete URL inventory.
Assuming a reported link still exists.
Comparing different property scopes.
Interpreting a root-domain row as one specific subdomain.
Reading link counts as ranking value or unique endorsements.
Forgetting that targets are grouped by canonical URL.
Treating Latest links as gained links or a live-status feed.
Expecting interface sort order to control an export.
Uploading a disavow file because unfamiliar domains appeared.