Internal linking audit: the 6-step method and the action plan

An internal linking audit verifies, using evidence, that every useful page can be reached, that every topic remains readable and that anchors accurately announce their target. The rules to apply are set out in the guide to internal linking and its structural rules; here is the six-step method to check what already exists, turn your findings into a prioritised action plan and measure the effect of your fixes.

What an internal linking audit must establish

A linking strategy describes intent: which pages link to which pages, with which anchors. An audit measures the current state, highlights gaps, and turns those gaps into an action plan.

The four starting questions

Before running a single export, set the questions the audit must answer:

• Can pages be reached? Every useful page receives at least one internal link and sits a few clicks from the homepage and from the pillar of its cluster.
• Are topics readable? The pillar links to all its zooms, each zoom links to its pillar, and pages covering the same topic link to each other.
• Are anchors consistent? The same intent points to the same page, with anchors that describe what the target promises.
• Do links work? No internal link leads to a 404 or to a redirect.

The scope covers URLs reachable via HTML links, templates (menu, footer, breadcrumbs, automated blocks) and Google Search Console reports.

The metrics to track

Each question translates into simple metrics. The table below shows, for each one, what it measures, how to read it and the action it triggers:

Step 1: crawl the site and model the link graph

Everything starts with a crawl: a bot moves through the site as Google would, following HTML links from the homepage and the main sections. What matters most is repeatability: being able to compare two audits run a few months apart.

What to export

The crawl must produce a usable inventory of relationships: source page → anchor → target page. For each URL, export:

• The canonical URL and the page type (pillar, zoom, product or service page, conversion page, utility page).
• Depth in clicks from the crawl's starting point.
• Internal incoming and outgoing links, with the area they come from (main content, menu, footer, automated block).
• The anchor text of each link.
• The HTTP status of the page and of each linked target.

Check the technical form of links at the same time: an a tag with an href attribute, no JavaScript link obfuscation and no nofollow on internal links. A link the bot does not follow does not exist for it.

Reading the graph: contextual links and template links

Once the relationships are extracted, model the site as a graph: pages are the nodes, links are the edges, and the anchor is each edge's attribute. Filter by section: a cloud of thousands of points answers no question. Two families of links change how you read it:

• Contextual links, placed in the body of the text: they tell Google and AI systems what the target is about.
• Template links (menu, footer, breadcrumbs, "recent posts" lists): they count for the architecture, but do not replace contextual links.
• Automated module links ("related content"): they need rules when they connect pages with no topical relationship.

Step 2: find orphan and weakly linked pages

An orphan page receives no internal links. It is discovered only through the sitemap or external links. It is either a defect to fix or a sign that the page has no place on the site. Orphaned answer pages are also invisible to generative AI, which relies on pages it can reach.

Cross-check three signals

A single export is not enough to detect orphan pages reliably. Cross-check:

• The graph: nodes with no internal incoming links.
• Google Search Console: URLs that receive impressions or appear in the sitemap but that the crawl does not reach.
• Audience data: pages viewed only as entry points (email, campaign, direct access), with no internal navigation leading to them.

Also identify weakly linked pages, whose few incoming links all come from navigation.

Link, merge or remove

Each orphan page calls for a decision, not an automatic link:

• Link the useful page from its pillar and from pages on the same topic, with a descriptive anchor.
• Merge a page that repeats another page into the one that answers best, then redirect.
• Remove a page with no value for the reader, or exclude it from indexing if the site still needs it to function.

Step 3: measure depth from the homepage and from each pillar

Depth measures the number of clicks needed to reach a page. There is no universal threshold: the rule is that every useful page can be reached from the pillar of its cluster in a few clicks. A deep page is not a problem in itself; a deep priority page is.

Measure depth from two places. From the homepage, it shows how the bot discovers the page, and depends mainly on navigation. From the pillar of each cluster, it shows whether the topic is readable: a zoom that its pillar does not link to directly is a hole in the structure. Finally, compare average depth by page type.

Also analyse dominant paths. If a priority page can only be reached through a chain of paginated lists or a "recent posts" block that changes with every publication, its access is fragile. Create a stable path from the pillar or from a closely related page on the same topic.

Step 4: check anchors and self-links

The anchor tells Google and the reader what the target page promises. Export the anchors received by each priority URL and look for four red flags:

• Generic anchors ("click here", "learn more"): they give no information about the target.
• Repeated exact-match keyword used identically across dozens of links: the anchor should describe the page, not hammer a query.
• Intent mismatch: the anchor promises content the target page does not deliver.
• Anchor shared between two targets: the same wording points sometimes to one page, sometimes to another, and muddies the signal.

The last case explains a common situation: the wrong URL ranks for the query. This is SERP realignment. The fix is not to remove links: you add links with a consistent anchor to the right page, and clarify the topic of the page that ranks by mistake.

Self-links and duplicate links

Two anomalies easily go unnoticed. A self-link is a page that links to itself, directly or via a URL redirected to it: it gives the reader nothing and signals a forgotten redirect. A duplicate link is the same target linked several times in one article or in the same block (image, title and button): the rule is one link per target page per article, placed where it helps.

Step 5: compare crawl and index

The crawl shows what your links make reachable; the Search Console Pages report shows what Google has kept. The gap between the two is a structural health signal. To find out why a URL is not indexed, the URL Inspection tool and the guide to indexing in Google Search Console explain the steps to follow.

Reading the crawl vs index gap

Three scenarios arise, and they do not call for the same fix:

• Many URLs crawled, few indexed: internal linking exposes low-value pages (facets, tags, internal search, pagination). Fix the templates that link to them and check the rules in your robots.txt file and what it actually blocks.
• Indexed URLs the crawl barely reaches: they rely on the sitemap or external links. They lack contextual links from pages on their topic.
• Useful URLs neither crawled nor indexed: orphan pages, to be handled with step 2.

Group URLs by page type: you can see which template is driving the gap, and a single fix resolves hundreds of cases.

404s and redirects segmented by origin

Internal 404s typically appear after page deletions, URL changes, redesigns, or simple typing mistakes. Links to redirects, especially chains, slow down crawling. Segment them by where they originate: editorial content, menu, footer, automated blocks. A template error repeats on every page that uses the template; it comes first.

In every case, fix the link at the source: replace it with the final URL rather than relying on the redirect. If the page no longer exists, point to the page that answers the same intent.

Step 6: prioritise and plan the fixes

The first five steps produce findings; the sixth turns them into sequenced fixes, each with an owner and a date.

From finding to fix

Most audit findings come down to a few recurring causes. This table matches each finding with its likely cause and the fix that solves the problem at the source:

Weigh impact against effort

Priority goes to the pages that carry your business: 91% of pages never reach Google's first page (SEO.com, 2026), and other SEO statistics confirm that it is better to concentrate effort than to spread it. Rank each fix by its impact (page value, number of pages affected) and its effort (editorial update or template change):

Finally, separate hygiene fixes (404s, redirects, broken links), which you handle without debate, from structural fixes (depth, anchors, parent-child links), which go through the linking plan and the brief for each piece of content.

Control routine and measuring the effect

Internal linking degrades as the site grows: every publication, every deletion and every redesign creates new gaps.

The control routine

Spread checks by frequency, each with its data source:

Measure the effect by page group

An internal linking project is measured by page group, never page by page: a cluster, a section, a page type. As 60% of searches end without a click (Semrush, 2025), look at impressions as much as clicks. For each group, compare a period before and a period after the fix:

• Indexed pages and crawled pages, in the Search Console Pages report.
• Impressions, clicks and average position, filtered on the group's URLs in the Performance report.
• Internal links received by priority pages, in the Links report or in the crawl.
• The right URL ranking on queries affected by a SERP realignment.

Give Google time to recrawl and note any other changes made during the period. An improvement after a fix is a correlation, not proof of causation.

Depending on your situation

An internal linking audit is part of wider work. Depending on your starting point:

• If your clusters still need to be built or redesigned, start from the method for the semantic cocoon and its pillar-zoom structure.
• If internal linking is only one part of a wider diagnosis, follow the approach for a full website SEO audit.
• If the audit mainly reveals crawling and indexing issues, go further with a technical SEO audit.
• If you need to get to grips with the Links, Pages and Performance reports, read the guide to Google Search Console.
• If your priority pages lack authority despite clean internal linking, the answer lies in external links and netlinking.

To map your pages and keywords, find orphan pages and measure the depth of each cluster without assembling exports by hand, the 360 SEO & GEO Audit module of the Incremys GEO & SEO platform brings these checks together in a single diagnosis.

FAQ: internal linking audits

Which tool should you use to audit internal linking?

Two families of tools are enough for a first audit. A crawler rebuilds the internal link graph, with anchors, click depth and the HTTP status of each URL. Google Search Console provides Google's own view: the Links, Pages and Performance reports, plus the URL Inspection tool.

How often should you run an internal linking audit?

A full audit every quarter suits most sites that publish regularly, complemented by a check with every publication and a monthly review of internal 404s, links to redirects and newly orphaned pages. After a redesign or migration, run every step again without waiting for the scheduled date.

How do you find orphan pages reliably?

Cross-check three sources, because none is enough on its own. The crawl lists the pages it knows that have no incoming link. Search Console reveals URLs that receive impressions or appear in the sitemap without being reached by the crawl. Audience data shows pages viewed only as entry points, with no internal navigation leading to them.

Is a deep page always a problem?

No. Depth becomes a problem when it affects a priority page: a flagship service, a pillar, a conversion page. An archive or support page can stay far from the homepage if its path is logical. The rule is about access: every useful page must be reachable from the pillar of its cluster in a few clicks.

How do you measure the effect of an internal linking project?

Measure by page group, with a period before and a period after. In Search Console, track the group's indexed pages, impressions, clicks and average position, as well as the internal links received by priority pages. Note any other changes made during the period: an increase remains a correlation, not proof.

Should you remove internal links to strengthen a page?

No. You strengthen a page by adding useful links to it, with an anchor that describes what it promises, from pages covering its topic. Removing links to steer internal PageRank is PageRank sculpting, which the audit does not recommend. Only remove links that are broken, duplicated or unrelated to the topic.

What should you do when Google ranks the wrong page for a query?

Check the anchors first: the same wording pointing to two pages muddies the signal. Then add links with a consistent anchor to the right page, from pages on the same topic, and clarify the topic of the page that ranks by mistake.