Internal linking tools fall into three distinct categories that get lumped together and shouldn’t be: crawlers that map and diagnose your existing link structure, suggestion engines that recommend where new links should go, and automation plugins that insert links into published content on a rule. They solve different problems, carry different risks, and most sites need the first two long before the third.
The reason internal links matter is mechanical. They are how crawlers discover pages, how authority flows from your strong pages to your weak ones, and how you tell Google which page owns which topic. A page nobody links to internally is a page you have told Google not to care about.
Category one: crawlers and link mappers
These walk your site the way a search engine does and produce an inventory of what links to what. The outputs that actually change decisions:
- Orphan pages — URLs in your sitemap with zero internal links pointing at them. Almost every site over 200 pages has some, usually old campaign landing pages or paginated archive leftovers.
- Inlink counts per URL — how many internal links each page receives. Sort ascending and the bottom of that list is your work queue.
- Crawl depth — clicks from the homepage. Anything past four clicks gets crawled less often and ranks worse for it.
- Broken internal links — 404s and redirect chains, both of which waste crawl budget and leak the value they were meant to pass.
- Anchor text inventory — every anchor pointing at each page, which is how you catch fifteen different pages all linked as “click here.”
Start here regardless of what else you use. You cannot fix a structure you have not measured, and the diagnosis usually reveals that the problem is concentrated in a template rather than spread across individual posts.
Category two: link suggestion engines
These scan your content, find topical relationships between pages, and propose specific links: put a link on this phrase in this post, pointing at that page. Good ones rank suggestions by relevance and show you the sentence in context so you can judge it.
Suggestions are the sweet spot for most teams. You get the scale benefit — a tool can consider 400 pages in seconds, which no editor will do — while a human still approves each placement. Budget roughly two seconds per suggestion to review; a 300-suggestion queue is an afternoon, not a project.
The failure mode is over-acceptance. Approving every suggestion produces pages with 40 internal links in the body, which dilutes every one of them and reads terribly. Cap it at five to eight contextual body links per 1,000 words and prioritize the ones pointing at pages that currently have almost no inlinks.
Category three: automated insertion
Automation plugins take a keyword-to-URL mapping and insert links wherever that phrase appears, sitewide, without review. They are fast and they are the riskiest option, for reasons worth being specific about.
Exact-match keyword rules produce unnatural anchor distributions — the same commercial phrase linking to the same page from 200 posts is a pattern, and patterns like that have been devalued before. Blind matching also links words inside quotations, code samples, and headings where a link makes no sense. And when you change a URL, an automation rule can rewrite hundreds of pages at once; when it goes wrong, it goes wrong everywhere simultaneously.
If you do automate, constrain it hard: limit to two or three automated links per page, vary the anchor text, exclude headings and quotes, and never let a rule fire on more pages than you are willing to manually spot-check.
The structure principles the tools cannot decide for you
No tool knows your business. These decisions are yours:
- Decide which pages are meant to win. Usually ten to thirty commercial or pillar pages. They should receive the most internal links; everything else feeds them.
- Link down and across, then back up. Pillars link out to supporting articles, supporting articles link back to the pillar and sideways to genuine siblings.
- Keep everything within three clicks of the homepage where the site size allows it.
- Write descriptive anchors. “Backlink gap analysis” tells Google what the destination is about. “Learn more” tells it nothing.
- Vary the anchor text across links pointing at the same page. Natural profiles are varied; identical anchors at scale are a footprint.
- Link from the body, not just the nav. Sitewide navigation links are discounted precisely because every page has them. In-content links carry more weight.
- Link when it helps the reader. If the link would not be useful to a person mid-sentence, do not place it.
A practical audit you can run this week
Crawl the site and export every URL with its inlink count. Sort ascending. Anything with zero inlinks that you actually want ranking gets three to five contextual links from relevant existing posts, added by hand — this alone often produces movement within a few weeks, because you have taken pages Google barely knew about and put them on the map.
Next, pull your top twenty pages by organic traffic. These are your authority reservoirs. Check what each one links to internally; if a page pulling 4,000 visits a month links only to your contact page, you are wasting it. Add two or three links from each into pages that need help.
Finally, fix the plumbing: every internal link pointing at a redirect should be updated to the final URL, and every internal 404 should be repointed or removed. This is tedious and it is the highest-certainty item on the list.
Where deterministic linking beats suggestions
There is a fourth approach that sits between suggestions and blind automation: rule-based linking applied by deterministic code at publish time, against a known map of your pages, with an explicit cap. It is not a plugin firing on string matches across the whole site — it runs once per article, from a defined set of destinations, before the page goes live.
That is how SEO Rocket handles it. The AI writer produces the draft with no links in it at all, then a separate deterministic step applies internal links from the project’s own page map. The split exists on purpose: language models are unreliable at remembering which URLs exist and will happily invent one, so the writing and the linking are kept apart. The AI writes; deterministic code decides what publishes.
Measuring whether it worked
Give it four to eight weeks and look at three things in order. First, crawl stats in Search Console — pages that were previously ignored should start getting crawled. Second, impressions on the newly linked pages, which typically move before positions do. Third, position changes, judged as a trend across weeks rather than day to day, since ±2 to 3 positions of daily movement is normal noise and will otherwise convince you of results that are not there.
Internal linking is one of the few levers you control completely — no outreach, no budget, no waiting on anyone. It is also the one most sites never systematically work through, which is exactly why it is usually the fastest win available on an established site.
