The Internal Linking Graph Optimizer is a free GrowthGPT tool that takes your page inventory and a pillar URL and returns an entity-based linking plan: a silo map, ready-to-paste link recommendations with anchor text and context, a click depth analysis, and a list of orphan pages with the exact sources that should link to them.
Internal linking is the highest-leverage SEO work almost nobody schedules. It needs no new content, no outreach, and no budget, and it changes what crawlers reach, how quickly, and which of your pages look important. This guide covers preparing the inventory, choosing the right navigation pattern, and reading each of the four outputs the Internal Linking Graph Optimizer returns. For the strategy behind it, pair this with internal linking for AI search.
Image: A node graph of site pages grouped into topical silos, with a pillar node at the centre, dotted lines showing recommended new links, and two disconnected orphan nodes highlighted
What does the Internal Linking Graph Optimizer produce?
Four outputs, plus a summary bar showing total pages, links recommended, orphan pages, average click depth, and silo coverage.
- Entity silo map. Your pages grouped into topical silos, each page tagged with its role in the silo and the entities it covers. This is where you discover that two silos are really one, or that a silo has a pillar and no support.
- Link recommendations. Each one names a source page, a target page, the anchor text to use, the semantic reasoning behind the pairing, surrounding context copy, a placement position, and a priority. The recommendations export as CSV.
- Click depth analysis. Current depth versus optimized depth per page, with the specific fix that closes the gap.
- Orphan pages. Pages with no inbound internal links, each with recommended source pages and the anchor text to use.
How do I run it?
Step 1: Build the inventory
Paste one page per line as a URL followed by its topic. The topic label matters as much as the URL, because the silo grouping and the anchor text suggestions are built from it. Write real topics rather than page titles: “SaaS pricing strategy” is useful, “Blog post 14” is not.
If you do not have a list, crawl the site first with the XML Sitemap Generator, which gives you every page a link-following crawler can reach along with its current depth. Twenty to eighty pages is the useful range for one run. Include product, pricing, and service pages, not just posts.
Step 2: Name the pillar and its topic
The pillar URL is the page you want the silo to concentrate authority on. Choose the page that should rank and convert for the broadest term in the cluster, not necessarily the longest article. The pillar topic tells the tool which entity space this silo occupies.
Step 3: Pick the navigation pattern
| Pattern | Shape | Fits |
|---|---|---|
| Hub & Spoke | One central page, supporting pages linking into it | A single dominant topic or product line |
| Flat Silo | Several parallel categories, minimal cross-linking | Distinct audiences or product lines that should not blur |
| Topic Cluster | Pillar plus cluster pages, dense links inside each cluster | Content libraries and editorial sites |
| Hybrid | Commercial core with an editorial layer feeding it | SaaS and service businesses with an active blog |
Step 4: Set a focus goal
The optional focus goal steers the plan. Something like “boost the pricing page” or “lift the comparison cluster” changes which links get recommended and how priorities are ranked. Without it you get a balanced plan, which is fine for a first pass and less useful when you have one page that has to move.
Step 5: Read the summary before the detail
Two numbers tell you most of what you need. Average click depth above three means important pages are buried. Any orphan count above zero means you have content that is currently invisible to link-following discovery, which is the fastest thing on the list to fix.
How do I implement the recommendations without breaking things?
Work in the order the priorities suggest, and resist the urge to paste every recommendation in one afternoon.
- Orphans first. Each one is a page that already exists and currently receives nothing. Add the recommended links from the suggested sources and the page enters the graph immediately.
- Then depth. Apply the click depth fixes for commercial pages sitting at depth three or deeper. Usually one link from a hub page does it.
- Then the high-priority recommendations, placed in the body where the context copy suggests. Links inside relevant prose carry more weight than a related-posts block at the foot of the page.
- Vary the anchor text. Use the suggested anchors as written, and do not point twenty pages at the same target with the same phrase. Descriptive variety is what teaches a machine what the target page is about.
- Re-crawl and re-measure. After the changes are live, run the crawl again and confirm depth actually fell. Plans are easy, verification is the part that gets skipped.
Google’s Search Essentials state the underlying rule plainly: pages need to be reachable through links a crawler can follow. Everything in this tool is an application of that.
Why is the plan entity-based rather than keyword-based?
Because retrieval works on meaning. A silo built around an entity teaches both search and AI systems that your site covers a subject completely, while a silo built around keyword strings produces pages that repeat each other and links whose anchors all say the same thing. Entity SEO covers the model, and the Entity Cluster tool shows which entities a topic implies so you can see what your silo is missing before you start linking.
This is also why internal linking pays off in AI answers. Google’s AI features documentation confirms there is no special markup for AI surfaces, so structure and coverage do the work, and the Princeton GEO study found evidence-rich, well-structured content wins citations by up to 40 percent over keyword-optimised equivalents.
What should I check after implementing?
Re-run the crawl and confirm depth fell and orphan count hit zero. Then check the pages you promoted are actually citable: run the AEO Ready Checker and a GEO Audit, watch positions with the Google Ranking Tracker, and track assistant presence with the AI Visibility Score. If pages are now reachable but still not indexed, how to get pages indexed faster covers the next gate.
Frequently Asked Questions
What does the Internal Linking Graph Optimizer do?
The Internal Linking Graph Optimizer is a free GrowthGPT tool that takes your page inventory and a pillar URL and returns an entity-based linking plan: a silo map grouping pages by topic and role, ready-to-paste link recommendations with anchor text and surrounding context, a click depth analysis showing current versus optimized depth, and a list of orphan pages with the specific sources that should link to them.
What do I paste into the site inventory field?
One page per line as a URL followed by its topic, for example https://example.com/blog/saas-pricing - SaaS pricing strategy. Twenty to eighty pages is the useful range for a single run. Include the commercial pages, not just blog posts, because the point of the exercise is routing authority toward pages that convert.
Which navigation pattern should I choose?
Hub and Spoke suits one dominant topic with supporting articles. Flat Silo suits several equal categories that should stay separate. Topic Cluster suits a content library where a pillar page anchors each cluster. Hybrid suits a site with a commercial core plus a broad blog. Pick the shape you already have unless you are deliberately restructuring.
What is an orphan page and why does it matter?
An orphan page has no internal links pointing to it, so crawlers can only find it through a sitemap and it inherits almost no authority from the rest of the site. Orphan pages are usually old posts, landing pages built outside the CMS, and anything published during a redesign. They are the cheapest fix in SEO because the content already exists.
Does internal linking affect AI search visibility?
Yes, in two ways. Links decide what gets crawled and how deep, which sets whether a page is available to be retrieved at all. And descriptive anchor text plus tight topical grouping strengthens the entity associations that retrieval systems use to judge whether a site is an authority on a subject rather than an occasional visitor to it.
How many internal links should a page have?
There is no magic number, but a useful working range is three to eight contextual links in the body of a typical article, all placed where they genuinely help the reader. Value comes from relevance and position, not volume: a link inside a relevant paragraph carries more weight than the same link in a footer block repeated on every page.