I Audited Internal Linking on 300 Sites. Three Patterns Move Rankings. The Rest Waste Crawl Budget.

TL;DR

• I reviewed internal link graphs across 300 sites between mid 2024 and early 2026. Most look fine on the surface and waste crawl budget underneath.
• Four patterns show up over and over: hub-and-spoke, hierarchical, mesh, and contextual. Only three of them reliably move outcomes.
• Hub-and-spoke wins for ranking specific commercial pages. Contextual wins for AI citations. Hierarchical wins for sites with more than 5,000 URLs. Mesh wins almost nothing despite being the prettiest on a graph.
• The single biggest lever nobody talks about is internal anchor diversity. Sites with 70%+ exact-match anchors to their money pages rank worse than sites that mix descriptive, partial, and branded anchors.
• John Mueller has been pretty clear: anchor text on internal links helps Google understand context, and the location of the link on the page does not change that much. So stop obsessing over footer versus body. Obsess over what you say in the link.
• If you only do one thing this week: audit how many unique anchor variations point at your top five revenue pages. Most sites have one. They should have four to eight.

Why I Wrote This (And Who It's For)

I run a link-building agency. Most of my time is spent on external links. But over the last 18 months, almost every site I audited had the same hidden problem inside its own walls: an internal link structure that looked organised in Screaming Frog and did basically nothing for rankings or AI visibility.

The pattern got loud enough that I started tracking it. 300 audits later, I have data. This post is what I'd want to read if I were running SEO in-house and somebody told me to "fix internal linking" with no further instruction.

This is for SEOs, content leads, and founders who already know what an internal link is and want to stop guessing which pattern actually works in 2026. If you want a definition-led intro to internal linking, the existing top results have you covered. I'm going to skip the basics and get into what I actually see in audits.

How I Built The Sample

Quick on methodology so you can decide how much to trust this.

Between mid 2024 and February 2026, I or my team ran technical audits on 300 sites. The split:

• 180 commercial sites (services, ecommerce, SaaS)
• 70 publisher and content sites
• 50 local service businesses

For each site I pulled the internal link graph from a fresh crawl, layered Google Search Console data on top (clicks and impressions by URL), and where the client gave me GA4 access, conversions per URL. For the last 90 audits I also pulled AI citation data from ChatGPT, Perplexity, and Google's AI Overviews using brand-name and category-level queries.

The takeaway from all of that data is short, so I'm just going to give it to you.

The Four Patterns I Keep Seeing

Every site I audit falls roughly into one of four internal linking shapes. People rarely build them on purpose. They just emerge from how the site grew.

1. Hub-and-spoke

One central page ("the hub") links out to a cluster of supporting pages ("the spokes"). Each spoke links back to the hub. Spokes might link to each other sparingly.

This is what most people mean when they say "topic cluster." It's also the shape behind the knowledge graph and entity model I've written about for AI search.

Where I see it work: ranking a specific commercial or money page. The hub accumulates internal PageRank from many spokes and gets visible lift on its target keyword.

Where it fails: when the spokes have nothing useful to say. A cluster of thin pages pointing at one money page does not save the money page. Google notices.

2. Hierarchical

Classic tree. Homepage links to category pages. Categories link to subcategories. Subcategories link to product or article pages. Each level mostly links downward, with breadcrumb-style links back up.

This is the default shape for ecommerce and large publishers. Schema.org actually formalises this with the BreadcrumbList and breadcrumb properties on WebPage, which Google uses to render breadcrumb trails in search results.

Where I see it work: large sites. Past around 5,000 URLs, hierarchical structure is the only thing that keeps crawl behaviour predictable. Googlebot needs the spine.

Where it fails: small sites that ape this pattern. If you have 80 pages and a six-level taxonomy, you've made things worse, not better. Click depth balloons. Pages get orphaned. I saw this on a 60-page accounting firm site that had its main service three clicks deep behind a category nobody ever clicked.

3. Mesh

Every page links to many other pages. The graph looks like a tangled ball of yarn. "Maximum internal PageRank flow," the theory goes.

This is what you get when you install one of the AI internal linking plugins and let it auto-link aggressively. It also shows up on sites that use "related posts" widgets stacked on top of "you may also like" widgets stacked on top of a tag cloud.

Where I see it work: almost nowhere. Across the 300 audits, mesh structures consistently underperformed hub-and-spoke for ranking and contextual for AI citations.

Where it fails: everywhere. Crawl budget gets shredded. Anchor text becomes meaningless because every page has 40 links with similar anchors. Pages that should accumulate equity get diluted into the noise.

The mesh problem comes back to something the original PageRank paper from Brin, Page, Motwani, and Winograd in 1998 made obvious 28 years ago. PageRank transferred from a page is divided equally among its outbound links. The more links you add, the less each one passes. Wikipedia's PageRank explainer walks through the same math. Stack 80 links on a page and each one is worth a 80th of what one link on a focused page would be.

4. Contextual

Links are placed inside body content based on what's actually being discussed, not where the page sits in a taxonomy. Anchor text reflects what the linked page is about. Links are spread across paragraphs, not clustered in a sidebar.

This is the pattern Google's own internal linking guidance keeps pointing at when it talks about "descriptive" and "relevant" anchor text in surrounding context.

Where I see it work: AI citations. This is the pattern that consistently shows up when a site is being cited inside ChatGPT and Perplexity answers. I dug into the why in my piece on getting cited in the first 30% of content. Short version: LLMs read sentences and grab links the way a reader would, so contextual placement is way more visible to them than navigation or sidebar links.

Where it fails: when there is no editorial discipline. Contextual works because a human or a careful automation decided this link belongs in this sentence. Spray-and-pray contextual is just mesh with a different name.

Which Pattern Wins What

Here's the bit nobody seems to print clearly. I'll just say it.

Serious answer: a real site usually layers two or three of these patterns. Hierarchical for the structural spine. Hub-and-spoke for each major service or topic cluster. Contextual links woven through the content of every page.

Mesh is the one I keep telling clients to stop doing.

The Lever Nobody Audits: Internal Anchor Diversity

This is the part nobody talks about. I'll keep it tight.

When I pulled the anchor text distribution for every internal link pointing at the top five revenue pages on each of those 300 sites, the pattern was loud. Sites where 70%+ of internal links to a money page used the same exact-match anchor ranked worse, on average, than sites where the same anchor distribution was 20-40% exact-match with the rest split between descriptive variants, partial-match, and branded.

The top performers usually had between 4 and 8 distinct anchor variations pointing at each important page. The under-performers had one or two, repeated everywhere.

Google's SEO link best practices doc explicitly says "good anchor text is descriptive, reasonably concise, and relevant to the page that it's on and to the page it links to." It does not say "identical every time." John Mueller has reiterated multiple times that the anchor text on internal links helps Google understand context. If you keep using the same anchor, you're giving Google one signal repeated 80 times. If you vary the anchor, you're giving Google 80 facets of context.

Mueller also told Search Engine Roundtable that the location of internal links on the page doesn't matter much for understanding what a page is about. Body, sidebar, footer, the anchor is doing the work either way. So stop debating placement. Audit your anchor variety instead.

What this looks like in practice

Say you have a service page at /services/saas-link-building. Internal links pointing at it might use:

• "SaaS link building" (exact match)
• "link building for SaaS companies" (partial)
• "our SaaS link building service" (branded partial)
• "how we build links for B2B SaaS" (descriptive)
• "my approach to SaaS link campaigns" (first-person descriptive)
• The brand name itself (branded)
• A URL or button-style anchor (occasionally, where natural)

The goal isn't perfect ratios. The goal is variety that reads naturally. If your CMS or AI plugin keeps forcing the same exact-match anchor everywhere, that's a problem to fix.

What 300 Audits Taught Me About Crawl Budget

Most small sites do not have a crawl budget problem. If you have 200 pages and Googlebot is crawling 600 URLs a day, you're fine. The problem at this size is usually that your important pages don't get enough internal links, not too many.

The interesting cases were the mid-to-large sites. Anything past 2,000 URLs starts to show crawl efficiency issues, and internal linking is one of the biggest causes.

The two crawl budget killers I saw most often:

1. Faceted navigation with internal links to every combination. Filter URLs (color + size + price + sort) end up linked from product listings. Googlebot follows them. Equity bleeds into infinite filter combos. I've seen sites with 10,000 "real" pages and 480,000 crawlable URLs because of this. Fix it with parameter handling, noindex, and selective rel="nofollow". There's more in my technical SEO fundamentals piece.

2. "Related posts" widgets that link to dozens of pages with no editorial logic. A blog post on "best CRMs" pointing to 24 random posts in a sidebar isn't passing meaningful context to any of them.

The shift toward hierarchical-plus-contextual instead of mesh was the single biggest crawl efficiency improvement I saw across the audits.

How AI Search Changes The Calculus

AI search reads internal links differently than Googlebot does. This isn't theory. I've watched it happen across the last 90 audits where I tracked AI citations explicitly.

LLMs that power AI search platforms like ChatGPT, Perplexity, and Gemini consume content in chunks. A model isn't traversing your link graph the way Googlebot does. It's reading a page, seeing in-context links, and using those signals to either follow up, cite, or build a richer answer.

Three things changed how I think about internal linking once AI citations entered the picture:

1. Position of the link inside the content matters for AI, not Google

Mueller said body-versus-footer doesn't matter much to Google. For AI citation, position matters a lot. Links inside the first 30% of body content get pulled into model context more often than links in footers or sidebars. I wrote up the detail in the first 30% of content study.

2. Anchor text and surrounding sentence become a unit

LLMs don't read anchor text in isolation. They read "as I explained in my piece on [anchor]," sentence and all. So your link's value to an AI is your anchor plus the 1-2 sentences around it. This makes lazy "click here" links almost invisible and makes well-written contextual links much more powerful than they were 18 months ago. The LLM optimisation post digs into the surrounding-context idea more.

3. Mesh structures confuse models more than they confuse Google

Google has had 25 years to learn how to deal with messy link graphs. LLMs haven't. When a model is summarising a topic, a clean hub-and-spoke or contextual structure gives it cleaner signals to pull from. Mesh structures, where everything links to everything, give a model 30 weak signals instead of three strong ones.

This is why the contextual pattern wins for AI citations even though hub-and-spoke wins for traditional rankings. They're optimising for two different consumers.

Three Real Examples From Client Audits

Numbers and patterns are easier to trust with examples. Three short ones.

Pollack Peacebuilding: hub-and-spoke at scale

I worked on internal restructuring for Pollack Peacebuilding alongside their broader content programme. The site had hundreds of articles but linked through a flat structure. We built clear topical hubs (workplace conflict, mediation, harassment training, etc.) and rewired the article cluster underneath each hub to link both up to the hub and laterally to one or two sibling pieces. The hub pages started ranking for harder commercial terms within months. Article-level rankings also improved because each piece now sat in a clear topical neighbourhood instead of floating alone.

Hotrod Hardware: hierarchical for ecommerce

Hotrod Hardware is an automotive ecommerce site with thousands of SKUs. Hierarchical structure isn't optional at that size. The problem we found wasn't the hierarchy, it was anchor text. Category links from the navigation used the same exact-match anchor on every page, leaving Google to make sense of context purely from page content. Diversifying anchors across body content (especially in product descriptions and buying guides) lifted category rankings without changing the underlying tree.

CBD.co: contextual + hub

CBD.co is an ecommerce marketplace, so it needs hierarchical navigation, but the editorial content (buying guides, ingredient explainers, education) is where AI citation visibility comes from. Restructuring guides to link contextually into product and category pages, with varied descriptive anchors, gave the site both ranking lift and noticeable presence inside category-level AI Overview results.

None of these moves required new content. They required moving existing internal links into better positions with better anchors.

A Practical Audit You Can Run In An Afternoon

If you want to actually do something with all of this, here's the order I run the audit in. It maps to what I'd ask for on a free audit for a new client and what we work through on our SEO services engagements.

1. Pull a fresh crawl. Use Screaming Frog, Sitebulb, or your tool of choice. Export internal links by source and destination URL with anchor text.
2. Pick your top 5 revenue or strategic pages. This is usually service pages, key product or category pages, and 1-2 cornerstone content pieces.
3. Count internal links pointing at each. If a key page has fewer than 10 internal links from non-navigation pages, that's your first fix.
4. Count unique anchor variations. If a page has 80 internal links with only 1-2 distinct anchor texts, diversify. Aim for 4-8 variations that all read naturally.
5. Check click depth. Anything important should be reachable within 3 clicks from the homepage. Deeper than that, restructure.
6. Find orphans. Pages with zero internal links from anywhere. Either link to them or remove them.
7. Audit your hubs. For each topic cluster, pick a hub page. Confirm every cluster member links to the hub with varied anchors and the hub links back out to most cluster members.
8. Look at where links sit on the page. For AI citation goals, get key links into the first 30% of body content. For Google, less critical, but still better than footers and sidebars.
9. Kill mesh patterns. Audit auto-related-posts and tag-page link sprays. Replace mass-link widgets with curated 3-5 link blocks chosen for relevance.
10. Rerun and measure. Track impressions and clicks for your top 5 pages in GSC for 60-90 days post-changes. Track AI citation visibility separately with Perplexity, ChatGPT, and Gemini brand and category queries.

If you're doing this on a site with more than a few thousand URLs, this gets painful by hand. Most of the audits I run use a custom crawler plus the Common Crawl public dataset for external context, but a small site can be done in Screaming Frog and a spreadsheet in an afternoon.

What I Don't Do (Because I've Seen It Backfire)

A few things I avoid based on watching them fail.

• I don't install "AI internal linking" plugins on small sites. On a 50-page site you don't need a plugin. You need an hour with a spreadsheet. Most of these tools add mesh patterns by default and you spend more time pruning than you would have spent linking by hand.
• I don't force every page into a cluster. Some pages are just useful standalone resources. Forcing them into a hub structure with awkward links makes the cluster weaker.
• I don't optimise internal anchors purely for the target keyword. This is the single most common mistake. If every link to your service page says the same five-word anchor, you're hurting yourself. You'd never accept that for external backlinks. Stop accepting it for internal ones.
• I don't ignore the navigation. Site-wide navigation links pass equity too. If your main nav doesn't include a link to one of your top 5 revenue pages, that's a bigger problem than any clever contextual scheme.
• I don't do programmatic internal linking for cluster pages. I have written about why programmatic SEO can backfire if the surrounding content is thin. Programmatic internal links sit in the same bucket. If you can't write a sensible anchor and surrounding sentence, the link probably shouldn't exist.

What Google Has Actually Said (And What People Misquote)

The internal linking conversation is full of confident claims that turn out to be loose paraphrases of John Mueller. So a few things straight from Google's own published guidance.

• The SEO link best practices doc says "Every page you care about should have a link from at least one other page on your site." Minimum bar, but a lot of sites fail it.
• The same doc says good anchor text is "descriptive, reasonably concise, and relevant." It does not say "exact match every time."
• It also says "linking to a related page on your site can help users navigate to that page," framing internal links as a UX signal first. That framing matters. Anchor text and surrounding context are read by both humans and crawlers.
• For the bigger picture, the BreadcrumbList property in Schema.org is recognised by Google for site hierarchy. Use breadcrumbs. They reinforce hierarchical structure.

Mueller's quote on link location, reported by Search Engine Roundtable, is the one I see misquoted most. He didn't say location is irrelevant. He said for crawling and contextual understanding, body versus sidebar versus footer doesn't change the picture much. That's a much narrower claim than "placement doesn't matter."

A Note On E-E-A-T And Internal Linking

One pattern I want to flag because it's emerged stronger in the last year. Internal links from author pages, about pages, and editorial process pages now seem to matter more than they used to, especially for YMYL content.

If your medical, legal, or financial pages aren't linked from an author bio with credentials, or from an "about" page that establishes who is publishing the content, you're missing an E-E-A-T signal that has become noticeably more important in 2026. This isn't a new pattern, it's the same hub-and-spoke logic, but the hub is the entity (author, organisation) and the spokes are the content they publish.

For commercial content, the equivalent is linking case studies, testimonials, and review pages into your service pages. The case studies index is a good example of an entity hub: every case study links up to it, and it links down to each individual study. Service pages also benefit when relevant case studies are linked from them with anchors that name the client and the outcome.

What To Do This Week

Numbered, because it's easier to act on.

1. List your top 5 revenue or strategic pages. Write them down.
2. For each one, count internal links and unique anchor variations. Spreadsheet, 30 minutes.
3. If any of them has fewer than 10 internal links from body content, add 3-5 new ones from relevant existing pages. Use varied anchors.
4. Find your worst mesh pattern (usually an auto-related-posts widget or a tag cloud) and replace it with a 3-5 link curated block.
5. Move at least one important internal link per cornerstone page into the first 30% of the body content. This is a small change with disproportionate impact for AI citations.
6. Audit anchor diversity for your top 5 again in 30 days and check GSC impressions and clicks for those pages.

If you do nothing else from this post, do step 2 and step 3. Most sites I audit can pick up double-digit percentage lift in 90 days just from fixing internal link count and anchor diversity on five pages. No new content, no new backlinks. Just better use of what you already have.

That's the part I find satisfying. Internal linking is the only major ranking lever you fully control. External backlinks need outreach. Algorithm changes need patience. Content needs writers. Internal linking needs an afternoon and a willingness to stop building mesh patterns nobody asked for.

Get the patterns right and the rest of the SEO programme gets easier. Get them wrong and every other investment underperforms. That has been the most consistent finding across all 300 audits, and it's the one I keep coming back to with every new client.

If you want a second pair of eyes on your internal link structure or you're not sure which pattern your site is closest to, my team runs free audits that cover this in detail, and we work it into every SEO and link-building engagement. You can also see how this played out across other engagements in the public case studies, where the structural changes are usually quieter than the link-building work but tend to do more of the heavy lifting in the first 90 days.