Internal linking has been an SEO fundamental for two decades. Everyone knows it matters. Almost no one does it well. What we didn't expect: AI Overviews and ChatGPT-Search use internal link structure as a signal too — and they reward different patterns than classic Google did. A short field guide to what changed.
What classic SEO rewarded
- Anchor text relevance. Links from related pages with descriptive anchor text passed topical signal.
- Link depth. Pages within 3 clicks of the homepage got crawled more often and ranked better.
- Link volume. More internal links to a page = more PageRank (broadly).
These still all matter. They're not enough anymore.
What AI search additionally rewards
1. Hub-and-spoke topic structure
AI engines use internal linking to model topical authority. A site with one page about a topic plus eight pages linking to it is treated as more authoritative on that topic than a site with nine equally-weighted pages on related topics.
Practically: pick the page you most want to rank on a topic (the hub). Build 6-12 spoke pages that each address a sub-question. Each spoke links back to the hub with descriptive anchor text. The hub links to each spoke from a contextual section. The result is a recognizable cluster that AI engines weight as a topical authority signal.
2. Bidirectional links between adjacent topics
Classic SEO rewarded one-way link flow: deep pages link up to hubs. AI search additionally rewards lateral links between adjacent topics. If you have a page about "X for small business" and a page about "X for enterprise", linking them to each other (both directions) tells the AI engine these are facets of the same parent topic.
We've seen this single change move adjacent pages 5-15 positions on long-tail queries that span both contexts.
3. Anchor text that names the entity, not the page
For AI citation, "learn more about our SEO audit services" works better than "learn more." Obvious for classic SEO. For AI: the anchor text becomes part of the model's understanding of what the linked page IS. Vague anchors give the model nothing to bind to.
Rule of thumb: anchor text should be the query you want the linked page to rank for.
4. Orphan pages get filtered out of AI re-ranking
An orphan page (no internal links pointing at it) was always weakly indexed by Google. In the AI search era, it's functionally invisible. AI engines retrieving for a query re-rank based on signals that include "is this URL well-integrated into its host site?". Orphan pages fail that check.
Audit step: pull every URL from your sitemap. Compare against every URL referenced as an hrefin your site's content. Any sitemap URL not referenced elsewhere is an orphan. Either link to it or remove it from the sitemap.
5. Footer links pass less weight; body links pass more
Classic SEO weighted body links higher than footer links by ~3x. AI re-ranking weights them higher by ~10x in our measurements. Footer links are now mostly navigation; they don't carry topical signal.
Implication: don't rely on footer link blocks to do your internal linking work. They don't.
A practical audit you can run in 30 minutes
- Pick 5 high-priority pages. The ones you want to rank for important commercial or AI-search queries.
- Search your site for links to each. A simple Google query site:yourdomain.com inurl:/target-page or a crawl with Screaming Frog. Note the count.
- Note the anchor text used.Is it descriptive, or is it "click here"-style?
- Identify gaps.For each priority page, find 3-5 related existing pages that should be linking to it and aren't.
- Add the links.Contextual placement in body copy with descriptive anchor text. Don't batch them in a "related links" section.
- Re-audit in 30 days. Search Console will show impressions and rankings for the priority pages. Track the delta.
Most sites have 70-80% of the SEO lift they could capture sitting in their existing internal linking. Building new content is slower and more expensive than connecting what you already have.