THE SHORT ANSWER
Internal linking for AI crawlers is about discovery and context rather than distributing authority. Many of these crawlers execute no JavaScript, follow fewer links per visit and revisit less often than Googlebot, so anything more than about three clicks from the homepage may simply never be fetched. Plain anchor tags in the HTML, descriptive anchor text and hub pages that consolidate a topic do more than any link equity strategy.
Internal linking advice has been dominated for twenty years by the question of how authority flows between pages. That framing was always a simplification and it is close to useless for AI crawlers, which are not running a link-based ranking calculation on your site in the first place. They are trying to find pages and understand what those pages are about.
That change of purpose changes the tactics. Descriptive anchors matter more, because the anchor is one of the few pieces of context a retrieval system has about a page before fetching it. Depth matters more, because the crawl budget is smaller. And script-driven navigation matters enormously, because a link that only exists after JavaScript runs does not exist for a large share of these agents.
The numbers, at a glance
Primary function: discovery and context, not authority distribution, since retrieval crawlers are not computing an internal ranking flow
Depth threshold: keep commercially important pages within three clicks of the homepage, since shallow crawls stop before deep pages
Script dependency: several AI crawlers do not execute JavaScript, so navigation rendered by a framework at runtime is invisible to them
Anchor text function: the anchor is often the only description of a page a system has before fetching it, so generic wording wastes the signal
Why these crawlers behave differently from Googlebot
Googlebot has two decades of infrastructure behind it: a rendering pipeline that executes scripts, deep crawl budgets and revisit scheduling tuned per site. Most AI crawlers are considerably simpler. They fetch HTML, parse it, follow some links, and move on, on a schedule that is often far less frequent.
The practical consequences follow directly. A page discoverable only through a filtered menu that builds itself in the browser is not discoverable. A page eight clicks deep in a paginated archive is unlikely to be reached. A page reachable only from a sitemap, with no HTML link pointing at it, gets fetched by some agents and not by others. None of these matter much for Google and all of them matter here.
The structural pattern that works
A hub page per topic. One page that introduces a subject and links, in body text, to every page in that cluster. This is the single highest-value structure for retrieval, because it also tells a system which pages belong together.
Bidirectional links. Cluster pages link back to the hub and sideways to two or three siblings. One-way structures leave orphans whenever the hub is missed.
Plain anchor tags in the source. Real href attributes present in the served HTML, not click handlers or router components that resolve at runtime.
Descriptive anchor text. The anchor should name the subject of the target page. Read more and click here transmit nothing.
Flat depth for commercial pages. Anything you actually want cited belongs within three clicks of the homepage, with the hub doing that job.
None of this is exotic. It is the site architecture people were told to build in 2012 and then abandoned in favour of framework-driven navigation that looks better and crawls worse.
Anchor text as an entity signal
Because retrieval systems work at passage level, they lose the page context that a human reader keeps in their head. The anchor text arriving from other pages is one of the few ways context gets restored: when eleven pages link to a page using wording that names the same subject, that repetition is a strong statement about what the target is.
This argues for consistent rather than varied anchors, which is the reverse of old external link building advice where over-optimisation was the risk. Internally, on your own site, describing the same page the same way each time is a feature. Reserve variation for cases where the target genuinely covers several subjects, and then vary along real distinctions rather than for the sake of it.
Two failure patterns worth checking for today
The first is the orphaned programmatic set. A team generates four hundred pages, submits them in a sitemap, and links to none of them from anywhere in the site. Google will crawl them from the sitemap eventually; several AI crawlers will not, because they follow links from pages they already fetched. Every generated page needs at least one HTML link from a page that is itself reachable.
The second is the infinite filter space. Faceted navigation that generates a URL for every combination produces millions of near-identical paths, and a small crawler that wanders into it burns its whole budget on nothing. Block filter parameters in robots, use plain links only for the facet combinations that deserve pages, and check the crawl logs for agents drowning in query strings.
An internal link review focused on retrieval
Fetch your homepage and two hub pages as raw HTML and confirm the navigation links are present in the source.
Find every page with no inbound internal link and give each one at least one link from a reachable page.
Build a hub page per topic cluster that links to every member in body text.
Rewrite generic anchors so each one names the subject of the page it points to.
Check crawl logs for agents looping through filter parameters and block those paths in robots.
Want leads like this in your pipeline?
Flock runs the campaigns, screens the enquiries and hands you only the ones that match your service area, job size and capacity. You pay per lead, not per month.
Book a 15-minute fit check | See lead package pricing
Related answers
Frequently asked questions
Does link equity still matter internally?
For Google, yes, though far less than the folklore suggests. For AI crawlers it is largely beside the point, because they are not computing an internal authority score. Optimise your internal links for discovery and context and the ranking benefit follows anyway; optimise purely for equity sculpting and you can easily make discovery worse.
How many internal links per page is right?
Enough to reach everything related, few enough that each one is a deliberate choice. A body-text link that a reader might actually follow is worth several in a footer block. There is no magic ceiling, but a page with eighty links in the template and three in the prose has its priorities inverted.
Do sitemaps replace internal links?
No. A sitemap assists discovery for crawlers that read it and says nothing about relationships between pages. Some AI crawlers do not fetch sitemaps at all. Treat the sitemap as a supplement to a properly linked structure, never as a substitute for one.
Should breadcrumbs be used?
Yes, and they are undervalued. Breadcrumbs create shallow paths back to hub pages from anywhere in a deep set, they render as ordinary links in the served HTML rather than through a framework, and with BreadcrumbList markup they state your topic hierarchy explicitly rather than leaving it to be inferred. For the effort involved they are among the highest-return structural additions available.
NEED A CLEARER PLAN?
Let’s turn your next move into momentum.
Talk to us →