Inbound Internal Links and AI Citation Probability
Abstract
We audited all 91 posts in our own blog corpus across four AI engines and sorted them by how many inbound internal links each page received. The pattern was steep: pages with zero inbound internal links were self-cited 4.5% of the time, while pages with ten or more were self-cited about 44% — roughly a tenfold swing on the same content, from the cheapest lever a publisher fully controls. Reach varied by engine, with Perplexity citing our pages more widely than any other at about 59%. The finding is correlational on our corpus at one point in time, but it is the strongest on-page effect we have measured, and it is wired at publish rather than retrofitted.
Method
We took all 91 posts in our own blog corpus and ran a four-engine citation audit on 2026-07-11, recording for each page whether an engine cited it. We then sorted the pages by inbound internal link count — links pointing at the page from our own other pages — and compared the self-citation rate of the pages with none against the pages with ten or more.
Results
| Inbound internal links | Self-cited rate |
|---|---|
| 0 | 4.5% |
| 10 or more | ~44% |
Same content, different link graph, roughly a tenfold swing. Reach also varied by engine: Perplexity had the widest reach in our 91-post audit — it cited our pages more often than any other engine, at about 59%.
Practical implications
Inbound internal links are the cheapest lever a publisher fully owns, and on our corpus they were the strongest. The move is to wire the link graph at publish — point ten or more of your own pages at anything you want cited — rather than retrofitting links after the fact. It costs nothing but editorial discipline.
Limitations
This is a correlation on our own corpus at a single point in time, not a controlled causal trial — high-link pages may differ from low-link pages in other ways. We report the two endpoints we measured (zero and ten-plus) and deliberately do not interpolate a curve between them. Treat it as a strong prior for your own site, not a guarantee.
Source: The AI Citation Institute per-post 4-engine audit of 91 blog posts, 2026-07-11 (own-cited rate by inbound-internal-link count).
Cite this record
Open accessAI Citation Institute. "Inbound Internal Links and AI Citation Probability." Research record 2026-002, 2026-07-23. https://aicitationinstitute.org/research/internal-links-and-ai-citation (CC BY 4.0).
Released under CC BY 4.0 — quote it, chart it, cite it. All we ask is attribution back to this record.
Related: the voice experiment, how many inbound links you need, the live scoreboard, and our methodology.