Entity salience is how central an entity is to a given page or passage, as opposed to being mentioned in passing. Natural language systems have estimated salience for years: the subject a page is really about scores high, an aside scores low. For AI visibility it is the quality dimension of a mention, and it decides how much a given piece of coverage actually builds.
In one sentence
Salience is the difference between a page about you and a page that mentions you, and systems can tell them apart.
How entity salience works
Systems infer salience from structural and linguistic signals: whether the entity appears in the title and headings, how early it appears, how often, whether it occupies subject position in sentences, and whether the page’s claims are about it or merely adjacent to it. Salience then conditions what a mention contributes: high-salience coverage strengthens entity association between the entity and the page’s topic, while a low-salience aside contributes little however authoritative the page.
Worked example
A brand appears in two listicles. In the first it has a dedicated section: named in a heading, discussed across three paragraphs, its trade-offs stated. In the second it appears once, inside a comparative clause about a rival. Both count as mentions; only the first is coverage in any meaningful sense, and only the first meaningfully files the brand under the listicle’s topic. Illustrative example of the mechanism.
Why entity salience matters
Salience is why mention counting misleads. A footprint of a hundred passing namings can build less association than ten pieces of genuinely about-you coverage, which changes what outreach and content placement should optimise for: dedicated sections, not name-drops. It also applies to owned pages, where the entity a page is structurally about is the association it builds; a services page that is really about twelve things is salient for none of them.
What affects entity salience
Evidenced factors: position, frequency and grammatical role of the entity within the text, the signals salience estimation is built on in natural language processing. Factors with practitioner evidence: presence in title and headings, dedicated sections rather than scattered references, and one clear subject per page. These are treated as working hypotheses until tested.
Related concepts
References
- Google Search Central, Optimising your website for generative AI features on Google Search: developers.google.com
Author: Harpal Singh · Last reviewed: 7 August 2026