visual/harpers/1 matches the site

public static VisualItem Harpers_Visual_Content()
{
    var harpersVis = new VisualItem()
        .WithContext()
        .WithId("https://linked.art/example/visual/harpers/1")
        .WithLabel("Visual Content of Harpers");

    harpersVis.ReferredToBy = [
        new LinguisticObject()
            .WithClassifiedAs(Aat.Description, Aat.Statement)
            .WithContent("The text and image are primarily red and black")
    ];

    harpersVis.Represents = [Aat.Term("300025943", "Woman")];

    return harpersVis;
}

• Metatype classification The second argument classifies the first, not the object: WithClassifiedAs(Aat.Description, Aat.Statement) says this is a Description, and Description is a kind of Statement. The Plain C# form shows the nesting.
• Terms outside the registry The vocabulary registry holds only the terms linked.art itself publishes, so a dataset's own concepts, places and people are written with Aat.Term, Tgn.Place and Ulan.Person, giving the id and the label.

var visualItem = new VisualItem
{
    Context = ["https://linked.art/ns/v1/linked-art.json"],
    Id = "https://linked.art/example/visual/harpers/1",
    Label = "Visual Content of Harpers",
    Represents = [Aat.Term("300025943", "Woman")],
    ReferredToBy =
    [
        new LinguisticObject
        {
            Content = "The text and image are primarily red and black",
            ClassifiedAs =
            [
                new Type("http://vocab.getty.edu/aat/300435416", "Description")
                {
                    ClassifiedAs = [Aat.Statement],
                },
            ],
        },
    ],
};

Generated from the object the example builds: every property that is set, as object initialisers, with the vocabulary constants where a term is in the registry. The fluent helpers are a convenience, not a requirement; this form and the one as written produce the same JSON.