Information Gain.

Many pages on a topic essentially say the same thing. Read ten guides on 'having a website built' and you find the same steps ten times in a slightly different order. For a search engine, the question is therefore: which page still offers something new to a person who has already seen several of them? That is exactly the idea information gain describes. The term comes from information theory and appears in a Google patent describing how an information gain score for documents could be calculated.

It is important to put this in context: a patent does not prove that Google uses a method in exactly that way. The direction is, however, supported by Google's own guidance on helpful content, which explicitly asks whether content provides original information, reporting, research or analysis, and whether it provides substantial value compared with other pages in the search results. Information gain is therefore less a measurable metric than a standard for your own editorial work.

AI answers have given the principle more weight. Systems such as AI Overviews or Google AI Mode summarise widely known knowledge themselves. They need sources mainly for what is not everywhere: concrete figures, local specifics, first-hand experience, clear statements about prices and limits. A page that only repeats what is known becomes interchangeable for these systems. A page with its own contribution has a real chance of being named as a source, which shows up in the citation rate.

Information gain does not come from more words. A long text that writes around the topic offers less gain than a short paragraph containing information that is otherwise missing. Typical sources of genuine value are project experience, your own analyses, photos and examples from practice, a clear statement of when a solution does not fit, and regional details that only a local provider knows. This ties in closely with E-E-A-T, where experience is explicitly named as a quality characteristic.