Information gain is the amount of new, useful information a page adds beyond what a searcher has already read or what the other results for the same query already say. A page with high information gain teaches the reader something the rest of the search results do not; a page with low information gain rearranges what is already there.
How information gain works
The term was borrowed from statistics and machine learning, where information gain measures how much a piece of data reduces uncertainty. In search, it became widely discussed after Google was granted a patent in 2022 describing how a system could score documents by how much new information they offer a user, relative to documents that user has already seen. A patent shows what Google has explored, not what its ranking systems do today, so treat the patent as context rather than proof.
The idea is easier to accept from Google’s public guidance. Its advice on helpful, people-first content asks whether a page provides original information, reporting, research or analysis, and whether it offers substantial value compared with other pages in the results. Those questions describe information gain in plain words.
In practice, information gain comes from things only you can supply:
- first-hand experience: what actually happened when you did the work, including what went wrong;
- original data: your own survey, test results, prices you have observed or anonymised customer figures;
- worked examples using real situations rather than hypotheticals;
- a clear position where the consensus is wrong or incomplete, with the reasoning shown;
- local or sector detail the national articles skip.
Why it matters
When ten pages say the same thing, a search engine has little reason to prefer yours, and an AI system building an answer only needs one of them. At the time of writing (October 2026), AI Overviews and AI chat tools summarise the consensus well. My judgement, not a measured finding, is that pages which repeat that consensus are the easiest to replace, while pages adding something distinct are the ones still worth citing or clicking.
Take a UK accountancy practice writing about Making Tax Digital for Income Tax. Dozens of pages restate HMRC’s dates and thresholds. A page showing what the quarterly updates looked like for a real sole trader client in the first months, which software problems came up and how long each update took, has information nobody else can copy. It also demonstrates the first-hand experience that E-E-A-T describes.
Common mistakes
- Writing by summarising the top ten results. The output is, by construction, the page with the least new information.
- Adding length instead of substance: longer pages with more headings but no new facts.
- Calling a rephrased definition “original research”.
- Hiding the unique part. If your own data sits in paragraph twelve, many readers will never reach it.
- Producing large volumes of AI-drafted articles on the same topics as everyone else, which tends to drive information gain towards zero.
How to act on it
Before writing, run a SERP analysis: read the pages that already rank and list what they all say. That list is the baseline you must cover briefly and then go beyond. Next, list what you know that they do not: questions customers ask you, numbers from your own records, mistakes you see repeatedly, how the rules differ in Scotland or Northern Ireland, or what the job costs in your area.
Put the new material near the top, label it clearly (“from my 2026 customer survey”, “what I found in this audit”) and show the method behind any figures. A content gap analysis tells you which topics are missing from your site; information gain is about what is missing from each page.
If you want an outside view on whether your pages add anything, building briefs around original material is central to the content SEO and strategy work I do.