Information gain is now a ranking signal you can measure
Google has held a patent on “information gain scores” since 2020: a way to rank a page not by how well it covers a topic, but by how much it adds beyond the documents the user has already seen. For years that patent was a curiosity. The March 2026 core update made it the story. Sites whose model was to read the current top 10 and publish a cleaner version of the same article lost visibility across whole templates, while pages carrying original numbers, test results, and defensible opinions moved up without a single new backlink.
The frustrating part is that “add something new” has always been the advice, and it has always been unfalsifiable. New compared to what? The Information Gain Analyzer makes it concrete. It reads the actual top 10 for your keyword, breaks each page into information units (the distinct points it makes), and clusters the units that say the same thing. What comes out is a map of the SERP's knowledge: which points are consensus, which appear on only one page, and how much of the whole results page is repetition.
The SERP echo chamber
Most keywords are echo chambers, and the mechanism is obvious once you see it. Everyone researches an article the same way: open the ranking pages, note what they cover, write a “more complete” version of the same outline. Each new page is optimized against the existing ones, so the consensus hardens. Run the analyzer on a commercial keyword and you will routinely see Echo Scores in the 70s and 80s, meaning the large majority of points on page one are repeated on other pages of page one. Ten URLs, one article.
A high Echo Score cuts two ways. It explains why yet another skyscraper post is not working: you are adding an eleventh copy to a pile Google already considers redundant. But it is also an opening. When the incumbents all say the same thing, even two or three genuinely new points give the ranking systems a reason to surface your page. The hardest SERPs to enter are the low-echo ones, where each result already earns its slot with something distinct.
How to add genuine information gain
Not all “new” content counts. Rephrasing a consensus point, adding a longer intro, or bolting on an extra H2 that every competitor also has does not move the score, yours or Google's. Three kinds of material reliably do:
- Original data. A survey you ran, pricing you collected, a benchmark you measured. Even a small dataset produces claims no other page can make, and other sites citing your numbers compounds the advantage.
- First-hand testing. Screenshots of the actual product, the failure you hit at step three, the setting the docs never mention. Testing produces the specific, verifiable detail that summaries structurally cannot.
- Expert point of view. A practitioner willing to say which popular advice is wrong and why. Judgment is the cheapest form of information gain to produce and the rarest one on most SERPs, because aggregators cannot fake having an opinion.
One caution: unique is not the same as complete. The consensus points on a SERP are consensus because searchers need them. Cover the table stakes, then spend your word count where the top 10 is silent. The analyzer shows both lists, so you are never guessing which is which.
Where the analyzer fits in your workflow
Before writing, run the keyword alone. The cluster map is your brief: the consensus points are mandatory coverage, the Echo Score tells you how much originality the SERP will reward, and the single-page points show angles competitors found but did not own. After drafting, or for any page that slipped in rankings since March 2026, add your URL. The unique-contribution score tells you what share of your points the SERP cannot echo, the unique-points list shows exactly which ones they are, and the missing-consensus list shows the gaps that make a page look thin to both readers and rankers.
Pair it with the free SEO checker to cover the on-page basics, and if you would rather not do the differentiation work by hand, see how ProofWrite researches and writes articles that are built on sourced facts instead of a paraphrase of page one.