- Information gain is the new knowledge your content adds beyond what Google already indexes for a query. It is the single thing that separates content that ranks from content AI can write in seconds.
- Google measures it. The Information Gain patent (US 2020/0349181 A1) scores whether a page adds anything new, and the 2024 Search API leak exposed an originalContentScore field.
- You cannot fake it. The fix is adding what only you have: first-hand experience, original data, a defensible contrarian view. The free Information Gain Finder skill shows you exactly where.
Anyone can publish the consensus now. Type a query into any AI and you get a competent, average, completely unoriginal page in seconds, and so can every competitor. The only content that wins is content that adds something the consensus does not have. That something has a name: information gain. Here is what it is, how Google measures it, and how to add it to a page so it earns its rankings.
What Information Gain Actually Is
Information gain is a term borrowed from information theory, where it measures how much uncertainty a piece of data removes. In a decision tree, the algorithm picks the split that delivers the most information gain, the one that tells you the most you did not already know. Google took the same idea and pointed it at the web.
The plain-English version: for a given query, Google already has millions of pages. Information gain is how much your page adds on top of all of them. Restate what the top results say and your gain is near zero. Add a number, an experience, or an angle nobody else has and your gain is high. That gap is what Google rewards.
This is not a soft, fluffy idea. It is a measurable score, and it is the mechanism behind why thin, rephrased content stopped working after the Helpful Content era. If you are new to how Google thinks about content quality, our 3 Ranking Laws Cheat Sheet covers it as Law 2: how Google measures your content.
How Google Measures It
Two sources confirm Google measures information gain directly, not as a vibe.
- The Information Gain patent. Google's 2020 application "Contextual estimation of link information gain" (US 2020/0349181 A1, granted as US 11,354,342 B2) describes scoring a page on whether it adds new information to what is already indexed for a query, and applies across text, images, and video. Read the patent on Google Patents if you want the primary source.
- The 2024 Search API leak. It exposed an originalContentScore field, an integer from 0 to 127. Most content sits near the floor. Original research, proprietary data, and real case studies push it up. It is the clearest signal yet that originality is a number inside Google, not a guideline.
Put together, the message is blunt: Google can tell the difference between a page that says something new and a page that rephrases the top five results, and it ranks them accordingly.
Why Most Content Has Almost None
Here is the trap almost every content strategy falls into. Find a keyword. Google it. Read the top five results. Write something that covers the same ground in the same order with slightly different words. The output reads fine, ticks the brief, and adds nothing. Its information gain is roughly zero, and now that AI can produce that page instantly, a zero-gain page is worth nothing.
The brands that lost traffic in the Helpful Content updates were almost all in this bucket: high volume, competent, completely derivative. The brands that gained were the ones publishing things only they could publish. Information gain is the line between the two.
The Five Sources of Real Gain
You cannot manufacture information gain by adding words. It comes from having something the consensus does not. There are five reliable sources.
- First-hand experience. "When we ran this for a client, here is what actually happened." AI has never done the thing. You have.
- Original data. Your own numbers, a small survey, anonymised client results, an internal benchmark. One real statistic beats a thousand words of opinion.
- A defensible contrarian view. Where the standard advice is wrong, said plainly, with the evidence to back it. Agreement adds nothing; a well-argued disagreement adds a lot.
- Process detail. The exact steps, settings, and screenshots the consensus skips. Specificity is information.
- A fresh synthesis. Connecting two ideas nobody has connected for this query. The insight is the gain.
Notice what is not on the list: more words, more headings, more keywords. Length is not gain. A 600-word page of first-hand experience beats a 3,000-word page of consensus every time.
How to Find Your Gaps in Minutes
Knowing the theory is one thing. Spotting exactly where your own page is restating the consensus is harder, because you wrote it and it all feels original to you. That is what the free Information Gain Finder Claude skill is for.
Paste a page or a draft and your target query, and it scores how much of the page is consensus versus original, flags the lines that add nothing, protects the parts that are genuinely yours, and gives you the specific additions to make: the experience to add, the number to insert, the contrarian angle to take. It is the difference between "make this more original" and a shippable list of exact changes. Install it free, drop it into Claude, and run it on your top pages before you touch them.
Commodity vs High-Gain, Side by Side
The same topic, written two ways. The difference is information gain.
| Commodity (near-zero gain) | High-gain |
|---|---|
| "Title tags are important for SEO and should include your keyword." | "We rewrote 40 title tags to lead with the exact query. CTR rose from 3.1 to 4.8 percent in 14 days. Here are the three patterns that worked." |
| "Backlinks help your rankings, so build high-quality links." | "One placement in a top-tier publication moved our client from position 66 to number one. Here is the pitch that landed it." |
| "Make sure your content is helpful and answers user questions." | "We surveyed 312 buyers. 68 percent abandoned at the same unanswered objection. We added one paragraph and conversions rose." |
The left column could be written by anyone, or anything, in seconds. The right column could only be written by the person who did the work. That is the whole game.
FAQ
Is information gain a real Google ranking factor?
Google does not publish a single "information gain score" you can check. But it holds a patent (US 2020/0349181 A1) describing exactly this, and the 2024 API leak exposed an originalContentScore field. The mechanism is real, even if the number is not exposed to you.
Does more content mean more information gain?
No. Gain is about new knowledge, not word count. A short page of first-hand experience has more gain than a long page of rephrased consensus. Length can even hurt if it buries the original parts.
Can AI write high-information-gain content?
AI is excellent at the consensus and terrible at the gain, because it has never run your campaign, surveyed your customers, or held your contrarian opinion. Use AI to draft the scaffolding, then add the experience and data only you have. The Information Gain Finder shows you where.
How do I measure my information gain?
Compare your page against what already ranks for the query and ask, honestly, what does mine contain that those do not. If the answer is "nothing", the gain is zero. The skill does this assessment for you and quantifies it.
Where should the original content go on the page?
Up top. Roughly 44 percent of AI citations come from the first third of a page, so front-load your unique value. Do not save the good stuff for a grand finale nobody reads.