Information gain is the new value a page adds beyond what a reader has already seen on the pages that rank for the same query. It comes from a Google patent, "Contextual estimation of link information gain", granted in June 2022, but as of 14 October 2026 Google has not confirmed that an "information gain score" is a live ranking system. The practical takeaway does not depend on that: Google's own helpful-content questions ask whether a page offers original information and substantial value compared with other results, so a page that only restates the top ten is exposed whatever the mechanism.
Key Takeaways
- Information gain = what your page says that the other top results do not, measured by meaning rather than wording. Rewording a competitor adds nothing.
- The patent is real (US11354342B2, granted 7 June 2022), but Google has not said it is used in ranking. Treat it as a useful lens, not a confirmed signal.
- Google's helpful-content guidance asks for "original information, reporting, research, or analysis" and "substantial value when compared to other pages in search results". That is the confirmed, citable version of the idea.
- The cheapest sources of new information are things only you hold: support tickets, sales calls, product data, first-hand tests and customer outcomes.
- Add one genuinely new element per page and say so near the top. A page needs a reason to exist next to the ten that already rank.
- Do not pad with filler numbers. Unsourced or trivial statistics are not information gain.
What "information gain" actually means
The term comes from machine learning, where it describes how much a new piece of data reduces uncertainty. SEOs use it for a narrower idea: given what a person has already read, how much does this next document tell them that is new?
The patent behind it is titled "Contextual estimation of link information gain". Google Patents lists it as filed on 18 October 2018 by inventors Victor Carbune and Pedro Gonnet Anders, published as US20200349181A1 and granted on 7 June 2022 as US11354342B2. In the patent's description, a machine-learning model produces an "information gain score" for a document, so documents can be presented in a way that reflects how much a user would likely learn from viewing them.
Two details matter for how you write:
- Score depends on the reader's session. Per Ahrefs' explanation, the score is relative to what a specific user has already seen, so the same page can be worth more to one searcher than another.
- It is semantic. The comparison is between meanings, not strings. Ahrefs and Search Engine Land's 2023 explainer both note that paraphrasing an existing article does not make it new.
What is proven and what is not
This is where most information-gain posts overreach, so here is my sorting of the claims.
| Claim | Status as of 14 October 2026 | Source |
|---|---|---|
| A Google patent on information gain scores exists and was granted | Verified | Google Patents |
| Google uses an information gain score in ranking | Not confirmed by Google; ranking use is inference | Ahrefs notes Google has not confirmed it and that patents often do not ship |
| Google asks whether content offers original information and value beyond other results | Verified, in Google's own guidance | Creating helpful content |
| Pages with many unique data points score higher on a third-party content score | A vendor's study reported by Ahrefs (On-Page.ai, 150 pages); not a Google metric | Ahrefs |
| Information gain helps you get cited in AI answers | Plausible, not demonstrated | No primary source |
That last row deserves a note. Google's AI features documentation says there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary", and that these features can run several related searches across subtopics. A page that covers a sub-question nobody else answers has a plausible route into that retrieval, but I have not seen evidence that proves it. I use it as a hypothesis to test, not a promise.
Where new information comes from
I sort sources of new information by how hard they are for a competitor to copy. The harder to copy, the more durable the gain.
Tier 1: things only you have
- Support tickets, chat logs and sales call notes: the exact words customers use and the objections they raise.
- Product usage data, aggregated and anonymised, where you are allowed to publish it.
- Results of tests you ran: your own A/B tests, crawl experiments, price changes.
- Customer outcomes with permission, including the unflattering ones.
Tier 2: things anyone could gather but few bother to
- A small original survey, even 50 to 100 responses, with the method stated.
- A manual audit of 30 to 50 competitor pages, tabulated.
- Screenshots and step-by-step walkthroughs from actually doing the task. Google's helpful-content questions call out expertise that comes from having used a product or visited a place.
Tier 3: synthesis and stance
- A clear position, with reasoning, where the top results hedge.
- A decision framework or template the reader can apply.
- A comparison that ranks options against stated criteria, with the criteria shown.
Tier 3 is the weakest on its own, because opinion without evidence is easy to copy. Pair it with something from Tier 1 or 2.
The 5-step process I use
Step 1: Read the top results as a reader would
Open the top 5 to 8 results for the query in a fresh browser. For each, write down the three to five distinct claims or sections. You are building an inventory of what a reader would already know after reading them.
Step 2: Find the gap
Compare the inventories. Look for:
- Questions raised in the People Also Ask box, Reddit threads or your own support inbox that none of the pages answer.
- Steps that are skipped ("then configure the settings" with no settings shown).
- Claims made without a source or an example.
- Angles that no page takes: a specific audience, region, budget, or edge case.
Ahrefs suggests the same move: check Reddit and question reports for the questions the SERP misses.
Step 3: Choose one new element
Pick one thing per page and commit. Examples I have used or would use:
- A table built from your own audit of 40 pages, with the method in one paragraph.
- A worked example with real numbers, and the mistake you made on the way.
- A template file or checklist that readers can copy.
- A screenshot sequence of a process that competitors describe only in words.
Step 4: Put it where readers and machines will see it
Say what is new in the first screen: "I audited 40 course landing pages in September and found..." Passage-based retrieval and impatient humans both reward this. Give the new element its own heading so it can be quoted on its own. For how I structure briefs so every section stands alone, see content briefs for AI extraction.
Step 5: Attribute, date and maintain
Every number gets a source or a "from my own audit, September 2026" label. The 2023 Princeton paper on generative engine optimization ("GEO") reported visibility gains of up to 40 percent for some strategies in its benchmark, and vendors often cite adding statistics and sources as one of them. Treat that as a lab result on a research benchmark, not a guarantee for your site. Then add a review date and re-check in a quarter.
A worked example (illustrative)
Suppose you run a course site and want to rank for "data analytics course fees". The top pages list fees, durations and a brochure link. Your inventory shows nobody explains what the fee includes, what is paid separately, or how EMI changes the total.
Your information-gain element could be a table titled "What the listed fee covers, and what it does not", built by reading the terms of the course and by listing questions your admissions team actually gets. No invented data is needed: it is clarity from sources you already hold. The same logic applies to any program page; see how I structure a degree or bootcamp page.
Information gain vs related ideas
| Idea | What it asks | How it differs |
|---|---|---|
| Information gain | What does this page add beyond what the reader already saw? | Relative to other documents and the session |
| Originality | Is the text copied? | Plagiarism check, not value check |
| Depth | How much detail is there? | A long page can still add nothing new |
| E-E-A-T | Can this source be trusted on the topic? | About the source, not the page's new content |
| Freshness | Is the information current? | New is not the same as recent |
What not to do
- Do not pad with filler data points. Ahrefs warns against stuffing pages with numbers to chase a score. A statistic you cannot source is a liability.
- Do not spin. If the claims match the top results, the page adds nothing, however it is phrased.
- Do not claim you can measure it. No tool outputs Google's score. Tools that show a score are modelling something; label it as such in reports.
How to measure whether it worked
Because the score is not observable, measure outcomes:
- Record the page's impressions, clicks and average position in Search Console for 28 days before and after the update.
- Check whether the page is cited in AI answers for the target prompt, using a fixed prompt set. I describe the method in my 50-prompt tracking panel.
- Look at engagement qualitatively: do people link to or quote the new element?
- Allow time. Google's SEO starter guide says some changes take effect in hours and others take months, and suggests waiting a few weeks before judging.
Change one thing at a time where you can, and note the date.
A pre-publish information gain checklist
- [ ] I can state in one sentence what this page says that the top results do not.
- [ ] That statement appears in the first 150 words.
- [ ] The new element comes from first-hand work, an original dataset or a clear, reasoned stance.
- [ ] Every statistic has a named source or is labelled as my own data, with the date.
- [ ] The method is stated for anything I measured.
- [ ] The page still answers the query directly in its opening paragraph.
- [ ] A review date is in the calendar.
FAQ
Is information gain a confirmed Google ranking factor?
No. A Google patent describing an information gain score was granted in 2022, but Google has not confirmed that it is used in ranking. What Google has said publicly is that it wants original information, analysis and substantial value compared with other results.
Who invented information gain in SEO?
The concept comes from machine learning. The SEO use traces to a Google patent by Victor Carbune and Pedro Gonnet Anders, filed in 2018 and granted in 2022. SEO writers popularised it from around 2020 onward, and early explainers like Search Engine Land's 2023 piece treated it as a hypothesis.
How is information gain different from a content gap analysis?
A content gap analysis finds topics your competitors cover that you do not. Information gain looks the other way: what you can say that none of them do. Use gap analysis to reach parity, then information gain to differentiate.
Do I need original research to get information gain?
No. Original data is the strongest source, but first-hand experience, a clearer worked example, a skipped step or an unaddressed edge case can all add value. A small, honest survey with its method stated beats a large claim with no source.
Can AI-written content have information gain?
Only if the information in it is new. A language model summarising what already ranks adds nothing the reader has not seen. A model used to organise your own data, interviews or test results can help you present something new. Google's own AI content guidance focuses on accuracy and value, not the tool.
Does information gain matter for AI Overviews and AI Mode?
Possibly, but this is unproven. Google says there are no special requirements for these features, and that they may issue several related searches. Covering a sub-question that others skip gives you a chance to be retrieved; I would test it rather than assume it.
How long should a page be to add information gain?
Length is not the measure. A 700-word page with a new dataset can add more than a 3,000-word roundup. Cut anything that repeats what readers can find everywhere, and keep the new element prominent.
How do I show information gain to a client?
Show the inventory: what the top results say, the gap you found, and the element you added. Then report outcomes (impressions, clicks, AI citations on a fixed prompt set) rather than a made-up score.
Work with me
I'm Younus Fardeen, an organic growth, SEO and AEO strategist in India with 4+ years of marketing experience across edtech and startup clients. If you'd like help finding the gap on your key pages and building the new element that earns them a place beside what already ranks, see my work and send a note through the contact form at younusfardeen.in.