Information gain in SEO is the useful new information a page adds beyond what a reader has already seen. Original data and first-hand explanations count. Different wording alone doesn't give readers anything they couldn't learn from the pages they've already read. Google's information-gain patent describes a contextual scoring method, not a public website score; its helpful-content guidance asks whether you contribute something existing pages don't.
I'm Brenden House. I run Searchmaxxed, an SEO and AI SEO agency.
In summary
- Information gain means new substance. Rewriting a competitor's article leaves the reader with the same information.
- Google's patent compares documents against information already presented to a user. It doesn't establish a universal originality score.
- Your product team, engineers and service records can supply useful evidence without a large research project.
- Keep the evidence attached. A finding needs its source, scope and publication permission before it becomes a public claim.
- Originality gives an answer engine a reason to use your page. Crawl access, relevance and credible sourcing still matter.
What does Google's information-gain patent actually say?
Google's patent describes estimating how much additional information a document offers after a user has received information from other documents. The comparison is contextual.
The patent is *Contextual estimation of link information gain*, US11354342B2. Google Patents records filing on 18 October 2018 and the grant on 7 June 2022.
Its abstract describes additional information beyond:
“information contained in documents that were previously viewed by the user”.
That last part matters. The patent doesn't describe checking whether your article contains a predetermined percentage of words missing from the internet. It describes what another document could add for this user, given information they've already received.
Imagine an enterprise buyer researching a software migration. They've read the standard advice about backing up data and mapping fields. Another page repeating those instructions adds little.
A documented example explaining how to preserve audit history during that migration could answer their next question. Same topic, different contribution.
Does the patent cover search results or only assistants?
It covers both kinds of implementation. Claim 1 describes an automated assistant responding to natural-language input. But don't stop reading there.
| Part of the granted patent | What it describes | What that means for your reading |
|---|---|---|
| Claim 1 | An automated assistant selecting information from new documents | Assistant answers are explicitly included |
| Claims 5 and 6 | Previously accessed documents, new documents and a ranked list | The granted claims extend beyond the assistant example |
| Claim 7 | A search-results interface after a user visits a document and returns | Search-interface behaviour is explicitly included too |
These are descriptions in the granted patent. They aren't evidence that a particular Google product runs this method today.
For your content brief, keep the useful question: what can this page explain after the reader already knows the basics?
Is information gain a confirmed Google ranking factor?
The patent doesn't establish a confirmed, universal ranking factor called information gain. Google's published content guidance does explicitly ask for originality and additional value.
Separate those claims.
In its helpful-content guidance, Google asks:
“Does the content provide original information, reporting, research, or analysis?”
It also asks whether a page avoids simply rewriting sources and provides substantial additional value. That's a clear editorial direction. You don't need to prove a patent's implementation before following it.
Useful original content still answers the search. A manufacturing buyer looking for material compatibility needs the compatibility answer before an unusual story about your production line.
Cover the essentials first. Then add the test conditions, failure example or engineering explanation that helps them choose correctly.
At Searchmaxxed, keyword research and competitor checks inform the page's subject, while the company's own facts supply what those other pages can't.
Can you calculate an information gain score?
You can't calculate Google's internal information gain score from its public documentation, and third-party scoring methods don't provide access to Google's system either.
Some tools compare semantic similarity using embeddings, which are numerical representations of meaning. That can flag pages discussing much the same thing. It doesn't verify a new claim.
A fabricated benchmark could look very different from existing articles. It would still be fabricated.
Review the draft against these questions before you commission more writing:
| Question | Evidence to put beside the draft |
|---|---|
| What does the reader learn here that competing pages omit? | The exact finding or explanation |
| Who knows this first-hand? | The named specialist or original record |
| Can another person check it? | A source, method, screenshot or reproducible example |
| Where does the claim stop being true? | Product version, population, conditions or exclusions |
| Why does this belong on this page? | The specific customer question it answers |
This is an editorial check. Use the answers to identify where a writer needs more evidence or a clearer explanation before the article goes live.
Passing a plagiarism check doesn't settle that question either: a page can avoid copied sentences without teaching its readers anything they didn't already know.
Where can B2B teams find useful original information?
Product decisions, implementation records and technical explanations from inside your business can answer questions that public summaries leave unresolved. Start with what you already know.
First-party data means information your organisation collects directly, but it isn't automatically publishable or representative of the wider market your article wants to describe.
Start with the question a customer asks before they can choose your product, then find the person or record that can answer it.
| Company or team | Source to inspect | Useful contribution to a search page | Boundary to preserve |
|---|---|---|---|
| Enterprise SaaS | Migration tickets and product documentation | Which records transfer, which need mapping and which cannot transfer | Product version and supported configurations |
| Venture-backed software company | Onboarding observations | Where implementation stalls and what changes the sequence | Observed accounts, not the whole industry |
| Manufacturer | Test reports and engineering notes | Why a component failed under a particular load or environment | Test method, material and operating conditions |
| Public-sector organisation | Approved service evaluations | Where an application process confuses users and what was changed | Privacy, accessibility and publication approval |
| Enterprise services team | Delivery records and specialist interviews | What must happen before a project can start | Contractual limits and client permission |
These are proposed examples, not Searchmaxxed client results.
Ask the engineer what changed their recommendation, or which condition makes the usual advice wrong for a customer trying to choose between those products.
A job title isn't evidence. The explanation is.
Our guide to original research for SEO covers larger research assets; here, we're choosing evidence that will answer the question on a particular page.
What would information gain look like on a SaaS migration page?
A useful migration page explains the product's actual limitations and the work required to handle them. General migration advice can't do that job.
Consider this hypothetical example. A SaaS company wants to rank for migrating legacy records into its platform, but its draft repeats the standard export-and-import instructions.
A buyer in a regulated organisation needs to know what happens to historical approvals before they can recommend the software internally. That's the missing answer.
| Draft element | Generic version | Evidence-led version |
|---|---|---|
| Main answer | Export your records and import them | Explain which historical fields the product accepts |
| Proof | An unsourced claim that migration is easy | An approved test using a documented configuration |
| Practical detail | Check the data afterwards | Show the fields to reconcile and the expected exceptions |
| Limitation | No exclusions stated | Identify unsupported approval records and the available alternative |
| Next step | Ask for a demo | Tell the buyer which system details to bring to the migration discussion |
No invented success rate is needed. The contribution is knowing what will happen to the buyer's records.
That gives the article a reason to exist beside the product page. Link the explanation to the relevant integration or service page so readers can check the supported scope.
Our content strategy work connects keyword research with product, industry and comparison pages that answer the questions holding up an enquiry.
How do you turn internal evidence into a publishable page?
Keep the proposed claim beside its source, limitations and publication permission, so a useful internal observation doesn't become an overstated public claim during editing.
Use this handoff sheet. The example entries below refer to the hypothetical SaaS migration, not a completed project.
| Field | Example entry for the hypothetical migration page |
|---|---|
| Customer question | Will our approval history survive the migration? |
| Evidence owner | Product engineer responsible for the import feature |
| Source | Current documentation and an approved test record |
| Proposed finding | State exactly which approval fields the tested configuration preserves |
| Limits | Name the tested version and excluded record types |
| Permission | Record what may be published and who approved it |
| Page placement | Put the answer beside the migration instructions |
| Update trigger | Review the claim when the import feature changes |
Keep confidential records private. Publish only the supported explanation and approved evidence needed to understand it.
For Searchmaxxed work, the business owns its website, content and first-party data, keeping those useful facts in your hands as the product changes.
Before approving a draft, ask someone outside the writing task to follow the source back. Can they reach the same conclusion?
If they can't, narrow the claim to what the evidence supports rather than asking the writer to make the original statement sound more confident.
Does original content help you appear in AI answers?
Original content gives an AI answer something distinct to cite, but originality alone doesn't establish that your page will be selected. It also needs to be accessible and relevant.
Google's AI search optimisation guide recommends:
“Create valuable, non-commodity content for your audience”
Its examples include first-hand perspectives rather than summaries of existing pages, which supports the case for original content without establishing how the patent is used.
Our own research shows why discovery and citation need separate measurement: an Australian agency-question panel we conducted on 2 October 2026 recorded both:
We put 300 Australian agency-buyer questions to ChatGPT three times each, with search on. It pulled 17,870 search results into those 900 answers and kept 4,389 as sources. Being found is the easy part. About three in four results it reads never make the answer.
That panel didn't test originality. It shows the difference between being retrieved and being cited, not a causal lift from adding new information.
Keep the finding visible in text. Give it a clear heading, name its source and explain its limits beside it, rather than hiding everything inside a graphic.
Google's AI-feature requirements also require a supporting page to be indexed and eligible for a search snippet. Originality doesn't replace technical SEO.
How should you check whether an improved article works?
Measure search visibility, citations and useful enquiries separately, because a better article can make progress on one without moving the others yet.
Save the original page. Record the content changes alongside titles, internal links and technical fixes, so you can see what else might have affected the result.
| Measure | What to examine | What it doesn't establish |
|---|---|---|
| Google visibility | Relevant queries, impressions, clicks and landing pages in Search Console | That originality caused the change |
| AI citations | The same customer questions, with cited URLs retained | That every future answer will cite you |
| Enquiries | Relevant form submissions and sales feedback | That every reader arrived through search |
| Content accuracy | Whether the published claim still matches the product | That a fresh date makes old evidence new |
Searchmaxxed combines Search Console and GA4 measurement with ongoing page and technical work, so findings can guide changes rather than merely decorate a report.
Check which searches are growing. If they don't relate to what you sell, revisit the page's purpose before treating those extra visitors as a result.
Is Searchmaxxed right for this work?
Searchmaxxed fits teams that want original evidence turned into SEO and AI SEO pages, with the website work handled too. Access to your specialists matters.
We do keyword research, answer-first articles, industry and comparison pages, schema and internal links, alongside the measurement and technical SEO those pages need.
The plans differ in what they cover:
- Growth System includes new pages or substantial rebuilds, alongside technical, schema and internal-link work.
- Agentic Search Engineering adds comparison, industry and integration page families, plus a quarterly proof asset built from the client's own wins.
- Custom & Enterprise can include original first-party research, multiple markets, dedicated engineering and custom reporting.
Those are service inclusions, not claims that every business needs a large study when a documented product explanation would answer the question better.
Use these checks to decide whether the service fits your team.
| Searchmaxxed is a good fit if | Look elsewhere if |
|---|---|
| You want SEO and AI SEO pages with implementation handled | You want strategy advice with no changes to your site |
| Your specialists can provide evidence approved for publication | You want rewritten competitor summaries without original evidence |
| You need outside support for content and website work | Your in-house writers and developers can handle both |
Frequently asked questions
What is the difference between information gain and unique content?
Unique content can use original wording while repeating familiar facts, whereas information gain adds useful substance beyond what the reader already knows.
Is information gain the same as topical authority?
No. Topical authority concerns a site's depth and credibility across a subject; information gain concerns what a particular piece adds to information already available.
Does every article need original research?
No. A tested example, specialist explanation or documented limitation can provide new information without a formal study. It still needs evidence.
Can AI-written content provide information gain?
It can present original evidence supplied to it, but that evidence needs verification regardless of which tool or person produced the draft. Our guide to LLM writing quality and SEO examines that separate writing question.
Should you remove information competitors already cover?
Keep the context readers need. Then add useful evidence where existing coverage stops, rather than making readers visit another page to understand the basics you left out.
How much information gain does a page need?
Google publishes no required percentage or target score in the documentation cited here, so judge the contribution by its usefulness and supporting evidence instead.
What should you do next?
Choose a page that already matters commercially. Find the question it leaves unanswered, then locate the product record or specialist who can supply a supported answer.
That's a useful brief. If you want Searchmaxxed to turn it into the page and search work around it, Let's talk pricing.