An information gain content framework is a scoring system editors use to grade how much new information a draft adds before it ships. The IG-9 Rubric grades every draft on five dimensions: proprietary data, first-hand evidence, original framework, expert attribution, and freshness hook. The total is 9 points. A draft must score 7 or higher to publish. Anything lower is a ship-blocker. This is the rubric Metaflow runs on every post we publish, codified here for the first time.
Ahrefs analyzed 900,000 newly created web pages in April 2026 and found 74.2% contained AI-generated content (Ahrefs study of AI in new content). When three out of four pages competing for your query are AI-synthesized restates of the same SERP, adding new information becomes the only way to earn a citation in an AI Overview, a Perplexity answer, or a top-10 rank. A "this feels good" review does not survive that math. Editors need a numeric pre-publish gate.
TL;DR
- The IG-9 Rubric scores a draft on five dimensions: proprietary data, first-hand evidence, original framework, expert attribution, freshness hook.
- 7 out of 9 is the ship threshold. 0–3 is a ship-blocker. 4–6 needs work. 7–9 is ready.
- Google's granted patent US12013887B2 grounds the concept but does not give editors a rubric.
- Score the angle in the brief, not the draft. Late scoring wastes 80% of writing cost.
- A 30-minute IG-9 pass changes which pages you commission, not just how you edit them.
What information gain actually is (and what the SERP gets wrong)
Most posts on the SERP define information gain as "add original info to your content." That advice is correct and useless. It tells a writer to do the thing without telling an editor how to check it.
The Google patent in one paragraph
Google was granted patent US12013887B2, "Contextual estimation of link information gain," in June 2024. It describes an information gain score that ranks a second set of documents based on what the user has already seen. The Search Engine Journal teardown by Roger Montti makes the common misread visible: the patent does not say "rank the most comprehensive page higher." It says rank the next document by how much new information it adds beyond what the user already viewed (Search Engine Journal on the patent).
Why most SEO posts misread it
The popular reading treats information gain as a bigger-is-better signal. The patent treats it as an additive signal. Those are opposite instructions. Bigger-is-better produces 4,000-word guides that repeat the SERP. Additive produces 1,800-word posts that contribute one new dataset, framework, or first-hand teardown the SERP lacks.
The shift from comprehensive to additive
Animalz called this shift early: the goal moves from displacement to differentiation, because AI already synthesizes comprehensive coverage from the existing top 10 (Animalz on information gain). Once that synthesis is free, "comprehensive" becomes baseline, not differentiator. Your job is to be the source AI pulls from for the part nobody else covered. That is what the IG-9 Rubric scores.
Why an information gain content framework is now a publish gate, not a nice-to-have
In 2022, information gain was a theory. In 2026 it is a budget decision.
74 percent of new pages contain AI content
The Ahrefs 900K-page study is the number to anchor on: 74.2% of pages published in April 2026 contained AI-generated content. The same study reported that the top 10 Google results contain a similar share. The race to publish more pages is now a race to publish more pages that already exist in slightly different words. Without a scoring gate, your team is funding that race.
Differentiation is the only remaining moat
Kevin Indig's Growth Memo essay "Information Gainz" puts the strategic point plainly: prioritizing information gain means rethinking how we create content, not how we edit it (Kevin Indig, Growth Memo). Once you accept that, the rubric stops being optional. If you cannot show, dimension by dimension, what is new about a draft, the right answer is to not commission the draft.
Why editors need a score, not a vibe
Mike King's iPullRank treats this as a content engineering problem, not a content writing problem (iPullRank content engineering). Engineering disciplines run on rubrics. The IG-9 score is the engineering scorecard for an editorial pipeline. It lets two editors look at the same brief and agree on whether it is worth funding.
The IG-9 Rubric: a 9-point information gain content framework
The IG-9 Rubric grades a draft on five dimensions. Four dimensions are worth 0, 1, or 2 points. One dimension (freshness) is capped at 0 or 1. The max is 9. The ship threshold is 7.
| Dimension | What it measures | 0 | 1 | 2 |
|---|---|---|---|---|
| Proprietary data | First-party data only you can publish | Repurposes public stats | Combines public datasets in a new way | Original study, survey, or product data |
| First-hand evidence | Practitioner reps, teardowns, lived workflow | Generic "best practices" | One worked example | Multi-case practitioner teardown |
| Original framework | Named, scoreable, reusable rubric or model | No framework | Borrowed framework, applied | Named original framework with rubric |
| Expert attribution | Verifiable named expert with public track record | No attribution | One quoted source | Multiple named experts with linked work |
| Freshness hook | A 2026–2026 shift that changes the answer | None | One named shift tied to a stat | (capped at 1) |
The weights are not arbitrary. Proprietary data, first-hand evidence, original framework, and expert attribution are the four signals that consistently show up in pages that get cited by AI Overviews, ChatGPT, and Perplexity. Freshness caps at 1 because freshness without one of the other four is a news brief, not a citation asset.
How to score each dimension with examples
Score in the brief, before you commission the draft. A 30-minute pass is enough.
Proprietary data (0 to 2)
Score 0 if the draft cites only public stats anyone can pull from Statista. Score 1 if you combine two public datasets in a way no one else has published. Score 2 if you publish first-party data: a customer survey, a usage benchmark, an internal A/B test, or original research. Once published, you become the source other articles cite. That compounding effect is what Animalz documented.
First-hand evidence (0 to 2)
Score 0 for generic best practices. Score 1 if the post walks through one worked example with outcomes. Score 2 if it shows a multi-case practitioner teardown: how three teams ran the same workflow and what they did differently. Clearscope's Amanda Johnson makes the point in the webinar embedded above: information gain at scale comes from converting customer calls and product reviews into citable claims.
Original framework (0 to 2)
Score 0 if there is no framework. Score 1 if you apply someone else's framework (PEST, AARRR, Jobs-to-be-Done) to a new context. Score 2 if you introduce a named, scoreable rubric like IG-9 that another team can run on their own work. Named frameworks become citation magnets because LLMs grab them verbatim.
Expert attribution (0 to 2)
Score 0 if no named expert appears. Score 1 for one quoted source. Score 2 for two or more verifiable experts whose own work the article links to. The expert needs a search footprint: Mike King at iPullRank, Kevin Indig at Growth Memo, Aleyda Solis on AI search. Quoting "industry experts" without names scores 0.
Freshness hook (0 to 1)
Score 1 only if the post is anchored to a 2026–2026 shift that changes the answer for the reader, paired with a stat or named event. Score 0 if the freshness is cosmetic (a year in the title with no substance behind it).
Interpreting the score: ship-blocker, needs work, ship-ready
The score band tells you what to do, not just whether to ship.
| Score | Band | What it means | Recommended action |
|---|---|---|---|
| 0–3 | Ship-blocker | Commodity content with no information gain | Kill the page or change the angle |
| 4–6 | Needs work | Some gain, not enough to earn a citation | Add evidence, narrow audience, or pair with proprietary data |
| 7–9 | Ship-ready | Crosses the threshold; safe to commission | Brief the writer; QA on output |
Why 7 is the hard gate
Seven matches Google's Quality Rater Guidelines emphasis on first-hand experience, expertise, and original information (Google Search Quality Rater Guidelines overview). A post that hits 7 has cleared three of the four 2-point dimensions and contributes content the SERP cannot dilute. A 6 is close but usually missing either proprietary data or a named framework.
What to do at 4 to 6
This is the band where most drafts die quietly because no one calls it. Three concrete moves: pull one customer dataset to raise proprietary data to 2; commission a 20-minute expert interview to take attribution to 2; or introduce a named rubric to push original framework to 2. If none of those are feasible, kill the page and reallocate. The skill of saying no is what compounds.
Before and after: a commodity post versus an IG-9 post
The contrast makes the rubric concrete. Consider the topic "what is product analytics."
| Dimension | Commodity 2,500-word guide | IG-9 rewrite |
|---|---|---|
| Proprietary data | 0 (Statista charts only) | 2 (in-product usage benchmark) |
| First-hand evidence | 1 (one screenshot) | 2 (three customer migrations) |
| Original framework | 0 | 2 (named buyer-fit rubric) |
| Expert attribution | 0 | 2 (two named PMs with linked work) |
| Freshness hook | 1 (2026 in title) | 1 (Amplitude pricing change) |
| **Total** | **2 / 9** | **9 / 9** |
The commodity post scores 2 and reads like 30 other ranking articles for the same query. The IG-9 rewrite scores 9 and contributes four assets no one else has. The second post is shorter and harder to write. It is also the one that gets cited.
How to install the IG-9 gate in your editorial workflow
A rubric only works if it lives upstream of the writing.
Where the gate lives
Install IG-9 in the brief, not the draft. Each brief carries a scored field per dimension and a total. If the brief scores below 7, the draft is never commissioned. This kills the cost-of-rework problem: roughly 80% of editorial cost lives in the draft, so scoring early saves the most. Metaflow's own pipeline (`blog-publish-requirements.mjs` and `blog-brief-schema.mjs`) hard-codes the same five dimensions and the same 7-point threshold. We refuse to write the post if the brief does not clear the gate.
Who scores and who overrides
Use two raters per brief: the editor and a subject-matter reviewer. If scores diverge by more than 2 points, escalate. Override authority sits with one named editor, not a committee. That is the governance pattern Mike King recommends for content engineering, and it is what a GTM engineer does for SEO operations more broadly. If you are spinning up the editorial side of an AI agent stack, Claude skills for SEO is a useful next read.
What happens when a draft fails
Three options: rewrite the brief with new evidence, narrow the audience until you can hit 7, or kill the page. The instinct is to ship a 6 because the work is done. Resist it. Shipping a 6 trains the pipeline to produce 6s. The bigger payoff is the topics you stop commissioning: tighter budget, fewer pages, higher AI-citation rate per page. For broader context, see what AI search visibility means for growth teams, query fan-out in SEO, and how to build AI agents that actually get stuff done. More playbooks live in the Metaflow learning center.
The IG-9 is one rubric, not the only rubric. The point is to have one. Information gain is no longer a guideline; it is a budget decision, and budget decisions need numbers.
Frequently Asked Questions
What is the information gain content framework?
An information gain content framework is a scoring system editors use to grade how much new information a draft contributes before it ships. The IG-9 Rubric is one such framework: it scores drafts on five dimensions (proprietary data, first-hand evidence, original framework, expert attribution, freshness hook) and uses 7 out of 9 as the publish threshold.
What is the IG-9 Rubric?
The IG-9 Rubric is Metaflow's pre-publish scoring system for information gain. Four dimensions score 0–2; freshness caps at 0–1. Total: 9. Ship threshold: 7. The same five dimensions are hard-coded in Metaflow's editorial pipeline so every post is gated before drafting.
Is information gain a Google ranking factor?
Information gain is described in Google's granted patent US12013887B2. The patent governs ranking of a second set of documents based on what the user has already seen, with strong applicability to AI Overviews. Whether it is a live ranking factor in the classic ten-blue-links sense is unconfirmed; either way, it maps directly to how AI synthesis engines select citations.
What score does a draft need to ship?
At least 7 out of 9. 0–3 is a ship-blocker; 4–6 means rework before commissioning the draft.
How long does it take to score a draft with IG-9?
A trained editor can score a brief in under 30 minutes. Scoring happens at the brief stage, before the writer is engaged, so the rubric protects roughly 80% of editorial cost from being spent on commodity content.
Can AI write content that passes the IG-9 Rubric?
Only when given proprietary inputs: original research, real customer transcripts, a named framework, and verified expert quotes. AI cannot generate those inputs on its own. The score is determined by the inputs you supply, not the model that drafts the prose.
