Last week’s note covered the two Google patents behind information gain and why the mechanics matter more than most SEOs realize. The short version: Google can measure whether a page says something the other pages on the same topic don’t, down to specific entities, numbers, and the relationships between them. Content that adds something new has a structural advantage. Content that restates what already exists adds to the pile.
This week is the practical side. I built a Claude skill that runs a systematic information gain analysis for any topic or search query. Give it a keyword, and it tells you what the current SERP covers thoroughly (your topical foundation), what it misses (your opportunities), and where competitor content is stale.
The Problem This Solves
Most content workflows go: keyword research, content brief, write. What’s usually missing is a deliberate look at what the existing results already cover and, more importantly, what they don’t.
Without that step, you end up producing content that reads like a remixed version of everything already ranking. The patents note explained why that’s a structural disadvantage. The top results have already provided those information nuggets. Your page, with the same nuggets arranged slightly differently, registers as redundant.
This skill fills that gap. It turns information gain analysis into a repeatable process you can run before writing anything.
What the Skill Does
Give it any topic or search query, and it runs a six-step analysis:
- SERP Content Analysis. Searches the top 10 results and catalogs what each one covers: questions answered, entities included, depth of coverage, and any unique angles. It tracks specific entities (names, numbers, tools, data points), not vague summaries.
- Topical Foundation Mapping. Identifies the well-covered areas that form the baseline for topical authority. What do all or nearly all results cover? What examples does everyone use? These are essential to cover thoroughly, not because they’ll differentiate you, but because skipping them signals incomplete coverage. Information gain is the layer you build on top of this foundation.
- Gap Detection. The most important step. It looks for questions in People Also Ask that are poorly answered, entity attributes mentioned but never explained, user intent aspects not fully addressed, and expert-level depth missing from general content.
- Information Gain Scoring. Every potential addition gets scored by how many of the top 10 results cover it:
- High IG: Not covered by any result. Must include.
- Medium IG: Covered by 1-2 results. Strong include.
- Low IG: Covered by 3-5 results superficially. Include with more depth.
- None: Covered by 6+ results. Topical foundation. Cover thoroughly.
- Unique Angle Identification. Recommends specific differentiation strategies. Not “add unique insights” (that’s useless advice), but concrete recommendations like “include actual cost breakdowns, none of the top 10 results provide specific pricing.”
- Content Freshness Assessment. Checks when competitor content was published and flags outdated information, recent developments not yet covered, and whether gaps are evergreen or time-sensitive.
The output is a structured Information Gain Map with four sections: Topical Foundation, High-IG Opportunities, Differentiation Strategies, and Freshness Opportunities. It also includes a content structure recommendation that prioritizes high-IG content early in the piece, because the 2006 patent explicitly weights content near the top of the document more heavily.
What Are Claude Skills?
If you haven’t used them before, skills are reusable instruction packages for Claude. Instead of writing a detailed prompt every time you want Claude to do something specific, a skill teaches it a workflow it can run consistently. You install it once, and Claude follows it automatically whenever you ask it to do that type of work.
Skills work in Claude.ai (web and app), Claude Code, and the API. You need a paid Claude plan (Pro, Max, Team, or Enterprise).
I covered skills in more detail in the Entity Analysis note if you want more background.
How to Install It
- Download the skill file: DOWNLOAD LINK
- Open Claude.ai
- Go to Settings → Capabilities
- Scroll down to the Skills section and upload the file
- Make sure the skill is toggled ON
- Start a new chat
To use it, give Claude a topic or search query and reference the skill:
- “Using the information gain skill, analyze ‘how to soundproof a home office'”
- “Using the information gain skill, find opportunities for ‘best backyard vegetable garden for beginners'”
- “Using the information gain skill, analyze ‘how to train for a half marathon'”
The skill requires web search to function properly. If web search isn’t available, it will flag that limitation in the output.
Sample Output: “How to Soundproof a Home Office”
Here’s a condensed version of what the skill produces. The actual output is longer and more detailed.
Information Gain Map: How to Soundproof a Home Office
SERP Summary
The top 10 results are almost entirely lifestyle and home improvement listicles. Most follow the same structure: explain soundproofing vs. acoustic treatment, then list 5-10 tips (rugs, acoustic panels, solid core doors, weatherstripping, curtains). Very few include specific performance data, cost breakdowns, or guidance on prioritizing fixes by impact. Several results are 3-5 years old.
Topical Foundation (Baseline Coverage)
| Topic | Coverage | Notes |
|---|---|---|
| Replace hollow-core door with solid core | 9/10 | Universally recommended. Cover the why (mass blocks sound transmission) and include cost context. |
| Add weatherstripping and door sweep | 8/10 | Most results list this but don’t explain that air gaps are the single biggest sound leak. Worth emphasizing the physics. |
| Acoustic foam panels on walls | 8/10 | Cover thoroughly but correct the common misconception. Foam helps echo and reverberation, not sound transmission through walls. Most results blur this distinction. |
| Area rugs on hard floors | 7/10 | Cover as part of impact noise reduction. Most results mention it generically. |
| Soundproof curtains | 7/10 | Include with honest assessment of limited effectiveness for sound blocking vs. absorption. |
| White noise machine | 6/10 | Cover as a supplementary approach, clearly distinguished from actual soundproofing. |
High-IG Opportunities
| Opportunity | IG Score | Implementation |
|---|---|---|
| STC ratings for common wall assemblies | High | Only 1 result mentions STC ratings at all. Include a reference table: standard interior wall (STC 33-35), with added drywall + Green Glue (STC 45-50), with resilient channel (STC 50+). Readers have no way to compare options without this. |
| Actual cost ranges by method | High | No results provide specific pricing. A comparison showing weatherstripping ($15-30) vs. solid core door ($150-400 installed) vs. additional drywall layer ($200-500 per wall) vs. window inserts ($300-800 per window) would be genuinely new. |
| Prioritized approach by noise source | High | Most results list tips randomly. No one structures the advice by “if your noise comes from X, start with Y.” A decision-tree approach based on whether noise is airborne vs. impact, internal vs. external, would be novel. |
| HVAC noise and ductwork | Medium | Only 2 results mention that sound travels through ductwork and vents. Most readers with home offices have HVAC running through the room. Covering duct lining, flex duct, and vent covers would fill a real gap. |
| Realistic expectations by budget tier | Medium | No result tells readers what level of improvement they can realistically expect at different price points. A framework like: $50-100 (noticeable, 5-8 dB reduction), $200-500 (significant, 10-15 dB), $1,000+ (substantial, 20+ dB) would be unique. |
Differentiation Strategies
| Strategy | Specifics |
|---|---|
| Before/after dB measurements | Include actual decibel measurements using a free phone app (NIOSH SLM) for common scenarios. No competing result includes measurable data. |
| The “diminishing returns” framework | Explain that the first $100 in soundproofing gets you more improvement than the next $500. Seal gaps first, then add mass. Most results list tips as if they’re all equal. |
| Zoom/Teams-specific audio tips | Several results mention video calls but none cover software-side settings (noise suppression in Zoom/Teams, microphone selection, virtual background audio isolation). This is a user intent gap since remote workers are the primary audience. |
Freshness Opportunities
| Area | Competitor Age | Opportunity |
|---|---|---|
| Product recommendations | 3 results from 2020, 2 from 2023 | Window insert technology and acoustic panel options have improved. Current products and pricing needed. |
| Software noise suppression | Not covered at all | Zoom, Teams, and Google Meet have all added AI noise suppression since most of these articles were written. This entire angle is unaddressed. |
Sample Output: “How to Start a Backyard Vegetable Garden”
Here’s a second example showing how the skill works on a different type of query.
Information Gain Map: How to Start a Backyard Vegetable Garden
SERP Summary
Dominated by authority sites (Old Farmer’s Almanac, university extensions, Gardener’s Supply) and established gardening blogs. Content quality is generally high but formulaic. Almost every result follows the same structure: pick a sunny spot, start small, improve soil, choose easy crops, water consistently. Very little regional specificity, almost no first-year cost data, and minimal guidance on what to expect month by month after planting.
Topical Foundation (Baseline Coverage)
| Topic | Coverage | Notes |
|---|---|---|
| Choose a spot with 6-8 hours of sun | 10/10 | Universal. Cover clearly but briefly. Every competitor handles this adequately. |
| Start with a small plot or raised bed | 10/10 | Well-covered but most results give generic size advice. Add specifics on why 4×8 works for beginners (reachable from both sides, fits standard lumber). |
| Add compost to improve soil | 9/10 | All results say this. Few explain how much compost or what kind. Opportunity to be more specific even within the foundation. |
| Beginner crop list (tomatoes, lettuce, beans, zucchini) | 9/10 | Include but add zone-specific context that most results lack. |
| Raised beds vs. in-ground comparison | 7/10 | Cover the tradeoffs. Most results default to recommending raised beds without explaining when in-ground is better. |
| Get a soil test | 6/10 | Worth covering with specifics on where to get a test and what to do with the results. Most results mention it and move on. |
High-IG Opportunities
| Opportunity | IG Score | Implementation |
|---|---|---|
| Realistic first-year cost breakdown | High | No result provides a total startup cost estimate. A breakdown showing raised bed materials ($80-200), soil/compost ($50-100), seeds vs. starts ($20-75), tools ($40-100), and watering ($0-150 for hose/timer) would be genuinely useful. |
| Month-by-month timeline for first year | High | Every result covers what to do before planting. None walk a beginner through what the first 6 months actually look like: what to expect in weeks 2-4, when the first real problems show up, when harvest starts, when to plant fall crops. |
| Specific failure points for beginners | High | Most results mention “common mistakes” in a generic list. None describe what failing actually looks like: blossom end rot on tomatoes (and why it happens), bolting lettuce in summer heat, squash vine borers, and what to do when each one hits. Showing the problems with photos and fixes would be novel. |
| Grocery cost offset analysis | Medium | Only 1 result mentions growing what you already buy. None calculate the actual economics: what a 4×8 raised bed of tomatoes produces vs. grocery store cost, which crops have the best ROI, which aren’t worth growing for savings. |
| USDA zone-specific planting windows | Medium | Most results are generic or region-specific to one state. A framework that says “enter your zip code at [Almanac frost date calculator], then use this timeline” would be more actionable than the vague “plant after last frost” advice every result gives. |
Differentiation Strategies
| Strategy | Specifics |
|---|---|
| The “what nobody tells beginners” section | Cover the unglamorous realities: how much time weeding actually takes per week, what pest damage looks like (not stock photos), why the first year’s yield is usually disappointing, and why that’s normal. This framing is absent from every result. |
| First harvest to table walkthrough | A concrete example of a single day’s harvest from a beginner-sized garden, what you picked, how much, and a simple meal made from it. Every result talks about the end state abstractly. None make it tangible. |
| Water cost and time estimate | Include actual watering requirements in gallons per week for a standard raised bed, plus time required. “Water consistently” appears in every result. Specific quantities appear in none. |
Freshness Opportunities
| Area | Competitor Age | Opportunity |
|---|---|---|
| Tool and product recommendations | Several results reference outdated products | Current raised bed kits, soil delivery services, and watering systems have changed. |
| Climate shift impact | Not covered | Growing zones have shifted in recent years. Some regions can now plant earlier or grow crops they couldn’t before. USDA updated the hardiness zone map in 2023 and most gardening guides haven’t incorporated the changes. |
How This Connects to Last Week’s Note
The Information Gain patents note explained the mechanics: Google breaks content into information nuggets and entity interactions, scores them by rarity across the corpus, and weights content near the top of the document more heavily.
This skill operationalizes those mechanics into a repeatable workflow. The SERP analysis identifies which nuggets form the topical foundation. The gap detection finds which ones are rare or missing. The scoring prioritizes what to include and the content structure recommendation tells you where to put it.
The patents explain why information gain matters. This skill tells you where to find it.
