The Information-Gain Content Heuristic
Judge content by what new it adds, not how typical it is
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 3
- Confidence
- 85%
A heuristic for producing content that wins in search, answer engines, and general audiences: did you say something others didn't (information gain), how typical are you (so typical you look like a rewrite of someone else's content), and did you bring original research or domain expertise. It doubles as a diagnostic for why most SEO content is worthless — non-experts endlessly rewriting each other with content-scoring tools that optimize toward the average.
Origin
Ethan Smith's proposed solution to the 'everyone rewrites each other's content' problem, offered as how he would build an algorithm to rank quality; Lenny notes it applies to newsletters and podcasts too.
Core principles
- 01Information gain: reward saying something no one else said
- 02Typicality penalty: if you read like an averaged rewrite, you add nothing
- 03Original research and stated domain expertise raise quality
- 04Content-scoring tools that push you toward what everyone else says produce the sameness problem
How to run it
- 1
Test for information gain
Before publishing, ask whether the piece contains a point, data, or insight that the existing articles on the topic do not already contain.
Pro tip Original research is the strongest form of information gain because it can't be rewritten from others.
Watch out Running everything through a 'be more typical' scoring tool actively removes your information gain.
- 2
Check your typicality
Assess whether your content is so average that it reads as a rewritten version of someone else's article; if so, it will be treated as a derivative.
Watch out The cheap path — hiring a non-expert to rewrite a competitor's article — is exactly what produces low-typicality-value, unrankable pages.
- 3
Show domain expertise
Bring and explicitly surface genuine expertise (e.g., a certified professional's perspective) so the content demonstrates authority beyond a paraphrase.
Pro tip Concentrate spend on the few pages that matter and write them extremely well rather than thinly funding many rewrites.
Watch out Expertise signals may be under-weighted by current algorithms, so treat this as future-proofing plus genuine quality, not a guaranteed ranking lever today.
In the wild
Graphite's analysis found one out of 20 landing pages drives roughly 85% of traffic, so 19 of 20 pages drive little to nothing. The cheap industry response is non-experts rewriting each other's content to cover many pages.
→ Concentrating money on the few pages that work and writing them well beats spreading spend across rewritten pages.
Common mistakes
Optimizing toward typicality
Using scoring tools that tell you to add whatever everyone else already said makes your content an averaged rewrite with no information gain, which is what fills search results with unhelpful content.
Is it for you?
Best for
Content leads, newsletter writers, and marketers deciding what makes a piece worth publishing
Not ideal for
High-volume programmatic content operations optimizing purely for coverage and cost per page
From the transcript
“one out of 20 landing pages drive roughly 85% of all your traffic”
“did you say something that somebody else didn't say”
“Are you so typical that I think that you're a rewritten version of somebody else's content”
“ideally did you do original research and do you have some domain expertise”
From the episode
The ultimate guide to AEO: How to get ChatGPT to recommend your product
Ethan Smith (Graphite)