Source: ai-research/similarweb-information-gain-2026-06-30.md — Similarweb (https://www.similarweb.com/blog/marketing/geo/information-gain)
Information gain measures how much genuinely new information a page adds versus the other pages already ranking or being cited for the same query. Similarweb’s study (Shai Belinsky, Jun 23 2026) found that most of the SERP is occupied by pages that say nothing the other pages don’t already say — the median top-3 page carries just 4 unique data points. The thesis: depth and specificity win AI-search visibility precisely in the categories where established brands publish generic, “mostly-shared” content.
Key Takeaways
- Information gain = how much new information a page adds versus pages already ranking/cited for the same query.
- “A substantial portion of the SERP is occupied by pages that say nothing more than the other pages do” — the median top-3 page carries just 4 unique data points.
- B2B SaaS and ecommerce sit at the bottom of the originality distribution — the majority of top-ranked content in those categories scores in the “mostly-shared” (low-originality) range.
- High-information-gain pages overperform on visibility in exactly the categories where incumbents are generic — depth and specificity beat brand weight there.
- Scoring is quantified: a page at 45/100 information gain with an 8% citation share across tracked prompts “has a quantified problem and a quantified target.”
- Pair information-gain audits with AI brand visibility tracking — “the two signals together tell you what fixing alone cannot.”
What the Study Found
- The median top-3 page carries just 4 unique data points — most of the page restates what competitors already cover.
- Blunt framing: “a substantial portion of the SERP is occupied by pages that say nothing more than the other pages do.”
- B2B SaaS and ecommerce score lowest on originality — the majority of top-ranked content in those verticals lands in the “mostly-shared” range.
- High-information-gain pages are the ones that overperform on visibility in categories where established brands are “mostly-shared” — i.e., depth and specificity win where incumbents are generic.
- Scoring example: a page that scores 45/100 on information gain with an 8% citation share across tracked prompts “has a quantified problem and a quantified target” — both the gap and the goal are expressed as numbers, making them trackable. ^[inferred: the quote frames 45/100 + 8% as a measurable baseline to improve against]
Pair It With AI Brand Visibility
- Similarweb’s momentum score measures citation acceleration — how fast a brand’s citation frequency is growing over the tracked period.
- B&H Photo posted an AI visibility momentum score of 296.9 — meaning its citation frequency grew nearly 3x over the tracked period.
- This is in a category where Apple sits at 100/100 on brand visibility — so a fast-rising challenger can show strong momentum even next to a category-defining incumbent. ^[inferred: momentum (acceleration) and brand-visibility level are distinct metrics; B&H’s 296.9 momentum sits alongside Apple’s 100/100 visibility in the same category]
- Guidance: “Track your AI brand visibility alongside your information gain audits. The two signals together tell you what fixing alone cannot.”
- Read together: information gain tells you whether a page deserves to be cited; brand visibility and momentum tell you whether the citations are actually accruing. ^[inferred]
Try It
- Run an information-gain audit on your top-ranked pages: count the unique data points each adds beyond what already ranks or is cited for the query — beat the median of 4.
- Prioritize B2B SaaS / ecommerce pages, where originality is lowest and the visibility upside from added depth is highest.
- Set quantified targets the way the study frames them: move a page from 45/100 toward a higher information-gain score and grow its citation share past 8%.
- Track AI brand visibility and momentum (citation acceleration) in parallel with originality fixes — content depth alone won’t tell you whether citations are landing.
- Remember the closing line — “The visit is going to happen. The question is whose site it goes to.” — and treat originality as the lever that decides whose.