LEARN
The Equaliser Effect: Who Gains Most

IN ONE SENTENCE
After applying these methods, pages ranked fifth gained 115.1% in visibility while first-ranked pages lost 30.3% on average — the return is higher for brands outside the top tier.
The Princeton GEO paper (KDD 2024) found an asymmetry: after applying the three writing methods, pages originally ranked fifth gained most in visibility (+115.1%) while first-ranked pages lost about 30.3% on average.
OUR POSITION
This has a practical consequence: if you already lead a subject, the incremental gain is limited and there is a real risk of bloating content that worked because it was concise. Concentrate effort on pages close to breaking into the top ten.
The specific numbers
The paper measured across GEO-bench — 10,000 queries over 9 domains — on Position-Adjusted Word Count and subjective impression.
The three effective methods produced a 30–40% relative lift overall; split by starting position, fifth gained +115.1% while first lost 30.3% on average.
Why this might happen
One reasonable reading is that first-ranked content was already the most usable version, and adding length and quotation diluted its liftability.
Content in the middle was missing exactly the checkable basis, and supplying it moved it from candidate to used.
⚠️ That is an interpretation of an observation, not a mechanism the paper establishes — platform mechanisms are undisclosed, so any mechanism story is inference.
Two implications for investment
First, prioritise the 11–20 and 4–10 bands. Since this category returns almost nothing past eleven, pushing those into the top ten has the highest marginal return.
Second, be careful editing content already at position one. Record the current state before changing it and re-measure after; if visibility drops, reverting beats further tuning.
Data behind this page
+115.1% / −30.3%
Visibility change for pages ranked fifth versus ranked first
Source:The 'equaliser effect' in the Princeton GEO paper, KDD 2024,2024
30–40%
Relative visibility lift from adding statistics / citations / quotations
Source:Princeton GEO paper, KDD 2024, GEO-bench 10,000 queries,2024
Sources
- [1]GEO: Generative Engine Optimization.Aggarwal et al., KDD 2024.2024
Updated 2026-08-10