Blog

GEO, the five studies that could prove us wrong: the strongest published evidence against our position, and the conditions that would change our mind.

Test GEO on your own site in 30 days: our protocol with a control group, a baseline, targeted changes and repeated measurements you can read honestly.

What GEO really costs: 55% of marketers have a GEO budget yet no practitioner discloses an amount. Our rates, and the GEO services we refuse to bill for.

Publisher traffic decline: minus 58%, minus 70% or minus 10%? Each camp measures something different with the same word. How to read AI traffic loss numbers.

16% of the sources AI cites are written by AI, per an audit of 712 queries on four engines. Why GEO rewards what degrades the corpus, and what to do.

Manipulating AI answers: injection attacks work in the lab, but a 1,000-site black-hat testbed shows 99.78% neutralization in production. The real weakness.

The sources AI prefers are not yours: generative engines favor third-party sources over brand sites, so most GEO budgets target the wrong place.

GEO democratizes nothing: across 100,000 AI answers, global brands reach 73% visibility and niche brands start at 11%. What it changes for a small business.

Position bias in AI answers: model accuracy follows a U-shaped curve by position in the context, so part of GEO visibility is a placement artifact.

GEO tracking tools measure noise: visibility in AI answers is a distribution, not a point, so one monthly measurement proves nothing. The minimum protocol.

AI citations misrepresent their source 30.6% of the time, across 761,495 citation pairs. For a brand, the real GEO risk is not absence but error.

The GEO market is booming while usage is flat: the Reuters Institute sees no growth in chatbot use for news, as 55% of marketers open a GEO budget.