Becoming the answer ChatGPT, Claude, Gemini and Perplexity cite — sector by sector. The GEO method, authority signals, and KaiZen Barometer data to steer your AI visibility.
949 citation failures analysed: most relevant pages an AI engine skips miss the question’s intent or entities, not the style. Targeted fixes chosen by diagnosis beat generic rewriting while changing about 5% of the text.
37,927 AI agent journeys: sites agents can read were clearly recommended almost twice as often. The others were described by someone else.
Researchers gave an AI retrieval system placebo headings. They bought nothing. What helps is where the text is cut, and headings that tell the truth.
A National Taiwan University study shows an AI recommender that waits until the need is clear, then replaces a rejected product with its nearest neighbour. What it changes for measuring AI visibility.
AI Overviews reached France on 22 July. Google's roadmap read with a measurement lens: the figures, the verification step, and a baseline before peak season.
Being cited by an AI engine is not being recommended. Why AI share of voice must separate citations, mentions and recommendations, with the formulas to measure each.
Retrieve-for-Train names the flaw behind many AI visibility panels: paraphrastic collapse. What the study shows, what it doesn't say about AI Mode, and how to measure.
Since 15 September, Cloudflare separates Search, Training and Agent. Block and Disallow AI Training no longer do the same thing — and the third control is the one nobody is watching.
No Google tool allows it. Search Console shows your impressions only, AI Overviews and AI Mode combined. The observational protocol, and its five conditions.
An AI crawler can get a 403 on a public page that allows it. And OpenAI's advertising crawler is Cloudflare-verified while its search crawler is not.
SEO decides whether a system can index your pages. GEO decides whether an AI keeps them as evidence. Indexing, AI crawlers, corroboration, measurement: what each one solves.
AI engines supposedly agree on brands and diverge on sources. Measured against chance the order inverts: 42× for a shared page, 17× for a shared brand.
ChatGPT cites 6.88 sources per prompt, Perplexity 16.35 — yet uses each one more than four times as deeply. And Q&A formatting does not improve absorption.
The search web is becoming a recommendation web: AI, Social Search, Dark Social, Retail Media and creators are reshaping brand discovery.
Nine agencies scored on what they can prove, five audited on 25 criteria. Primal leads the market, KAIS leads on method. Full grid published.
Google shipped a whole infrastructure for paid and a single sentence for organic. Two economies are forming inside the AI answer — and one of them isn't for sale.
The correlation between Google's Top 10 and answer-engine citations fell from 76% to 38%. Three out of four citations now come from sources outside the Top 10.
82% of French firms are invisible on Perplexity; one firm owns the market at KVS 100. The Muda concept, the three KaiZen levers, and the niche strategy that beats a giant.
A director looking for a lawyer now asks ChatGPT before Google. GEO decides which firms the AI cites — and it rewards nothing like the same players as SEO.
Thirty legal queries, three AI engines, 1,870 citations. Bredin Prat crushes business law while the international giants vanish from the ranking.
No published article in this sector yet — the pillar is on its way. Test your AI visibility in the meantime →
Find out where you appear (or don't) in ChatGPT, Claude, Gemini and Perplexity. Your KaiZen Visibility Score and three concrete actions, in 48 to 72 hours.
Get my free audit →