[Blog]

Insights on AI Search and the agentic web.

Expert research, product updates, and practical guides on helping marketing teams win in the age of AI search.

Sitefire research chart, 'AI slop has a shape. You can see it without the words.': structural rarity of 13,500 blog posts by source, human posts concentrated in the rarest configurations at 0.84 against five AI models between 0.33 and 0.55
[Research]

How to Tell AI Slop From Human Writing: Ten Structural Tells From 13,500 Blog Posts

AI-generated blog posts have a shape below the words. Structure alone finds them at 98.0 macro-F1 in a 13,500-post corpus, and rewording does not remove it.

Jochen MadlerSep 4
Sitefire research chart, 'A citation doubles your AI visibility. Being sourced does not.': brand mention rate on cited versus not-cited days across Gemini, ChatGPT and Google AI Mode, from 354,004 AI answers
[Research]

Do AI Citations Actually Matter? What 354,004 AI Answers Show

Sitefire research across 354,004 AI answers on what an AI citation is worth, and why being sourced without being cited pays almost nothing.

Jochen MadlerJul 31
Sitefire research chart, 'AI models mention similar brands. But they cite different pages.': for the same prompt, pairs of ChatGPT, Gemini and Google AI Mode overlap on 30 to 36% of brand mentions but only 6 to 10% of cited web pages
[Research]

Why AI Models Recommend the Same Brands but Cite Different Sources

Sitefire research on why ChatGPT, Gemini and Google AI Mode cite different sources for the same prompt, and what it means for your AI visibility.

Jochen MadlerJul 21
Under the headline 'Google is liable for what its AI says', a Google AI Overview card asserting a publisher is known for dubious business practices, with the false claim marked, the court's ban of up to EUR 250,000 per violation, and cited sources that never made the claim
[Industry]

Google Is Liable for What Its AI Overviews Say: Inside the Munich Ruling

A Munich court held Google liable for false AI Overviews. The legal reasoning, the precedents, and what AI Overviews liability means for brands.

Jochen MadlerJun 10
Three-layer diagram of AI visibility measurement: AI answers, bot fetches via Cloudflare, and AI referral sessions in GA4
[Playbooks]

How to Measure GEO Beyond Prompt Monitoring

Prompt monitoring is a useful start, but it only measures synthetic visibility. The real GEO measurement stack adds Cloudflare bot data and GA4 referrals.

Jochen MadlerApr 18
Trustpilot FY2025 earnings card under the headline 'The first earnings call where AI visibility influenced the stock price': 1,490% growth in AI search click-throughs, a 25% one-day stock move, 5th most-cited domain on ChatGPT, 12 of 12 analysts on Strong Buy
[Industry]

Trustpilot's AI Search Earnings: The First Proof That AI Visibility Drives Revenue

Trustpilot's AI search click-throughs grew 1,490% YoY and the stock surged 25%. What the first AI search earnings report reveals about brand visibility.

Jochen MadlerMar 29
Sitefire research chart, 'Most cited domains are cited once': 43.3% of 129,904 cited domains appear in a single AI answer, 28.9% in two to five, with a long tail below, from 25M citations across 6 AI models
[Industry]

AI Content Is Good Enough - But Good Enough for What?

AI content can rank in Google. But can it earn citations in AI search? Analysis of 25M+ citations reveals what actually gets cited - and it's not production quality.

Jochen MadlerMar 19
Comparison of browser versus AI agent Accept headers, highlighting text/markdown as the signal that flags agent traffic
[Playbooks]

How to Detect AI Agents Like OpenClaw Visiting Your Website

AI agents like OpenClaw send Accept: text/markdown - a header no browser uses. Learn how to detect agent traffic on Vercel, Cloudflare, Nginx, and more.

Jochen MadlerMar 8
The prompt 'best CRM for small business' fanning out into six searches, the sources each one retrieved, and the single answer synthesized from them with its citations
[Explainers]

Query Fan-Out: What AI Search Actually Does With Your Query

AI search rewrites every query into 5-20 separate searches before answering. Learn how query fan-out works and why 68% of AI-cited pages aren't in the top 10.

Jochen MadlerMar 3
Sitefire visibility card next to the headline 'How to audit your AI visibility': visibility score over 14 days, competitor ranking, and one AI answer with the brand mention, sentiment and citation marked
[Playbooks]

How to Run a GEO Audit on Your Website

A 6-step framework to audit your AI search visibility across ChatGPT, Gemini, and more. Covers content structure, schema markup, and competitor benchmarking.

Jochen MadlerFeb 28
The AI search pipeline from prompt to citation: fan-out into 5 to 15 searches, a shortlist merged by RRF, a full page fetch, the answer with citations and brand mentions, and the website, with the foundation-model bypass below
[Explainers]

How AI Search Engines Actually Work: From Prompt to Citation

The full pipeline behind AI search: query fan-out, Reciprocal Rank Fusion, and why only 12% of ChatGPT citations overlap with Google results.

Jochen MadlerFeb 26
Two stat cards under the headline 'ChatGPT is still small compared to Google. But AI search is now everywhere.': AI engines at 5.2% of daily search sessions against 78% of Google sessions that already show an AI Overview
[Industry]

Global AI search engine market share

Google Search is now showing AI-infused content in 78% of sessions. ChatGPT has a 56% share of the AI search market.

Jochen MadlerJan 22