Research
Original studies built on the AIVIS corpus of real scanned domains — every number reproducible from an open dataset.
The AI Readiness Index 2026
How ready is the web for AI answer engines? Measured across 1,831 real domains, not a survey.
Open dataset (CC-BY)
The Agent Commerce Index
How readable are online stores to AI shopping agents, by platform? Measured across 20,149 real stores, not a survey.
Open dataset (CC-BY)
AI Brand Concentration Report 2026
How concentrated AI recommendations are across 62 software categories, and where ChatGPT, Perplexity, Gemini and Claude disagree.
Open dataset (CC-BY)
llms.txt Adoption Report 2026
How many websites actually publish an llms.txt file for AI assistants? Measured on 1,459 real websites, not a survey.
Open dataset (CC-BY)
AI Crawler Blocking Report 2026
Which AI crawlers do websites shut out, and which do they let in? Measured on 1,401 real robots.txt files, not a survey.
Open dataset (CC-BY)
Structured Data Coverage Report 2026
How many websites describe themselves to machines with schema.org, and with which types? Measured on 1,459 real homepages, not a survey.
Open dataset (CC-BY)
JavaScript Rendering Report 2026
How many sites are invisible to AI crawlers because their content needs JavaScript? Measured on 450 real homepages, not a survey.
Open dataset (CC-BY)
AI shopping agents: buyer questions answered by online stores
Shipping cost, delivery time, return window, cancellation, in-store pickup, returns of opened items: what an agent finds before it buys for someone.
Open dataset (CC-BY)
Sources
On a benchmark of 10,000 queries, adding statistics, quotations and cited sources raised a page's visibility in generative-engine answers by up to 40%. Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
AI search engines answer with retrieval-augmented generation, introduced by Lewis et al. in 2020: they fetch live web pages first, so a page they cannot read cannot be cited. Lewis et al., Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (NeurIPS 2020)