What 400,000 prompts reveal about how brands enter AI answers
Patterns from daily monitoring across commercial, comparison, and category-discovery questions.
Read the study →Research
We study how AI systems find, judge and recommend brands, every day. Our research is the foundation of Digraph’s product, and we publish what we learn.
9AI systems monitored
400K+prompts analysed daily
5industries in the public index
Opendata and methodology
AI Visibility Index
Which brands AI systems recommend for high-intent buyer questions in automotive, beauty, finance, marketing software and retail.
Explore the index →AI search visibility software benchmark
A dated, reproducible snapshot of the category with scope, observations, limitations and a recheck method.
Read the benchmark →Benchmark dataset (JSON)
The machine-readable observations behind the benchmark, free to reuse with attribution.
Download the dataset →Studies and analysis
All journal articles →The new competitive field is the answer, not the result page
A framework for understanding which brands AI recommends and what evidence shapes the category.
Read the study →How ChatGPT, Gemini, and Perplexity recommend brands
The practical differences between major answer engines and what they mean for visibility work.
Read the study →Seven AI systems, one dashboard
What changes when brand mentions are monitored across every major AI system.
Read the study →The psychology behind AI brand recommendations
How authority, social proof, and cognitive shortcuts show up in answer-engine recommendations.
Read the study →Scarcity vs. social proof in paid search
What a large ad-variant review can teach teams about persuasion patterns.
Read the study →Which sources shape AI answers—and how to prioritize the ones that matter
A smarter way to map owned, earned, and third-party sources without reducing citation strategy to a checklist.
Read the study →The AI shopping shortlist: how product discovery shifts before the click
Why the products named early in an answer matter, and the signals retail teams should inspect before seasonal demand arrives.
Read the study →How much answer variance is normal? A guide to reading daily movement
Volatility does not make AI visibility unmeasurable. It changes how teams need to define a meaningful change.
Read the study →How we measure
Every number traces back to a recorded answer. We run defined prompt portfolios across AI platforms, markets and languages, store each response with its citations, and repeat the observation so that normal answer variance is not mistaken for change.
We publish scope and limitations with every study, and we never report a trend from a single answer.
Read the metric definitions →Cite or cover our research
Journalists, analysts and researchers are welcome to cite Digraph studies and reuse our open datasets with attribution to “Digraph (digraph.dev)”.
For data requests, commentary or a briefing with our research team, contact us.
Contact the research team →Put this into practice for your brand.
Digraph tracks how ChatGPT, Gemini, Perplexity and 6 other AI systems describe your brand, finds what to fix and plans the next move with our team.