Project update — June 2026: Canary 2.0 concluded on 4 June 2026 following completion of its pre-registered hypothesis test. The content on this page remains live as a record of the project methodology. For the full project conclusion and results, see the Project Conclusion and Final Report.
Every day, hundreds of articles are published about AI companies. Earnings reports, product launches, regulatory decisions, funding announcements, geopolitical developments. No individual can read them all, and even those who try cannot reliably turn what they read into a consistent, comparable signal across the whole sector at once.
Canary does.
What AI sector sentiment means
Sentiment here is not market sentiment in the traditional financial sense. It is the measurable tone of the news being written about AI companies on any given day, whether coverage is predominantly positive, negative, or neutral, and by how much. Journalists and analysts writing about AI companies are processing earnings calls, regulatory filings, product announcements, and executive commentary. The aggregate tone of that coverage, scored consistently and daily, is a real signal.
What Canary tracks
Six categories that together represent the principal layers of the AI value chain: AI Chips and Hardware, AI Platform Hyperscalers, AI Enterprise Software, Data Centre Infrastructure, AI GPU Cloud, and Pure Play AI.
Each morning Canary processes the previous day’s articles, scores each for sentiment and relevance, and produces a daily signal for each category with a 95% confidence interval and a statistical significance flag comparing the reading against that category’s own recent history.
Narrative framing
Sentiment scores tell you whether coverage is positive or negative. They do not tell you what kind of story is being told. Two days with identical sentiment scores can represent entirely different situations.
Canary addresses this through narrative frame classification. Every article is assigned to one of eight frames: Growth Momentum, Technical Breakthrough, Financial Results, Regulatory Risk, Geopolitical Risk, Competitive Threat, Market Correction, or Macro Environment. The distribution of frames on any given day shows what the sector conversation is actually about, not just how it feels.
A sustained shift toward Regulatory Risk and Market Correction frames, even before sentiment scores fall sharply, is often the earliest detectable sign that the character of coverage is changing.
The Semantic Volatility Index
The Semantic Volatility Index is a composite early-warning measure that aggregates five sub-components to detect narrative instability before it registers in headline sentiment figures.
Two of those sub-components use Jensen-Shannon Divergence to measure how different this week’s AI sector language is from last week’s. When the words being used to describe the sector change, that change typically precedes the sentiment shift that follows by several days.
The dashboard
The Canary Dashboard is updated every morning. It shows the overall AI sector sentiment for the day with a trend chart, per-category scores with confidence intervals and historical comparison, a 14-day sentiment heatmap across all six categories, the narrative frame distribution, and the Semantic Volatility Index with sub-component breakdown. It is free and requires no registration.