We Analyzed 50,000 News Articles with AI. Here's What They Tell Us About the Stock Market.
Over two months, we scored the sentiment of every business, politics, and geopolitics article from 80+ major news outlets — then compared it against S&P 500 daily performance. The market didn't care about the level of negativity. But it did respond to changes in direction.
50,464
ArticlesBiz, Politics, Geo-0.21
Avg SentimentPersistently negative+1.7%
S&P 500 ReturnDec 15 — Feb 67,002
S&P PeakJan 28 — first ever+0.38
Strongest SignalSentiment Δ → marketThe Setup
Using Flogent's Sentiment Index Pro, we collected and analyzed 50,464 articles published between December 15, 2025, and February 7, 2026. Each article was scored on a scale from -1.0 (very negative) to +1.0 (very positive) using a proprietary AI pipeline that combines transformer-based sentiment analysis with lexicon scoring.
We split the data into two cohorts: Business & Economy — 8,800 articles averaging 160 per day, and All Market-Relevant — 50,464 articles including politics and geopolitics, averaging 1,069 per day. On the other side: S&P 500 daily closing prices covering 37 trading days.
Finding #1: The News Is Relentlessly Negative
Across all 55 days we tracked, the average daily sentiment for market-relevant news was -0.21. Business-only articles scored slightly better at -0.09, but still negative every single day except one.
January 1st was the only day that business sentiment crossed into positive territory — barely, at +0.006 — driven by optimistic new-year outlook pieces. Every other day was negative. Meanwhile, the S&P 500 gained 1.7% and hit 7,000 intraday on January 28.
All Market-Relevant Sentiment vs S&P 500
1,069 articles/day avg · Business, Politics & Geopolitics combinedBusiness & Economy Sentiment vs S&P 500
160 articles/day avg · Business & Industry + Economy & MarketsFinding #2: Absolute Sentiment Has Weak Correlation with Market Moves
We ran several correlation analyses between daily sentiment scores and S&P 500 returns. None were strong enough to act on. The same-day correlation of +0.27 means there's a slight tendency for more negative days to coincide with down days, but it's far from predictive. Looking ahead — using today's sentiment to forecast tomorrow's market — the signal essentially disappears.
Finding #3: The Direction of Sentiment Change Matters More Than the Level
This was the surprise. When we stopped looking at the absolute sentiment score and instead measured the day-over-day change, a stronger pattern emerged.
The Strongest Signal We Found
The correlation between daily sentiment change and S&P 500 returns was +0.38 for business-focused articles — the highest signal in our entire analysis.
When sentiment improved from one day to the next — even if it was still deeply negative — the S&P 500 tended to rise. When sentiment deteriorated further, the market tended to fall.
It's not about how bad the news is. It's about whether the news is getting better or worse.
This is consistent with how professional traders think about markets. The adage “buy the rumor, sell the news” captures a similar idea: markets price in expectations, and what moves them is the delta between what was expected and what materialized.
Correlation Breakdown
| Metric | Business | All Market | Interpretation |
|---|---|---|---|
| Sentiment Change → S&P Change | +0.38 | +0.21 | Strongest signal — improvement in sentiment aligns with gains |
| Same-day sentiment vs S&P change | +0.27 | +0.21 | Weak — absolute level has limited connection |
| Negative article ratio vs S&P change | -0.18 | -0.19 | Consistent — more negative articles slightly predict declines |
| Today's sentiment → tomorrow's S&P | -0.15 | -0.08 | Very weak — today's news doesn't predict tomorrow |
Daily Positive vs Negative Article Count (All Market-Relevant)
The Notable Moments
December 17 — Alignment. Business sentiment hit its worst point (-0.25) after the Federal Reserve signaled fewer rate cuts than expected. The S&P 500 dropped 1.2%. Both data streams told the same story.
January 20 — Divergence. The S&P 500 fell 2.1% — its worst day in our window. But sentiment that day was only moderately negative (-0.21), no worse than average. The market moved on factors the news tone alone didn't capture.
January 24–26 — Resilience. Three consecutive days of the deepest negativity in our data (-0.30, -0.30, -0.27). The S&P was flat and then rose. Markets shrugged off peak gloom.
January 28 — Historic milestone. The S&P 500 crossed 7,000 intraday for the first time in history while our all-market sentiment sat at -0.23. AI/DeepSeek excitement drove action despite cautious coverage.
February 5 — Alignment again. Sentiment dropped to -0.26 and S&P fell 1.2%. Tariff fears hit both coverage and market simultaneously.
What This Means
News sentiment alone is not a market predictor. Over 37 trading days, we found weak-to-modest correlations between daily sentiment and S&P 500 returns. If you're using news mood as your only signal, you're working with noise.
The change in sentiment carries more signal than the level. A shift from very negative to moderately negative was more informative than the score itself. A sentiment-based signal should be momentum-based, not threshold-based.
Business-specific news carries a cleaner signal than the broad firehose. The narrower dataset (160 articles/day) produced stronger correlations than the broader set including politics and geopolitics (1,069 articles/day). More data isn't always more signal.
Important Caveats
This analysis covers 37 trading days — a small sample. Correlation does not imply causation. News sentiment captures public narrative but misses insider flows, algorithmic signals, options positioning, and macro data releases. The persistent negative skew in news is a known phenomenon — financial journalism skews toward threats and problems. A longer study with sector-specific breakdowns would produce more robust conclusions.
How We Built This
This analysis was built using Flogent's Sentiment Index Pro, which tracks daily sentiment across 80+ major global news outlets. Articles are scored using a proprietary AI pipeline that includes theme classification, named entity recognition, and automated trend analysis.
Track sentiment for your industry
Flogent monitors how brands and topics are perceived across news, AI systems, and search — in real time.
Explore FlogentData period: December 15, 2025 — February 7, 2026 · 50,464 articles analyzed · 37 S&P 500 trading days
S&P 500 data: Yahoo Finance (^GSPC) · Sentiment data: Flogent Sentiment Index Pro
© 2026 Flogent. All rights reserved.