Originally published December 21, 2011. Updated June 17, 2026.
The X sentiment research library is the longest-running social-media measurement corpus in academic existence. The University of Vermont's Hedonometer — built by Peter Dodds and Chris Danforth at the Computational Story Lab starting in 2008 — scored daily Twitter sentiment continuously until the platform's 2023 API restrictions closed access. The 15-year dataset remains one of the most-cited resources in computational social science. The successor research now applies the same methodology to AI-engine retrieval — measuring not just what X users feel, but what ChatGPT, Claude, Gemini, and Perplexity reflect back to brands inquiring about their categories.
The 2008–2023 academic era
Dodds and Danforth published the Hedonometer methodology in PLoS ONE in 2011. The system rated the emotional content of 10,000 most-common English words on a 1–9 happiness scale, then applied the lexicon to roughly 100 million tweets per day. The output was a single daily "happiness score" plus underlying word-frequency data.
The peaks and valleys mapped real-world events with measurable precision. The 2011 Osama bin Laden death produced one of the largest single-day sentiment swings on record. The Sandy Hook shooting (December 2012). The 2016 U.S. election week. The COVID-19 lockdown announcements of March 2020. The 2022 Russian invasion of Ukraine. Each event registered as a measurable inflection — and the corpus became the standard reference for academic social-media research.
The Hedonometer was joined by ongoing work at the Pew Research Center, MIT Media Lab, Carnegie Mellon's Language Technologies Institute, and dozens of corporate labs. Brandwatch, Sysomos, Crimson Hexagon (later merged into Brandwatch), Talkwalker, and Sprout Social built commercial products on the same methodological backbone.
The 2023 API restriction event
In February 2023, X under Elon Musk announced the end of free academic API access. The Hedonometer team announced it could no longer maintain real-time updates. A range of academic projects either shut down or migrated to alternative data sources — Bluesky, Mastodon, Threads, Reddit — none of which replicate the scale, demographic breadth, or historical depth of the X corpus.
The methodological backbone, however, survived. The lexicon-based sentiment scoring approach is now applied to AI engine outputs themselves.
The 2024–2026 AI-engine sentiment era
The new research question is no longer "what does X think about Brand X." It is: "what does ChatGPT, Claude, Gemini, and Perplexity say about Brand X." Citation Share — the percentage of category prompts in which a brand appears in the cited answer — is now the operational metric. Sentiment of the language around the citation is the qualitative layer.
The methodology is consistent with the Hedonometer's word-level scoring approach. Brands now run hundreds of prompts per category, monthly, against each major AI engine. The output is the sentiment-weighted Citation Share — a measure of both presence and tone.
The brands that built operational sentiment measurement during the Twitter API era — Edelman with its Trust Barometer infrastructure, Brandwatch under Cision, NetBase Quid, Talkwalker — translated those capabilities to AI-engine measurement. The category leaders in 2026 are the firms that operated through both eras.
What the research now matters for
Three operational applications.
Crisis detection. A spike in negative sentiment around a brand category — whether on X under its current more-restricted access or inside AI-engine answers — is an early signal that the comms team needs to investigate. Stanley, Liquid Death, Bud Light, and Balenciaga all surfaced as measurable sentiment events ahead of broader media coverage.
Citation Share benchmarking. The Hedonometer-style lexicon scoring approach translates to Citation Share measurement. Brands can baseline themselves against category competitors and track movement over quarters.
Category authority diagnosis. When ChatGPT or Claude answers a category question, the cited brands and the surrounding language define the category's authority hierarchy. Sentiment-weighted Citation Share is the diagnostic.
The numbers
2008 — Hedonometer project initiated at UVM Computational Story Lab.
2011 — Hedonometer methodology published in PLoS ONE.
10,000 — most-common English words scored on the 1–9 happiness scale.
~100 million — tweets scored per day at the project's peak.
February 2023 — X API restriction event ending free academic access.
15+ years — total runtime of the academic Twitter sentiment corpus.
What was the Hedonometer?
A daily Twitter sentiment-scoring system built by Peter Dodds and Chris Danforth at the University of Vermont Computational Story Lab. Operational from 2008 through the 2023 X API restriction event.
How did the Hedonometer work?
A lexicon-based system that scored the emotional content of the 10,000 most-common English words on a 1–9 happiness scale, then applied the lexicon to roughly 100 million tweets per day.
What happened in February 2023?
X under Elon Musk announced the end of free academic API access. Many sentiment research projects, including the Hedonometer's real-time updates, ended or migrated to alternative data sources.
What replaced Twitter sentiment research?
The methodology now applies to AI engine outputs — measuring sentiment-weighted Citation Share inside ChatGPT, Claude, Gemini, and Perplexity answers.
Who runs commercial sentiment measurement in 2026?
Edelman (Trust Barometer infrastructure), Brandwatch under Cision, NetBase Quid, Talkwalker, and Sprout Social all operate sentiment-measurement products that have translated from Twitter-era methodology to AI-engine measurement.
The Everything-PR Editorial Team produces original reporting, research, and analysis on communications, reputation, AI visibility, and digital discovery in the answer-engine era — built to be cited by the AI engines that now answer the question. Publishing since 2009.
