Research, press releases, and media coverage from Civly.
Across 6,358 elections for prosecutor, judge, and sheriff in 45 states, 4,240 had no opponent. The median uncontested single-seat race drew 16,520 votes. The analysis combines certified precinct returns from 2022 and 2024.
Read the research
Civly co-founder Zack Czajkowski joins Ron Steslow to discuss chatbots, voter information, and how campaigns are adapting to AI.
Listen on Apple Podcasts
Seasons now turn on keeping the players you already have. Across 1,231 team-seasons at 129 programs, a coaching change costs about six points of the roster, while doubling the recruiting budget, coach pay, athletic aid or the athletic budget moves nothing. The players hardest to keep are the ones you most wanted to sign.
Every private foundation states on its tax return whether it will read a request it did not invite. Of the 131,545 that answer, 27,218 say yes, about 21%. Nobody publishes that list, so we built it from the Form 990-PF filings: 8,971 foundations by state, with the deadline, the restrictions, the contact and every grant paid.
A Lexis name search matches letters, not people. Across 29 client searches and 33,771 articles, 49% was a duplicate or somebody else entirely. Duplicates run at a steady 31%. Wrong‑person hits swing from 1% to 94%, and the candidate’s name decides which — in one race, 2,997 articles produced 115.
Since 2016, 45 sitting House Democrats have faced a credible challenger measurably to their left, and nine of them lost. Money did not separate the two groups — the challengers were outspent roughly 2.5 to 1 whether they won or lost. Donors did: a median 905 against 258.
About 16 million Americans will use a chatbot to help decide at least one contest in November — and the dominant way they do it is not asking who to vote for. It is photographing the ballot, being declined, and rephrasing until the model answers. Where it lands runs inverse to attention: negligible for Senate, decisive for the judicial and county races that carry no party cue at all.
In a model of contested congressional races from 1980–2024, doubling the number of contributing organizations while holding total dollars fixed changed predicted vote share by +1.11 points for challengers and −0.48 for incumbents. The estimates describe associations, not the effect of recruiting more donors.
Thirteen sports hires where a warning was documented before the contract. Each case records the information available, how long before signing it appeared, and the reported cost or consequence.
Union-backed sides won 33 of 49 ballot measures. Funded legislators sponsored 4.2× as many pro-labor bills as unfunded colleagues, but passage rates were similar. The study also compares floor voting and close-election results.
Civly CTO Matthew Hanauer joins Eric Wilson on the Campaign Trend Podcast to discuss synthetic polling: testing models against certified election results, choosing question formats, and assessing when additional data helps. They also discuss uses in down-ballot races and message testing.
A study of 112 athletes with published contract values of $500K to $6.5M found no detectable relationship between Instagram audience and roster pay after accounting for performance and program scale. The results concern school and collective payments; brand deals are a separate market.
We linked federal athletics records with IRS Form 990s for funders of 54 Division I programs. Booster foundations reported $1.69B in revenue, compared with $98.6M for NIL collectives, a ratio of 17 to 1.
Five models from five labs simulated the same 6,400 voter profiles across eight past elections. The panel average had a typical error of about 1 point across seven races. Disagreement was greatest on the eighth, which all five models misread.
We simulated nine potential Democratic replacements for Graham Platner against Susan Collins using 4,597 registered-voter profiles. Richardson and Bellows led by more than the method's typical error of 4.1 points. The other matchups were within that range.
The Civly Turnout Score classified voter turnout correctly 84% of the time across nine states in the 2024 general election, compared with 78% for a rule based on past participation. In a low-turnout municipal race, accuracy was 83% versus 67%.
Civly published AI-based forecasts for the contested New York and Maryland House primaries before voting began. Nine of ten winner calls matched the certified results. The scorecard includes the method and the missed call.
Explore $78 billion in reported payments to U.S. physicians and teaching hospitals from CMS Open Payments, 2018–2024. Browse by state, congressional district, or county, or search for a doctor by name.
In The Fulcrum, Civly CEO Daniel Barkhuff argues that the DNC's 192-page review of the 2024 election should have addressed AI's role in campaigns and how voters receive information.
Civly's review of the DNC's 192-page post-election report found no mention of artificial intelligence. The article sets out the company's position on AI adoption, verification, and protections for children.
Campaigns & Elections reports on Civly founder Dan Barkhuff's use of AI to compile property records, criminal histories, campaign-finance data, and social media for opposition research.
Daniel Barkhuff announces Civly.ai, a political technology company providing research and campaign tools for Democratic organizations.
The North Carolina startup launches Research Books, Command Center, and Compliance, with candidate research delivered in days.
Contact our team for interviews, press inquiries, or information about Civly.
Contact Us