QUID · GEORGIA U.S. SENATE · OCT 3 - OCT 9, 2026
Seven days of conversation about Jon Ossoff and Mike Collins, split between posts that can be placed in Georgia and posts from the rest of the country. Six days were ordinary. The seventh was a debate that produced more conversation than the other six combined.
Published polling, shown to set context. A poll asks a sample of voters which candidate they intend to support. Everything else on this page measures something different: the balance of positive and negative labeled posts about each candidate in the captured conversation. A candidate can lead the vote and still take the worse of the argument online, and where the two diverge this page explains the conversation, not the electorate.
The two measures agree, narrowly. Ossoff leads the polling average by about 8 points and takes the better of the conversation inside Georgia, though both candidates sit within three points of zero there. The figure to watch is Collins's: on posts naming him and not Ossoff his national net falls from +12.2 to −52.5.
Net sentiment runs on the left axis from −100 to +100. Gold bars carry total daily volume for both candidates combined; hover any bar for the narratives driving that day. Six flat days and a cliff. Friday carries 133K posts against a weekly average of under 10K for every other day.
Georgia's two in-state cells are almost identical: both candidates sit within three points of zero, on roughly a hundred stored documents each. The state is split down the middle.
Georgia
1,833 unique authors · 270 original posts
(6% of volume; the rest are reposts and replies).
Sentiment from 1,090 labeled posts, 24.2% of this slice;
stored sample behind the labels 109 documents.
Inside Georgia the case for Collins is built on immigration and on blaming Ossoff for a government shutdown that delayed farm aid. It is the most locally specific argument either side makes all week.
Jon Ossoff delayed aid to Georgia's farmers by shutting down the government to defund ICE and Border Patrol.
@TeamOverhaulGAOct 9Source
The in-state reaction tracks the national one closely, including from accounts that had expected Collins to perform better. His worst daily figure of the week, -75.6, comes on debate eve rather than on debate night itself.
A few months ago I told one of my friends Mike Collins is too stupid to realize he's the sacrificial lamb, and boy did tonight prove me right.
@0liviajuliannaOct 9Source
Collins's in-state cells are the thinnest in this race: 55 stored positive documents and 54 negative. The resulting +0.9 is best read as a state that has not moved rather than as a precise balance.
I hate you, Mike Collins.
@KevinMOliverATLOct 4Source
Georgia
2,269 unique authors · 400 original posts
(7% of volume; the rest are reposts and replies).
Sentiment from 1,010 labeled posts, 16.8% of this slice;
stored sample behind the labels 101 documents.
Ossoff's in-state cell is the quietest of the four relative to his national presence: 6,000 posts against 83K outside the state, and only 17 percent carry a sentiment label. Georgia is watching a race the country is arguing about.
My 'Vote your Ossoff' tote just arrived and I love it.
@GeorgiaVoter2026Oct 9Source
Where Georgians criticize Ossoff directly rather than through the debate, the subject is advertising and health care costs, and the charge is that his ads misattribute blame.
You are running deceptive ads. Health care costs are skyrocketing, not by any actions of Collins.
@ArmyMom224Oct 6Source
Both candidates finished Friday within a fifth of a point of each other, +29.5 for Collins and +29.3 for Ossoff. A debate that national accounts scored decisively produced no separation at all in the daily figures.
Ossoff was brilliant. This man is absolutely direct, informative and his presence is calm.
@Jimmy10127925Oct 9Source
The same week reads differently depending on where you stand. Net sentiment is plotted from Georgia on the left to the rest of the country on the right; the solid line counts every post naming a candidate, the dashed line counts only posts that name that candidate and not the other.
All posts naming the candidate Posts naming that candidate only
Collins's +12.2 national figure is the most borrowed number on this microsite. 81 percent of his posts also name Ossoff, and on posts naming him alone he reads -52.5. The national conversation is not positive about Collins; it is large, it is about the debate, and he is in it.
Ossoff improves on every tightening. He is +11.3 nationally and +30.2 once his opponent's name is removed, and the in-state and national readings agree in direction. Among the twelve races measured, he is one of the few whose figures get better the harder the test.
Geography, not party. Both columns cover the same week and the same two candidates; what separates them is which arguments actually travel.
Georgia argues about consequences: a shutdown that delayed aid to farmers, immigration enforcement, and whose advertising is honest about health care prices. The in-state cells are small and almost evenly split.
Outside Georgia the race barely registered until Thursday night, then produced 133K posts in a day. The subject was Collins's ties to white nationalism and Ossoff's handling of it, argued on Bluesky and X rather than in the press.
Both men read positive nationally, which is an artifact. Remove the posts naming the other candidate and Collins falls 65 points while Ossoff rises.
Rest of United States
30,297 unique authors · 2,961 original posts
(5% of volume; the rest are reposts and replies).
Sentiment from 15,124 labeled posts, 26.7% of this slice;
stored sample behind the labels 2,589 documents.
Ossoff used the debate to raise Collins's documented ties to white nationalism, and that exchange is the single most repeated item in the entire corpus. It is what national coverage led on, and it is why Collins's figure collapses once his opponent's name is removed.
Mike Collins has brought dishonor to our state through his personal bigotry and his widely documented ties to white nationalists.
@AcynOct 8Source
Collins's own response was to turn the question back on Ossoff over a donor he declined to condemn. It is the one line of his that travelled at scale, and it is carried almost entirely by his own account and party research accounts.
This is the megadonor that Jon Ossoff refused to condemn tonight. Coward.
@MikeCollinsGAOct 9Source
The clearest case of classifier misattribution on the microsite. Posts celebrating Ossoff beating Collins are scored as positive toward Collins because they name him. With 81 percent of his posts naming both candidates, this is systematic rather than occasional.
Knockout punch. Sen. Jon Ossoff wipes the floor with his MAGA opponent during their debate with one of the most brutal soundbites of the cycle.
@OccupyDemocratsOct 9Source
Rest of United States
43,294 unique authors · 4,388 original posts
(5% of volume; the rest are reposts and replies).
Sentiment from 13,272 labeled posts, 16.1% of this slice;
stored sample behind the labels 2,855 documents.
Ossoff's national cell is the largest evidence base in the race at 2,855 stored documents, and most of it is reaction to one night. The praise is specifically about technique rather than about policy.
Jon Ossoff was the clear winner of Thursday night's debate against Georgia GOP nominee Mike Collins.
@JeffTrnkaOct 9Source
The organized national case against Ossoff is about the border, pushed through party research accounts in clipped video form. It is the mirror of the Collins material and reaches a comparable audience.
Mike Collins slams Jon Ossoff for helping Biden open our borders and putting Georgians at risk.
@RNCResearchOct 9Source
A recurring strand of the positive commentary is impatience with Democrats generally, using Ossoff as the counter-example. The classifier scores these as negative because of their tone, though their subject is approval.
Democrats have made a strategy out of 'have you, at last, no sense of decency' for too long. Ossoff is finally not doing that.
@Cassandra MillerOct 9Source
This page is the product of two different kinds of work. One half is data that no public API makes available; the other half is judgment about what the data does and does not support.
None of the following could be assembled from public endpoints inside a week.
The data does not interpret itself, and it arrives with traps in it.
Neither half can establish cause. This page measures what was said, how much of it there was, and who said it. It is not a poll, it does not model turnout, and it does not predict the November result.
primary: sentiment targeting, so the sentiment is scored
toward that named figure rather than toward the post as a whole.