Did X Community Notes miss 93% of election falsehoods?

A 2026 study found X Community Notes missed 93.1 percent of false midterm claims. Learn why consensus takes time and how to check primary sources yourself.

Fact-checking · Published 2026-10-26 · 7 min read · 1,571 words

AI-generated illustration. A smartphone screen displaying a social media feed with a warning label next to a magnifying glass icon.

When voters encounter viral political claims online, they often look for a crowdsourced correction below the post. However, relying on this system has serious limits. If you are researching community notes election misinformation, a recent analysis shows a massive gap between what goes viral and what actually gets corrected.

According to a NewsGuard Reality Check published on October 21, 2026, the crowdsourced fact-checking program caught only 6.9 percent of posts advancing false midterms claims. The watchdog found that X failed to flag 93.1 percent of false claims related to the US midterm elections. NewsGuard identified dozens of posts advancing false narratives, yet only 13 received a Community Note. The uncorrected posts drew millions of interactions ahead of the November elections. This highlights a structural flaw in relying entirely on user consensus for fast-moving political news.

The consensus bottleneck

To understand why so many false claims slip through, you have to look at how the notes are approved. Community Notes operates on a bridging algorithm. For a note to become public, it requires contributors from opposing political viewpoints to agree that the context is helpful.

A 2025 report by the DDIA highlighted that X pioneered this decentralized moderation model to avoid partisan bias. The system prevents a single political group from mobbing a post and forcing a false correction. However, requiring cross-partisan consensus means highly polarizing topics almost never reach agreement. Because users frequently refuse to rate notes that contradict their political stance, approximately 87 percent of drafted notes are never rated highly enough to be displayed publicly.

The platform is aware that speed is a problem. In October 2024, X introduced an update called lightning notes, re-architecting the scoring system so that notes could appear in under 20 minutes. But technical speed cannot solve human disagreement. If the required opposing groups do not agree on the facts of an election claim, the algorithm keeps the note hidden, regardless of how fast the system processes the votes.

The cost of waiting

When a note does manage to appear, it is highly effective. A 2025 study led by the University of Washington found that appending a Community Note significantly reduced engagement. Following a note, reposts dropped by 46 percent and likes dropped by 44 percent. The visual friction makes people pause before sharing.

But if 93 percent of false election claims never receive that note, the viral spread continues without any friction at all. The uncorrected posts in the NewsGuard study reached millions of users who likely assumed the absence of a warning meant the information was accurate. During an election, a false narrative can do its damage in the hours before opposing political factions finally agree on the wording of a correction.

How to verify claims manually

Since you cannot rely on platforms to flag every falsehood, you must take verification into your own hands. First, treat a Community Note as just one piece of evidence, not an absolute truth. Second, track down the primary source yourself. If a post claims a candidate said something controversial, find the raw video or the official transcript. Do not rely on screenshots, as they are easily fabricated.

Third, check the earliest publication date. Many false election narratives involve recycling old footage and claiming it depicts a current event. Running a reverse image search on a still frame from a video can often reveal that the footage was published years before the event in question. Finally, read across viewpoints. If a claim is entirely absent from major news outlets and primary sources, it requires heavy skepticism.

Automated verification with wyper

wyper builds tools to automate this verification process without acting as an ideological filter. The wyper Fact-Check extension runs directly in your browser. It uses a Dual-AI cross-check featuring Gemini plus Grok. Grok searches for its own data during the check, ensuring the two models do not share a single evidence set. They read the post, check the text, and return a truth score from 1 to 10 alongside an evidence chain of real, linked sources.

Sources are never filtered by origin. There is no mainstream bonus, and official sources are not automatically preferred. The reader decides what to trust. If a post already has a Community Note, the extension reads it as part of the context on the page. It evaluates the claims in the note alongside the original post and checks the primary sources for both.

You do not have to install the extension to test this capability. You can paste a post URL into the free wyper web app to get the same text fact-check and evidence chain. The web app does fact-checking only and requires no signup. Both products have a free tier that allows five text checks per day.

Why independent searching matters

Our own published measurements show why independent searching is critical for fact-checking. Two models trained on overlapping text are correlated, not independent, so confirmation by two AIs carries little information on its own. In our tests with 30 claims using frozen sources, Gemini and Grok agreed in 29 out of 29 cases, even on deliberately contested claims. That is why the standard check forces Grok to run its own search.

Regarding source reach, our tests showed the tool retrieves primary sources 63 percent of the time, legacy media 14 percent, and fact-checkers 1 percent. To ensure accountability, wyper maintains a public dispute register. It is append-only and hash-chained, meaning entries cannot be silently edited or removed. If an evidence chain is contested, the record remains visible for scrutiny.

Handling manipulated video

Election misinformation frequently relies on manipulated or mislabeled video. While the standard text fact-check is free, wyper offers a specialized AI Spotter for images and video on paid plans. It costs 4 credits per video because it calls three paid services on every run.

The AI Spotter is a provenance report, never a simple pixel detector. It never outputs a percentage of how fake something looks, because that number is the harm we exist to prevent. Instead, it extracts still frames from a video and searches each one separately. A hit on any frame is a hit on the video. The provenance report checks for content credentials (C2PA), verifying them against 55 official trust anchors. Only a trusted signature counts, as a merely valid one is self-signable.

Because platforms repackage files on upload, content credentials sitting in their own box beside the picture data are almost always stripped out. This is the normal case on a social feed, not a suspicion. What carries the weight for video verification is the earliest appearance and the event-versus-post date gap. Finding a date gap reveals recycled footage instantly, something a pixel detector cannot see at all.

Who needs which: Manual checks are best for users who want absolute privacy, as no data leaves the browser during a standard search. The wyper web app is ideal for users on mobile devices or shared computers who need a fast, automated evidence chain without installing software. The wyper extension is built for heavy readers who want to verify claims, videos, and existing Community Notes directly in the feed where they encounter them.

FAQ

The short version

Crowdsourced fact-checking struggles with political misinformation because requiring cross-partisan consensus causes severe delays. Instead of waiting for a platform to flag a false claim, readers should verify primary sources themselves or use automated tools to build an evidence chain.

Keep reading

Guides and tools