Bluesky will take down threats against election workers, false information about when, where or how to vote, deepfake impersonation, doxxing and fake account networks during the 2026 US midterm elections, while labeling unverified claims and premature declarations of victory, according to a post the company published on its blog on October 19, 2026. The post, signed by The Bluesky Team, also sets out an internal election command center, specialist review for reports involving candidates and officials, and proactive sweeps for impersonation and coordinated activity ahead of voting on November 3.

In Short

Bluesky has written down what it will delete, what it will tag with a warning label and what it will leave alone when people post about the US midterm elections on November 3. Posts that lie about how to vote, threaten election workers or pretend to be a candidate will be removed, while posts claiming a win before votes are counted will get a label and be removed if official results prove them wrong. If you use Bluesky, criticism, predictions, debate about election rules and satire stay up, and you can choose whether labeled posts are shown, hidden or placed behind a warning.

Three categories, one rulebook

The post is built around a three-way split. Some content is removed, some is labeled and some is explicitly left alone. That structure is the most specific part of the document, and it is worth reading in full because the boundaries between categories decide what millions of election-related posts will look like over the next two weeks.

According to Bluesky, the company will remove:

  • Threats or intimidation directed at voters, poll workers, election officials or candidates
  • False information about when, where or how to vote
  • Impersonation, including deepfakes
  • Doxxing
  • Coordinated harassment
  • Fake account networks and influence operations

The labeled category is narrower. It covers two things: "Unverified claims stated as fact, such as a report that a voting machine has failed," and "Premature victory claims while votes are still being counted." The second carries a conditional escalation. "If official results show that a claim is false, we remove it," according to the post.

The third category spells out what Bluesky will not touch. It lists criticism of candidates, parties and officials, "provided it does not violate our Community Guidelines, including our rules on harassment and discrimination"; opinion, predictions and "general statements about the validity of the election"; debate about maps, laws and election procedures; and satire.

That last group draws a line that other platforms have struggled to hold. A general statement that an election is illegitimate stays up. A specific, unverified factual claim, such as a named machine failing, gets a label. A false statement about voting logistics, such as the wrong polling date, comes down. The distinction is between opinion about the outcome, assertions of fact that cannot yet be checked, and misinformation about the mechanics of voting itself.

Users decide how labels appear

Labels on Bluesky do not work the way warning interstitials do on most large platforms. "You control how posts labeled as unverified claims appear to you. In your moderation settings, you can choose to show, hide, or place a warning on these posts," the post states.

That is consistent with the architecture Bluesky has described since early 2024. When the company opened its moderation stack to outside developers in March 2024, it presented Ozone, an open-source tool, as a way for independent moderation services to label content and for users to subscribe to those services alongside Bluesky's own default moderation team. The election post does not mention third-party labelers, and it is not clear from the text whether the new election labels will be applied only by Bluesky's own service.

Two commitments, applied to every party

Bluesky frames the program around two commitments, which it says "follow from our Community Guidelines." The first: "We act quickly on content that could cause harm." The second: "We protect people's right to talk about the midterms."

The company argues that the two must be held together. "We hold these commitments together because neither is sufficient alone. These rules apply in the same way to every party and every candidate," according to the post.

The language of neutrality appears more than once. It recurs in the section on specialist review, where Bluesky writes that reports made by and against candidates and elected officials "will receive specialist review, and the same standard will apply regardless of party." The repetition is not incidental. Content moderation during US elections has become a partisan flashpoint, and Meta, the largest social platform operator, ended its third-party fact-checking program in the United States in April 2025 in favor of user-written notes. Bluesky's post makes no reference to fact-checkers or to community notes. Its model rests instead on removal, labeling and the user's own display settings.

Only the US midterms

The scope is explicit. "This effort covers the 2026 US midterm elections. Elections elsewhere are governed by the same Community Guidelines but are not part of this dedicated program," according to the post. The rules themselves are global. The command center, specialist review and proactive sweeps described below are not.

How Bluesky says it is preparing

The preparation section lists six measures. Bluesky describes the effort as "a company-wide responsibility."

  • Election command center. "A cross-functional group will coordinate our election work across teams."
  • Shared moderator guidance. Moderators will "work from shared guidance on election-related policies so that decisions are fast and consistent."
  • Specialist review for candidates and officials. Reports made by and against candidates and elected officials go to specialist reviewers.
  • Proactive impersonation sweeps. "We will search for accounts impersonating candidates, campaigns, and election offices rather than waiting for reports."
  • Coordinated-behavior scans. Bluesky says it will "proactively look for coordinated inauthentic activity, including bot networks and influence operations."
  • Verification and account protection. The company says it is "engaging with campaigns and election officials through a dedicated process so that verification requests and account issues are handled consistently."

The post gives no staffing figures, no budget and no detail on how large the command center is or when it begins operating. It names no external partners for detection work. It does not say what tools the impersonation sweeps use, how often they run or whether they extend to accounts impersonating journalists or news outlets, which are not on the list.

Nor does the post set any time window. Meta has published a fixed restriction period for new political ads running from October 27 to November 3. Bluesky's commitments carry no start or end date beyond the reference to the November 3 vote and to "the days ahead."

The post does not mention advertising or paid political content at all. Its subject is organic posting, accounts and moderation.

The 18 months of systems behind it

A separate section, "Building on work already underway," lists systems that Bluesky says it developed "over the past year and a half." Several contain figures not previously disclosed in the same place.

3,619 accounts and influence operations

"In 2025, we removed 3,619 accounts for suspected involvement in influence operations, the majority linked to foreign state-aligned actors," according to Bluesky. The post points to the company's 2025 Transparency Report, linked at an address dated January 29, 2026, which it says describes a detection approach that "relies primarily on behavioral patterns and technical indicators rather than on content alone."

The post does not break the 3,619 figure down by country or by operation, and it does not define "majority." For scale, Google reported terminating 18,532 YouTube channels in a single quarter, July to September 2025, under its influence operations program. The platforms differ enormously in size, so the comparison says more about volume than about effectiveness.

Outside testing has questioned how well platforms, Bluesky included, remove inauthentic activity. A NATO Strategic Communications Centre of Excellence experiment that bought fake engagement across seven platforms between September and November 2025 found that Bluesky removed 55% of the fake accounts but none of the purchased engagement, and removed 23% of accounts the researchers reported. That test predates the election program, and the October 19 post does not address it. The fake account networks Bluesky now promises to remove are the same kind of sockpuppet infrastructure that study purchased, and the coordinated-behavior scans are aimed at the astroturfing such networks make possible.

From 6 reporting categories to 39

"We expanded post reporting options from 6 categories to 39, which routes reports to the right specialist team faster," according to the post. The change is described in a moderation process update that the post links at an address dated November 19, 2025. The jump matters for election reporting specifically: the post later tells users that reports of impersonation, misleading content, harassment and threats "reach specialist reviewers," which depends on the report landing in the right category in the first place.

A strike system scaled to severity

Bluesky describes a strike system that "assigns a severity level to each policy violation so that consequences scale with the potential harm, from warnings to suspensions to permanent bans." The post adds: "Critical violations, including those that threaten immediate real-world harm, result in permanent removal." It does not say how many strikes trigger each level, or which election violations count as critical.

Verification and domain handles

The post cites "Bluesky's verification badges, together with our network of Trusted Verifiers (independent organizations that verify accounts within their own areas of expertise) and domain-based handles." Bluesky introduced blue checks and Trusted Verifiers on April 21, 2025, building on domain-based handles that launched in April 2023 and that more than 270,000 accounts had adopted by May 2025. In the election context, verification serves a narrow purpose: helping users tell a real election office or campaign from an impersonator. Bluesky's advice to users later in the post is to look for "a verification badge or a domain handle if you are unsure whether an account is relevant."

Impersonation through manipulated media is covered by the removal list, which names "impersonation, including deepfakes." A deepfake of a candidate is therefore treated as impersonation rather than as a separate category of synthetic media. The post does not describe any detection technology for manipulated video or audio, and it says nothing about labeling AI-generated content that does not impersonate anyone.

14,000 submissions behind the guidelines

The current Community Guidelines "were shaped by more than 14,000 community submissions and include protections for journalism, education, and advocacy," according to Bluesky. Those guidelines were published for feedback in August 2025 and took effect on October 15, 2025, organized around four principles and including an explicit prohibition on deepfakes aimed at exploiting individuals. The submission count did not appear in that earlier announcement. The post also links an essay titled "Building Healthier Social Media," at an address dated September 19, 2025, which it says explains the thinking behind the guidelines.

Appeals: human review, and an inbox still to come

"Enforcement actions can be appealed, and every appeal receives human review," according to the post. The route depends on the action. Appeals of account suspensions are made inside the Bluesky app. Appeals of post takedowns go by email to moderation@blueskyweb.xyz. "Soon you will also be able to submit appeals through a new moderation inbox," the company writes, without a date.

That split matters for timing. A post taken down for a false voting claim in the last days before November 3, or a premature victory claim removed after results are certified, would be appealed by email. The post gives no target response time for appeals, during the election period or otherwise.

What Bluesky asks of users

The final substantive section addresses users directly with three requests. The first is to report content rather than amplify it. "Reporting is more effective than quote posting, which can spread harmful content further," the post states, naming "Misleading," "Harassment or hate" and "Violence" as examples of report categories.

The second is to "Confirm claims about deadlines, polling places, and results with your state or local election office or another official source before passing them on."

The third points to Democracy Works, which Bluesky describes as "a nonpartisan nonprofit that helps voters find accurate information about when, where, and how to vote," and to its TurboVote service. "Early voting is already underway in many places," according to the post. The TurboVote link in the post carries a Bluesky referral tag and blog campaign parameters.

Why this matters for the marketing community

Bluesky is a small advertiser destination but a large conversation venue. The company reported more than 40 million registered users when Toni Schneider became interim CEO in March 2026, and the post itself says that "a significant share" of election conversation takes place there. For brands, agencies and publishers that distribute content on the network, the practical consequence is a clearer map of what is likely to be removed or labeled near accounts they run or monitor.

The timing places Bluesky alongside a wave of election-period rules from other platforms. Meta will block new US political ads for eight days before the vote, from October 27 through November 3. Flipboard pinned seven human-curated midterm news collections to its home view in early October. And the money flowing into the cycle is large: AdImpact projected $11.6 billion in political ad spending for the 2026 midterms, about $400 million above the 2024 presidential cycle.

Meta's measures act on paid placements. Bluesky's act on organic posts and accounts. Together they show how differently platforms define their election obligations: one through the ad auction, the other through moderation queues and label settings. For campaigns and advocacy groups that rely on organic reach on newer networks, the Bluesky rules carry particular weight, because the impersonation sweeps and verification process directly affect how official accounts are found and trusted.

Industry commitments of this kind have a history. In February 2024, 20 technology companies signed an accord to combat deceptive AI in that year's elections. Bluesky was not among the companies PPC Land named at the time. Its own program, two years later, is narrower in technology terms, saying little about AI detection, but more specific about where it draws lines on ordinary political speech.

Several questions remain open after reading the post. It does not say how quickly removals will happen, how many moderators are assigned to election work, whether third-party labelers will apply election labels, or when the program ends. Bluesky closes by promising to "continue to be transparent about how we apply our policies" and to "keep refining our approach in the days ahead." Whether that includes publishing election-specific enforcement figures after November 3 is not stated.

Timeline

Summary

Who: Bluesky, the social network built on the AT Protocol, writing as The Bluesky Team. The rules affect voters, candidates, campaigns, election officials and anyone posting about the US midterms on the platform.

What: A dedicated program for the 2026 US midterms that removes threats, false voting logistics, impersonation including deepfakes, doxxing, coordinated harassment and fake account networks; labels unverified claims and premature victory claims, removing the latter if official results disprove them; and leaves criticism, opinion, procedural debate and satire up. It adds an election command center, shared moderator guidance, specialist review for candidates and officials, proactive impersonation and coordinated-behavior sweeps, and a dedicated verification process for campaigns and election officials.

When: The post was published on October 19, 2026. Voting takes place on November 3, 2026, and early voting was already underway in many places, according to Bluesky.

Where: The United States only. Elections elsewhere remain under the same Community Guidelines but outside the dedicated program.

Why: Bluesky says it wants people to "speak freely about the election" while keeping the conversation "free of content that harms the electoral process," and that neither goal is sufficient on its own.