How to fact check campaign ad claims?

Learn how to verify political ads about taxes and Medicare using primary documents like CBO estimates and legislative texts before trusting campaign claims.

Fact-checking · Published 2026-11-01 · 10 min read · 2,111 words

AI-generated illustration. A laptop screen displaying a government budget document next to a paused video player showing a political advertisement.

Every election season brings a flood of political advertising to our screens. When you need to fact check campaign ad claims, relying on the opposing candidate for clarity is a mistake. The opposing campaign has its own narrative to push. The most reliable method is to ignore the partisan messaging entirely and go straight to the primary documents. Whether an ad claims a candidate is gutting Medicare or raising taxes on working families, the truth is usually documented in official legislative records. By focusing on nonpartisan budget estimates and actual roll call votes, you can strip away the emotional voiceovers and evaluate the bare mechanics of the policy.

Start with the primary source

A primary source is the original, contemporaneous material that documents an event or a policy decision. For political claims about taxes or federal programs, this means the actual text of a bill, a roll call vote record, or an official cost estimate. Secondary sources, like press releases or campaign websites, exist to persuade. Primary sources exist to record.

Campaign ads often rely on emotional framing to obscure the mundane reality of legislative work. They might take a single procedural vote on a massive omnibus bill and frame it as a targeted attack on a specific demographic. To break this tactic down, you must first isolate the factual assertion. If a political action committee runs an ad saying a candidate voted to raise taxes, preserve the full ad and identify the specific piece of legislation it references. The disclaimer at the bottom of the ad tells you who paid for the airtime. It does not prove the claim is true. You must trace the claim back to the floor of the legislature.

When verifying local claims, public records are the authoritative baseline for campaign finance facts. You should corroborate these filings with local reporting to understand the place-based context, but always keep the exact document names and access dates so your verification work remains reliable.

Reading CBO scores for tax and Medicare claims

When an ad makes a specific numerical claim about how much a bill will cost or how deeply it will cut a program, the definitive source is usually the Congressional Budget Office. Established in 1974 to provide Congress with expert staff, the CBO produces nonpartisan analysis on federal economic and budgetary matters.

The agency produces about 1,000 cost estimates a year, giving lawmakers objective data on how a bill will impact the national deficit over a ten-year window. When you review a CBO estimate, you are looking at how legislation would change spending and revenue relative to a baseline of what would occur under current law. It is not an abstract judgment of whether a number is too large. It is a strict mathematical projection of a proposed policy.

News coverage often focuses on the political intent of a bill, but the CBO document answers who pays for it and exactly how much they pay. If an ad claims a new law drains Medicare funding, the CBO score for that specific law will show the exact projected impact on Medicare outlays. By comparing the dramatic claims in a thirty-second commercial to the dry, nonpartisan CBO data tables, you can immediately spot exaggerations and missing context.

How context changes the truth of a vote

A common tactic in campaign advertising is citing a real vote but leaving out the context of what the bill actually did. FactCheck.org frequently documents this method in action across the country.

In one recent example, Democratic ads exaggerated the impact of a major domestic policy law on Medicare. The ads accused opponents of gutting the program and cheering on an attack on seniors. The primary documents showed that while the law did make changes to Medicare, those changes were narrow in scope and affected relatively few beneficiaries. The broader cuts the ads implied were tied to a separate 2010 deficit law that required automatic spending reductions. Congress subsequently waived those cuts, just as lawmakers of both parties have done regularly in the past.

Another frequent distortion involves combining unrelated votes into a single attack. An ad in Alaska claimed a senator voted to raise healthcare costs by $1,800 while voting to give himself a taxpayer payout on the exact same day. The primary record showed the senator actually voted in favor of a compromise bill to end a government shutdown. The bill simply did not include an extension of enhanced federal healthcare subsidies. Framing a vote to reopen the government as a deliberate, targeted vote to raise healthcare costs is a classic omission of vital context.

Tax claims follow the exact same pattern. In Maine, ads claimed a candidate wanted to raise Social Security taxes and hurt seniors. The actual legislative proposal sought to remove the cap on the maximum amount of income subject to payroll taxes. This change would only affect high-income earners who were still working, not the seniors the ad claimed would be hurt. The same ads cited state legislature votes to claim the candidate increased taxes on gasoline and groceries, but the actual votes were for fees that did not directly pertain to those specific items.

Verifying claims efficiently with AI tools

Doing this research manually requires locating the bill number, finding the CBO score, and reading through legislative history. This is the most thorough method, but it takes significant time. You can also use tools designed to trace claims back to their roots and fetch the relevant sources for you.

The free wyper web app allows you to paste a claim or a URL and receive a truth score from 1 to 10, backed by an evidence chain of real, linked sources. This standard check runs a Dual-AI cross-check using Gemini and Grok. Because Grok searches for its own sources, the two models do not share a single evidence set. You get five free text checks a day without needing an account. The truth score is a useful metric, but you should never call it proof of trustworthiness. The tool does not judge intent. Both of those responsibilities stay entirely with the reader.

For a deeper look, the wyper extension Pro tier offers a Full Spectrum check. This is a Triple-AI reading: three models, one evidence set. The third voice does not search. It receives exactly the sources the main model received and reads them independently. A disagreement here is a different reading of the same legislative material, not a different search result. Two models trained on overlapping text are correlated, not independent. In our tests of two models agreeing on 30 claims with frozen sources, Gemini and Grok agreed in 29 of 29 cases, even on deliberately contested claims. A third model of different origin disagreed in 3 of 26 cases, every time with a machine-verified quote. This is the strongest thing we have, and it argues against a claim we used to make ourselves about two AI models being enough to guarantee accuracy.

Our published measurements show this approach works. In a test of 633 real citations re-fetched, we found 73 percent verbatim on the page. Out of 24 claims with the root named in advance, the tool reached 63 percent primary sources and 14 percent legacy media, reaching all 6 non-English roots. Sources are never filtered by origin. There is no mainstream bonus, and official sources are not automatically preferred by the algorithm. The reader decides what carries weight. If a user finds a flaw in a check, there is a public dispute register. It is append-only and hash-chained, so entries cannot be silently edited or removed.

If the ad is a video, verifying it gets harder. The wyper extension can check videos up to 10 minutes on the free tier by reading the subtitles. For YouTube videos on the Pro tier, it can handle up to 30 minutes. If subtitles are unreachable, an AI model actually watches the video within that time window. For X videos, the limit is 2 minutes on the free tier and 7 minutes on Pro. Longer videos are refused outright, not partially checked. For videos on X, both the web app and the extension take the same route, so the extension has no specific advantage there. The web app cannot see your YouTube session, however, so subtitles are often unavailable there.

If you are worried about manipulated media, the AI Spotter is available on paid plans for both images and video. It generates a provenance report, never a detector percentage. A percentage of how fake something looks is the exact harm we exist to prevent. Instead, the AI Spotter verifies content credentials against 55 official trust anchors. For video, it takes still frames and searches each one separately. A hit on any frame is a hit on the video. Because a content credential sits in its own box beside the picture data, any repackaging drops it. Platforms repackage files on upload, so a video taken off a feed will almost always return nothing on that layer. That is the normal case, not a suspicion. The report will advise checking the original file instead. What carries the weight for video is earliest appearance plus the event versus post date gap, which catches old footage relabelled as new.

Who needs which: Voters who want maximum privacy and have the time to read through congressional budget reports should stick to manual checks. Casual readers who want to quickly verify a text claim they saw in a news article without installing anything should use the wyper web app. Researchers and heavy social media users who need to check videos in their feed or run Full Spectrum checks with a Triple-AI reading should install the wyper extension.

The short version

Verifying campaign claims requires looking past the partisan framing and reading the primary documents, such as CBO scores and legislative texts. By comparing the emotional language of an ad to the nonpartisan financial estimates, you can quickly identify missing context and exaggerated figures. Whether you do this manually or use a tool to fetch the evidence chain, the original source remains the ultimate authority.

Keep reading

Guides and tools