Analysis frame
Reported evidence
How a platform policy that distinguishes internal campaign work from voter persuasion depends on identity, intent detection, disclosure, and consistent enforcement.
- voters receiving campaign messages
- political campaigns and consultants
- election officials and disclosure regulators
- AI providers enforcing political-use policies
- How many campaign subscriptions were used for voter-facing persuasion rather than permitted internal work
- How often prohibited prompts succeed across models, accounts, languages, and election periods
- Which enforcement layers operate beyond visible model refusals
- Campaigns may shift risky AI work to outside groups or vendors with weaker disclosure
- Uneven refusals could advantage users who learn to reframe prohibited requests
- Low public trust in synthetic messaging may push campaigns toward less visible AI assistance
Campaign use is visible but incompletely disclosed
A Washington Post analysis found that 39 congressional candidates reported paying for OpenAI subscriptions during the 2026 election cycle. Two explicitly described using the subscriptions for advertising, while another campaign disclosed using unspecified AI software to draft and personalize political messages or create synthetic media. Around 30 political action committees and parties also reported OpenAI payments.
The filings show that AI has entered campaign operations, but they do not reveal the full scope or purpose of every payment. Consultants told the Post that use is likely broader than disclosures suggest, especially for email, ad copy, scripts, research, planning, and internal productivity. Those claims describe industry practice; they are not a complete census.
The policy depends on a difficult intent test
OpenAI allows responsible, human-directed campaign work such as research, planning, administration, and budgeting, but prohibits targeted political persuasion and the use of its tools to generate campaign ads. The Post found inconsistent enforcement: in some tests ChatGPT drafted fundraising emails or demographic-targeted messages, while later versions of similar prompts were refused.
OpenAI says refusals are only one part of its enforcement and that its policies apply to everyone. The practical gap remains. A reliable system would need to identify campaign actors, distinguish internal work from voter persuasion, document enforcement, and support legitimate uses without monitoring all political speech. Until those controls are independently testable, a written ban is a commitment whose real coverage is unknown.
Go to the source
Read the evidence behind this analysis. External links open in a new tab.
The Washington Post — ChatGPT bans campaigns from using AI to make ads, but campaigns are using it
