OpenAI Shuts Down Pro-Marcos Campaign from the Philippines

It looks like artificial intelligence is becoming a new tool in the political playbook, but OpenAI isn’t having it. The company behind ChatGPT has recently taken action against accounts in the Philippines that were using its technology to flood social media with political comments.

According to a report released by OpenAI on June 5, some users based in the Philippines misused ChatGPT to post short, politically charged comments on platforms like TikTok and Facebook. These comments generally supported President Ferdinand Marcos Jr. and criticized Vice President Sara Duterte. The campaign, now referred to as “Operation High Five,” stood out because of its heavy use of emojis and cheerful tone, despite its political intent.

The operation was traced to Comm&Sense Inc., a PR firm located in Makati. Though contacted by the media, the company has yet to comment on the issue.

OpenAI described the activity as a low-level influence operation. Still, it involved a well-structured process. First, the group used ChatGPT to scan social media for trending political content. Then, based on what they found, the model was prompted to generate hundreds of brief replies, typically no longer than 10 words, aligned with certain themes. These were not just for fun or casual engagement. Some comments were used to make videos look more popular, especially on TikTok, by posting repetitive, positive responses.

In addition, ChatGPT was also used to create PR materials and data analyses, likely as part of a pitch to attract future clients. One detail revealed in the report was the creation of five TikTok accounts used to promote content favorable to President Marcos. These accounts began sharing videos around mid-February, just as the 2025 election campaign season kicked off in the Philippines.

However, while the operation was organized, its reach appeared limited. Most of the fake accounts had little to no engagement. On Facebook, the pages had fewer than 15 friends, and the comments often didn’t get more than a handful of likes or replies, if any. Even when the campaign tried to flood comment sections of news pages like ABS-CBN, it barely made a dent. For example, hundreds of AI-generated comments were buried in a sea of over 23,000 real ones.

OpenAI's investigation shows how AI tools can be misused to shape public opinion, even subtly. While the actual influence was minimal this time, it raises serious questions about the future of online political discourse and the role of AI in it.

Photo by Levart_Photographer on Unsplash