Facebook users circulated a video supposedly showing the Oct. 1 eruption of Taal Volcano. The video was generated using artificial intelligence.
The clip was uploaded on the same day as the eruption, depicting violent blasts of lava and smoke coming out of the volcano. One FB user published the reel with the following caption:
“Actual Video Taal Volcano
October 1, 2025
ALERT LEVEL 1..”
Details in the video contradict what was reported by the Philippine Institute of Volcanology and Seismology (PHIVOLCS). Expert analysis suggests that the video was generated with Artificial Intelligence.
PHIVOLCS reported a minor phreatomagmatic eruption from the main crater of Taal volcano at 2:02am on Oct. 1. The volcano then erupted twice on Oct. 2: the first for five minutes and the second, for 13 minutes. Footage of the eruptions show lava spewing out in a fountain-like pattern, not aerial explosions such as the ones seen in the AI video.
Alert Level 1 was raised over the area on both days, prohibiting boats and aircraft anywhere near the volcano. The AI video depicts a boat in the water from the vantage point of a helicopter circling the volcano, which would not have been possible.
Aerial footage taken by UNTV on Jan. 15, 2020 shows a similar vantage point with some land and rock formations that are noticeably absent in the AI video which places the volcano in a larger body of water.
Geologist Mahar Lagmay, executive director of the University of the Philippines Resilience Institute, dismissed the video as AI in a Facebook post and advised viewers not to believe it.
VERA Files sent the video for analysis to the Deepfakes Analysis Unit of the India-based Misinformation Combat Alliance.
After running the video through five different AI detection tools, results from two such tests expressed varying levels of confidence that the video was AI-generated.
WasItAI expressed high confidence that screengrabs from the video were created by AI. IsItAI gave an 85% confidence rating that select portions of the video were made with AI, while it ranked other portions of the video as “human created.”
AIorNot gave the video a rating of “likely real,” while Hive Moderation gave a high confidence rating that the video was not generated or manipulated with AI. Google Synth ID detector gave no indication that the video was made with Google AI.
Copies of the video have collectively garnered at least 7,700 reactions, 138 comments, 3,700 shares and 2.2 million views.
