In mid-August, we came across an op-ed written by a UC Berkeley mathematics professor advocating for the reintroduction of standardized testing in UC admissions. Within the argument, we recognized a pattern.

The editorial style was formulaic, with syntax and sentence structure commonly found in text produced or edited by artificial intelligence.

We ran the article through Pangram, an AI text detection platform, which identified the writing as — at least partially — AI-generated. The Daily Californian reached out to the author, who confirmed that AI helped produce the op-ed.

Since its publication, multiple national media outlets have circulated the story, and it has spurred a campus debate over AI writing tools, particularly in academia and public communications.

In the following weeks, the Daily Cal expanded the scope of our original inquiry. Staff from our news and data departments ran AI-detection technology on several genres of material, representing more than 1,000 independent published works and 1 million scanned words. We conducted this research to map how far AI-written text has permeated common correspondence and understand how institutions are approaching modern communications.

We narrowed our search to three key areas of public interest:

Communications of local leadership, including UC Berkeley Chancellor Rich Lyons, Berkeley Mayor Adena Ishii, UC President James Milliken, congressional representatives and other lawmakers.Legislation, including resolutions passed by the ASUC, bills approved or prompted by state officials and Berkeley City Council agendas.Public announcements, including obituaries, newsletters, press releases and student informational content.

AI writing was detected in all domains, across almost all genres and levels of public correspondence. Pangram detected AI use in obituaries, Substack posts, messages to students, recorded remarks, news articles and more.

This apparent widespread adoption of AI writing tools is, according to our research, a relatively recent phenomenon. The vast majority of positive test results found in text were published in the past year. For UC Berkeley student government, 42.1% of resolutions passed in the 2025-26 were flagged for AI.

Our reporting on this topic heavily relied upon the use of Pangram, our program of choice for AI writing detection. Pangram is seen as the industry leader in accuracy for AI text detection platforms, with independent reviews of the technology placing its accuracy rate above 99%. A study conducted by researchers at the University of Chicago in August 2025 found Pangram’s false positive rate at “near zero” — less than 0.5%. A recent study conducted by researchers at the Vrije Universiteit Brussel on student academic papers returned zero false positives and Pangram itself claims a false positive rate of 0.0041% on its most recent model.

Nevertheless, Pangram is imperfect and can incorrectly flag human writing as AI. When using the technology on a portion of text, Pangram provides an AI percentage score and highlights sections where it believes AI was used. The percentage score is variable, and in some cases, unreliable, which is why our reporting presents detection results primarily in aggregate. In cases of positive detection, we reached out to the text’s authors or affiliated public officials to provide them an opportunity to explain their approach to communications and ask if they had worked on text using AI.

Respondents claimed a wide variety of AI writing use cases, with a common trend arising in their assertions: descriptions of an emergent need for modern tools that might make some tasks more efficient or accurate. Among the offices of local leadership, including Mayor Ishii and Chancellor Lyons, few have adopted institutional policy governing the use of AI writing tools. Large institutions like UC Berkeley and the city claim to have entered an exploratory period with AI.

Our coverage on this issue has been separated into five reports covering how AI writing has been adopted in public-facing communications: UC Berkeley and UC administration, the city of Berkeley, the ASUC, obituaries and community reactions.

— Aarya Mukherjee, Managing Editor 

ASUC senators and officials have put forward nearly a hundred resolutions flagged for AI-generated or altered writing by the AI detector, Pangram.

— Madeleine Kashkooli

Last year, a faculty member in the theater, dance, and performance studies department passed away after a nine-year tenure at UC Berkeley. A week after his death, the department posted a 300-word obituary on its news site, cataloging his contributions to the campus community. That obituary and two others published by UC Berkeley departments may have been written or altered by AI.

— Elise Fisher

Various public communications from campus, including from known figures UC Berkeley Chancellor Rich Lyons and Vice Provost for Undergraduate Education Oliver O’Reilly, were flagged by AI detector Pangram as being generated or altered by an AI writing tool.

— Alice Conry & Cam Lippincott

On Aug. 24, Mayor Adena Ishii posted a 90-second video to her city Instagram account addressing public concerns regarding the Eighth and Harrison encampment. The video, or rather the script used for her address, was flagged as fully AI-generated by AI detection platform Pangram.

— Alice Conry & Paarth Mishra

The Daily Californian sat down with UC Berkeley Chancellor Rich Lyons on Aug. 19, in an expansive interview to discuss the most pressing issues for campus and university students.

— Aarya Mukherjee

This semester, for the first time, media studies lecturer Rich Jaroslovsky’s class will have an AI policy. Jaroslovsky will not allow students to write papers with AI, however he won’t bar them from using it. He will require students to disclose how they used AI in their assignments, including what AI models they used and what they used them for.

— Jackson Woodward