Lisa Braswick

From content integrity to the synthetic-media frontier

I work at the intersection of language, machine learning, and trust. For the better part of a decade I've been preoccupied with a single deceptively simple question: how do we know whether a piece of content is what it claims to be? What began as a fascination with plagiarism detection and editorial standards has, over six years, become a full-time focus on AI-generated text, synthetic audio and video, and the systems we build to tell the genuine from the generated.

I didn't come to this work through a lab. I came through newsrooms and integrity offices. I spent my early career inside editorial and academic-integrity teams, where I watched detection tools arrive with enormous confidence and very little transparency. Institutions were being asked to make high-stakes decisions - about a student's degree, a journalist's credibility, a job applicant's honesty - on the strength of a probability score that almost no one understood. That gap between confidence and comprehension is the problem I've built my work around.

Most of my time goes to testing. I run detectors, humanizers, and provenance tools through controlled benchmarks and publish the methodology alongside the numbers. The rest splits between explaining the underlying mechanics to decision-makers and helping universities, newsrooms, and security teams design workflows that account for what these tools actually can and can't do.

I'm independent on purpose. I take no equity in the vendors I review, I disclose every affiliate relationship, and I publish corrections in the open. The field is moving fast enough that overstated certainty is its own kind of harm - a false positive on a student essay or a missed deepfake in a newsroom carries real consequences. My commitment is to evidence over hype, and to showing my work so others can check it.

When I'm not testing tools, I'm usually writing about the arms race between detection and evasion, speaking with journalists and educators about practical policy, or digging into the next provenance standard before it reaches the mainstream. The organizations that will handle the next few years well are the ones building AI authenticity as a capability, not just buying a product.

How I approach the work

Every claim on this site is the product of hands-on testing, primary sources, or direct consulting experience - never a surface-level rewrite of someone else's reporting. If a piece of content could be written by anyone with a search engine and thirty minutes, it doesn't belong here. My full process is documented in the editorial standards.

Areas of focus

• AI text detection - accuracy benchmarking, false-positive analysis, and the linguistics of why detectors fail.
• Synthetic media - audio, video, and image forensics, including real-time detection in live settings.
• Content provenance - C2PA, Content Credentials, watermarking, and SynthID.
• Institutional policy - detection workflows for academic integrity, journalism, and enterprise.