If you were scared by the spread of the doctored Jim Acosta video, 2019 will terrify you.

Over the past year, technologists have witnessed entirely unprecedented advances in the field of AI-enabled forgeries and manipulations: from developments in motion transfer to better deepfakes to improved text-to-image generation.

Such progress did not go unnoticed. Opinion pages, think tanks, and even senators have since prophesied the collapse of democracy and the death of truth as we know it. Fake news was one thing, but manipulatable audio, image, and video — as metonymized by deepfakes — would assault the trust we instinctively place in our senses. More engaging and more believed than text, any and all photos or videos could become as doubted as a photoshopped magazine cover. Descartes’ wax would not just melt near the fire, it would disintegrate.

Rather than reiterate these dystopian forecasts, I would encourage newsrooms to treat such AI-generated forgeries as already arrived. In 2018, we saw algorithmically generated audio, image, and video reach unprecedented believability. In 2019, we will see how the public responds when such content is mass distributed.

The public response largely depends on how newsrooms will handle the detection of algorithmically generated content at scale. Last month, The Wall Street Journal was the first newsroom to publicly announce how it is preparing its team for the oncoming wave. Francesco Marconi and Till Daldrup argued that “traditional reporting” is the best way to detect deepfake videos. Certainly, the Journal should be applauded for its prescience and preparation. Yet detecting AI-generated forgeries is not simply a new version of an old reporting challenge. No amount of journalistic ingenuity and doggedness will match the scale, speed, and accuracy at which they will be distributed in the coming years. Instead, newsrooms will need to turn to software solutions — whether built in-house or purchased — for digital forensics. Such solutions will likely include metadata verification, trusted media capture, change level analysis, and public-facing proof points — in other words, the kinds of tasks humans cannot perform at scale.

To be clear: I do not believe that software alone will reverse the rise of AI-generated content. Nor do I believe that any single set of algorithms can restore the faith Americans once placed in journalism. But I do believe that in the coming year, newsrooms will have to turn to software if they want to combat the growing wave and ever-increasing quality of AI-generated content. Media companies can start this preparation by attaching forensic analysts to their newsrooms. Merely reporting on this new threat to journalism, democracy, and truth won’t be enough. To preserve faith in journalism, it will require a combination of software, continued reporting, and heightened public knowledge working against a problem that is already upon us.