OpenAI released a demo of Sora on Thursday last week. Sora is the most impressive text to video AI model we’ve ever seen. Fluid motion, realistic people, and video you’d believe was shot by a huge crew, it really is a sight to behold.
We can assume that Sora will be available in a beta within a couple of weeks. OpenAI will as usual make a big deal of AI safety with the beta, but we can expect restrictions or limitations to fall away over time.
Generative AI video makes video the latest untrustworthy format. Fake video is nothing new, but video has hitherto been a more trustworthy format as it’s expensive and inflexible. That is now no longer the case.
Equally, trust has been one of the defining issues in search over the last 12 months. With this, it’s just become even more important.
Video was once thought trustworthy; no longer trustworthy
Video was, until last Thursday, an inherently more trustworthy format as it’s so inflexible. It’s trivial to change text, but editing a video is – or was – inherently limited by the original source material. Until last Thursday, you couldn’t add in a new shot in post; if you didn’t film it, you couldn’t just add it in.
Deepfake video was a resource-intensive pursuit. This is no longer the case.
Sora is a technological marvel that’s a similar leap forwards to the GPT models for text. GPT-3, the precursor to the model used for ChatGPT that was released in 2020, was the first time reasonably accurate and useable text could be generated. Sora feels similar: this is significantly better than anything we’ve seen before and it’s impressive now, but have no doubt it’ll get wildly better over the coming months and years.
One can easily speculate about the future: it becomes easy-and-cheap to generate realistic-looking videos of anything. It’s most obvious to compare Sora to stock video at the moment, but this misunderstands “stock” in stock video: “stock” refers to generic and off-the-shelf materials. Sora offers the same ease, convenience, and presumably cost, but it promises to be indistinguishable from footage you’d otherwise shoot for the project.
There has been fake video for almost as long as there’s been video, but this has not been a mainstream topic. “Finance worker pays out $25 million after video call with deepfake ‘chief financial officer’” was major news merely weeks ago because the event was rare and noteworthy. The camera is now surely the ultimate unreliable narrator: when watching realistic video one typically assumes it’s real. That can no longer be the case.
100% human verified content
Sora is interesting because OpenAI is not telling us about the use cases it sees. This contrasts to its other models. The GPT-4 release notes, for example, promised “more useful responses”. The release of Sora focuses on the technological marvel and merely tells us that OpenAI is working to “identify positive use cases for this new technology”.
Pause for a second. This is remarkable: do they not know any positive use cases?
Sora is currently the most technologically advanced “stock” video library. You can have the perfect background video to accompany your content. Maybe even use text-to-speech to read your content, generate accompanying video with Sora, and you’ve automated YouTube content. I’m sure an app offering this will be at the top of Product Hunt within hours of the API becoming available.
Let’s run with a thought experiment for a minute: social media has given rise to a new level of parasocial relationships. We all have media personalities whom we trust. Sora makes the possibility that the person you trust doesn’t exist much more possible than it was a week ago. Trust is a critical component of the consumption of any information, whether it’s on a billboard or a piece of SEO Content, and Sora threatens one of the final at-least-generally-trustable formats.
How can your readers, viewers, or prospects trust your content in an untrustworthy world? How do we need to evolve our content to be the rare and valuable source that people do trust?
E-E-A-Ting the future
Google placed much more emphasis on E-E-A-T (Experience, Expertise, Authority, Trust) last year. The 4 major updates to the search algorithm trended towards rewarding content compatible with E-E-A-T requirements.
Given an influx of generated text, you can see how this makes sense: if you can better trust who is responsible for content and their experience, expertise, and authority, you can go some way to dealing with the challenges of AI-generated content. However, even with text, this has proven difficult so far: Google has already made plenty of mistakes and fixes in the last few months alone.
High-quality AI video makes this problem even worse and will accelerate the trend towards E-E-A-T. Those who move now to increase the trustworthiness of their content and get ahead of the competition are going to win.
It’s reasonable to expect searchers to start reading content with an ever-more critical eye. If your default assumption is that the text you’re reading or video you’re watching is AI-generated, then content that goes out of its way to show you why and how it can be trusted is suddenly noteworthy.
Respond to generative AI by making ever-more useful content
This is the opportunity in 2024: respond to generative AI by making your content radically more useful than ever. Cut the fluff and focus on the value. Need I remind you: SEO is winner-takes-all. Even the smallest gain over your competitors results in the win and aggregating all of the gains.
Generative AI makes it easier to create mediocre content. Bring it on! If your competitors are churning out mediocre, untrustworthy content, then it’s easier than ever to beat them with your actually useful content. Sora raises the bar on what we need to do to demonstrate trust.
Google’s SEO guidelines, paradoxically, say to ignore the guidelines and focus only on the usefulness of your content. The relentless genericism of Generative AI and Sora is currently incompatible with actually useful content. It’s certainly incompatible with trustworthy content.
This opens the door: trust is critical. SEO may even become easier as your competitors voluntarily reduce the usefulness of their content and thus decrease its performance by churning out generative AI spam. Get ahead on this now, and you’ll see the benefits in 2024 – all while your competitors will still be searching for answers.
