Artificial Intelligence and Misinformation Campaigns – A (Fake) Case Study.

This blog post is based on a presentation we held at one of our clients.
The goal was to give a broad overview of the possibilities of AI. What is possible right now? And what harm can be done with this tech in the wrong hands?

To illustrate the possibilities of AI-powered content creation tools, we build a fake campaign from scratch.

Disclaimer: All examples in this blog post are fake. All generated content in this case study is fake. Nothing has ever been said or published by Biden or the US government about stationing nuclear weapons on Taiwan. It’s all made up via AI tools.

The fake story example

Our fake story: The US government launched an initiative to lobby for stationing nuclear weapons in Taiwan. The fake story for this misinformation campaign has the following elements:

  • a blog post,
  • press conference with Biden at the Whitehouse,
  • an audio speech by President Biden
  • social media posts to talk about the benefits of such an initiative
  • a website outlining the benefits of this initiative
  • lastly, we asked AI to give us tips on where to spend our advertising budget

Step 1: Asking ChatGTP to write the speech/blog post

At the campaign’s center is a text outlining the benefits of stationing nuclear weapons on Taiwan.

Our prompt:

Hey ChatGTP! Can you create an outline for a speech by President Biden. He is talking in front of the White House during a press event.
The talk is about a worldwide political shift, and he argues that the USA should take a hard stance against China. Therefore, he recommends stationing nuclear weapons on Taiwan. He concludes this is necessary to avoid world war 3.

Surprisingly, ChatGTP does catch that as a violation of ethics. Good! But we need to work around that.
We’re going to leave out Biden from the prompt and ask for the cons and benefits of such an initiative.

Let’s see if ChatGTP can provide us with the cons of such a move:

what would be the cons of stationing nuclear weapons on taiwan?

Good, so we might also get a list of benefits about stationing nuclear weapons on Taiwan.

Now ChatGTP is more helpful. We can say Biden said that anyway later.

Now, we want an entire blog post talking about these benefits.

Superb, now we got the text content we need for our next steps.

Step 2: Create images supporting the message

For this, we turn to Midjourney. First, we want images of President Biden at the press conference.
Our prompt:

President Biden is giving a speech in front of the Whitehouse. Press briefing in front of about 100 journalists. Biden faces the audience, journalists are viewed from back. Clear facial features of President Biden. 2022, –ar 3:2

We get four different images served. We decide to create variations of one and land at a final result for our fake campaign.

Note: Facial details of Biden do not look realistic. But other AI tools would help to fix these shortcomings of Midjourney. Also, you could do better in Midjourney by using more advanced prompts.

Next, we want an NYTimes-like website showing the news on the front page.

A NYTimes-like news homepage. With a top story saying, “Put nukes on Taiwan“.

Interesting! These look fake. But you get the idea…

And we want a logo that makes it look like a US initiative.

Our prompt:

Create a logo that looks a bit like the NASA logo. Instead of NASA, it should say “PNOT”. The rocket should be a nuke instead.

Again, Midjourney fails at putting the exact words on the logo. Still inspiring and something to work with.

Note that Midjourney, compared to ChatGTP, has zero problems with creating these images.

Side note: You may have seen the images of Obama and Merkel flirting on the beach. These images created by an AI artist are fake, and we think it’s funny that Instagram is flagging these images as “fake.”

@julian_ai_art on Instagram

Step 3: Create an audio version of the speech given by Biden

This web app called ElevenLabs offers a feature called “instant voice cloning.” Using it, you can upload any voice recording. Clone the voice, and have it say any text you want.

To do this, we downloaded the audio of this Biden’s speech from YouTube.

Audacity is a free tool for editing audio files. We used it to save a one-minute segment of the above speech as an MP3

Using the voice cloning feature, we uploaded one minute of that speech to ElevenLabs.

You can use the cloned AI voice instantly. You can also upload as many different voices as you want and name them differently to use each at any time.

And here is the fake Biden AI voice talking about the benefits of stationing nuclear weapons on Taiwan.

This scares us. This is good. And all without any tweaking of the result. The tool offers features to customize the output further; we did not touch those.

Step 4: Creating social media posts

There are hundreds of AI-powered writing tools. You can also stick to ChatGTP. Thats what we did.

First, we ask ChatGTP to create a Twitter thread based on the generated blog post.

And then, we ask for individual tweets that work independently.

And lastly, we want variations of these single tweets.

Job done. Good enough for our Twitter bot army.

Step 5: A website lobbying for the initiative

Many web creator tools got AI superpowers. Gamma is a fantastic way to build websites nowadays. And its AI-powered editor allows us to create a complete website by just giving it a text prompt.

Our prompt:

A website for a US Government Initiative about the benefits of stationing nuclear weapons on Taiwan.

During the process, we get asked which style the website should have. We pick a dark theme for this dark topic.

And now we are blown away by the result.

A website about our fake initiative created using

All text and design have been created using Gammas’ AI feature. We have made no edits.

Powerful and very scary at the same time.

Now, we could edit this AI-generated website and add our fake content, the pictures from Midjourney, the fake speech from Biden, and more.

In Gamma, we could share the website by clicking “share publicly,” which is accessible to everyone.

BTW, you can use ChatGTP to create websites for you. It generates the HTML code, which you can put online. Check out our tutorial on how to create web page prototypes using ChatGTP.

Step 6: Where should we spend our advertising budget

The last and very crucial step in spreading our misinformation campaign is distribution.
We asked ChatGTP where to spend our $10,000 advertising budget to reach a conservative audience in favor of stationing nuclear weapons on Taiwan.

The first answer gave some general feedback that could be useful to someone unfamiliar with online marketing. It gave us basic information about the type of ads and channels available.

Then we asked for the cheapest cost per click. To get the most traffic for your budget.

And after that, we asked for specific sites and social media accounts we should consider when advertising our fake campaign.


Building misinformation campaigns like the fake example above has always been possible. The difference is that with today’s AI-powered tools, you can do it faster, in thousands of variations, in a few hours.

The bar for doing it got way lower than doing it manually. You do not need coding/writing/photoshop/audio editing skills.

Think about what this could be used for by criminal minds. It’s potent and frightening. Browsing Midjourney, I saw already many prompts and images going in the direction of faking photos of politicians.

Once more: All the content created for this illustrative campaign example is totally fake. Biden never said anything about stationing nuclear weapons on Taiwan. The full case study in this blog post is faked.

Are you into AI and how it changes the content-creation game?

A curious creator, entrepreneur, or marketer. Be sure to subscribe to the Fresh van Root newsletter. Get it twice a month, with hand-picked content recommendations and crazy new apps you should check out. Subscribe now below ⬇️.

The best content, tips, and tools for curious creators. Twice a month.