Author: Olena Churanova
Today, it is hard to imagine an area of life that has not been penetrated by artificial intelligence, which has offered alternative solutions or simplified ways of overcoming problems. For those fighting fakes and disinformation, the arrival of generative AI (a type of artificial intelligence capable of creating new content, such as text, images, music, or other media ) has brought new challenges and threats and additional opportunities to improve efficiency.
The work of a fact checker involves the use of specialized AI tools in the following areas:
- Transcription and text analysis;
- Translation;
- Search and systematisation of information;
- Content identification and verification (working with photos and videos).
Here’s a look at StopFake’s favorite AI tools, which can be useful in verifying information.
Text transcription and analysis
This free artificial intelligence-based tool for journalists from Google is a handy tool for transcribing long speeches, interviews, or Putin’s next ‘historical lecture’. It is enough to upload the desired video or audio file, pre-specifying the language of the recording (this is important for proper transcribing), and the program will automatically create a text file with a breakdown of the speech by timing. In addition, you will be able to see which people are mentioned most often in the document, as well as which organisations or places are quoted. Also on this site, you can download large documents and transform them into tables for analysing information (this option is still in beta version), and view documents that other colleagues – for example, The Washington Post – have made publicly available.
Another tool to help fact-checkers work with text is NoteGPT, which allows you to quickly turn a YouTube video into text, create a summary, and even answer your questions about the video. The features are fairly limited in the free version, but educators who sign up with a university email address are given free access for some time.
Translation
Professional translation tools are extremely important in the work of an international fact-checking editorial team. StopFake fact-checkers constantly encounter fakes when checking foreign media news sites, which are not always in English, so such tools become indispensable.
While we all have access to Google Translate, artificial intelligence has significantly improved translation services, making them more accurate, multilingual, and capable of AI-assisted translation editing. One such tool is Deepl. It can be used for free, but with limits on the amount of text to translate per day. Deepl allows you to download documents for automatic translation in PDF format. There’s also a Deepl Write feature that helps you rewrite text or suggest other options from the AI. You can even dictate text through a voice recorder and have it translated.
Search and organisation of information
Another useful tool from Google is the Google Fact Check Explorer. With this tool, you can check whether other fact-checking organisations have already written debunks on this topic, in which countries this fake has been spread, etc. There is an option to search for a refutation on a specific site, for example, StopFake.org, and you can also see what fakes about Ukraine have been spread abroad recently. It is also possible to search for refutations by picture because propagandists can use the same photo in different ways and with different disinformation narratives.
Many AI-based tools help track the spread of misinformation across platforms. StopFake, for example, actively uses the Osavul platform for this in its work. There are also free-access tools developed by universities or think tanks such as the Hamilton 2.0 Dashboard. This is a project of the Alliance for Securing Democracy at the US-German Marshall Fund. It monitors and provides summarised analysis of narratives and themes promoted by Russian, Chinese, and Iranian officials and state media on X, Telegram, YouTube, Facebook, Instagram, TikTok, state news sites, and in official press releases and transcripts published by their Foreign Ministry. On the platform, you can search keywords, statements, and events, and view context regarding certain narratives, including hashtags, URLs, and interactions with other accounts. This is important for understanding how misinformation is spread and what channels are being used.
Also, when analysing information, we always pay attention to where it first appeared. In particular, if it is a dodgy site, we need to check where it was registered, when it was registered, and whether there are other sites associated with it. The Domain Digger tool is convenient in that it immediately collects data from other platforms already known to fact-checkers, such as Whois, in one place.
Of course, when talking about tools based on artificial intelligence, it is impossible to bypass chatbots. Each user chooses the most convenient option for their current tasks, but in my opinion, Perplexity is the best among them for fact-checkers. It is not just a chatbot, but also a search engine that allows you to specify exactly what you want to find through dialogue.
An important advantage of Perplexity is that it provides links to the sources used in the answer. This is critical for fact-checking because we should not forget that there can’t be 100% trust in AI – it is just generating information available in the database of this tool.
How can Perplexity make a fact-checker’s job easier?
- Highlighting key points in a document and analysing large amounts of information without re-reading thousands of pages.
- Focus function that allows you to narrow your search to a specific category, e.g. only YouTube videos.
- Clarification of obscure concepts and topics, as well as links to sources and experts who can provide more in-depth explanations if the rebuttal is on a topic you are not an expert in.
- Translation features are similar to other chatbots.
It is important to remember that the quality of the response depends on the quality of your inquiry. It is also necessary to adhere to safety rules: in case of investigations on sensitive or dangerous topics, you should not upload such documents to chatbots.
Content identification and verification (working with photos and videos)
The topic of deepfakes (a technology that uses artificial intelligence to create fake video, audio, or images that look realistic) and their refutations certainly deserve a separate article. However, if we talk about the top 10 AI tools for fact-checking, a tool for analysing deepfakes should be among them. Of course, verification of this type of fake should be combined – not only the use of tools but also your observations of various dubious details on the video. But if a simple assistant for verifying video fakes can be offered, Deepfake-o-meter is just the thing. It automatically analyses content for signs of manipulation using various methods.
The Russian-Ukrainian war has prompted many fact-checkers to create their own AI-based tools. In particular, our colleagues from Faktisk Verifiserbar (Norway), together with researchers from the University of Bergen, have created a tool as Tank Classifier, which identifies and classifies tanks or artillery vehicles present in an image. They have also developed another useful tool to identify the language spoken in a video or audio file.
Lastly, a rather useful tool that demonstrates how much information AI can provide based on just one photo. After uploading the right photo, you will get a detailed description (maybe some details you didn’t pay attention to will be highlighted), EXIF data (metadata) of the image (if the original is uploaded), context, etc. As one of the starting points for investigation, this easy-to-use tool is very useful.
AI-based tools abound – and every day they evolve, become obsolete, and disappear, and completely new ones appear. However, they will never replace human critical thinking, observation, and ingenuity, without which one can find neither the source nor the necessary facts and arguments to refute misinformation.
Also read our previous article about how AI photos are used by Russians in information warfare, and what tools can be used to verify them.
The article was prepared with the support of NDI
