es 
The digital space is rapidly flooded with increasingly sophisticated synthetic audiovisual content, videos taken out of context or manipulated photographs. Much of this material is part of campaigns featuring stigmatizing narratives specifically designed to criminalize activists, journalists, and civil society organizations, using structures such as institutional amplification networks.
Given this scenario, forensic verification of multimedia files and automated discourse analysis have become key processes for examining the authenticity of a visual file and mapping the evolution of a malicious opinion campaign in real time.
There are four tools you can use to verify multimedia content and track stigmatizing narratives with AI:
1. InVID WeVerify
This free browser extension (essential for global fact-checking) allows users to break down any video from X, TikTok, or YouTube into keyframes to perform automatic reverse searches across multiple search engines simultaneously.
It also features advanced forensic filters, such as matrix quantization analysis and heat maps based on compression inconsistencies, that identify whether a specific section of an image has been altered, cloned, or digitally edited.
2. Google Lens y Yandex Search
These reverse image search engines use AI-based computer vision algorithms to identify visual patterns, faces, objects, and geographic locations.
Yandex stands out for its pinpoint accuracy in finding faces or video clips on platforms in Eastern Europe and Asia, while Google Lens excels at identifying the original context of Western images and extracting text embedded in screenshots.
3. Hive Moderation y Deepware
These platforms are trained to scan files for digital fingerprints left by generative models, such as Midjourney, DALL-E, or cloned voice generators.
When a suspicious image or video clip is uploaded, the system generates a probability percentage indicating whether the content was created synthetically (either entirely or partially) and highlights suspicious areas of the face or audio distortions.
4. GPT-4o y Gemini 1.5 Pro
These advanced multimodal models can simultaneously analyze video footage of a public speech, its transcript, and comments posted on social media to identify recurring rhetorical patterns.
They are ideal for processing large-scale state discourse and scientifically classifying the categories of stigmatization used against independent journalism; for example, by detecting the systematic repetition of labels such as “foreign agents,” “traitors,” or “false positives.”
How can we instruct AI to correctly detect a deepfake?
Having access to the right tool is only half the battle; the other half lies in how you interact with the AI. To uncover the structure of a manipulation campaign or attack based on a speech, a TV transcript, a social media thread, or a press release, it’s essential to require the AI to rely solely on linguistic markers and verifiable manipulation patterns, to back up every conclusion with an exact verbatim quote, and to avoid making political or emotional value judgments.
That’s why we’re presenting a prompt that can work in this case. Copy and paste this instruction, substituting what applies to your situation based on the result you want to achieve:
Act as a forensic linguist and digital manipulation analyst specializing in human rights and the dynamics of digital authoritarianism. Your goal is to break down the attached text to map out rhetorical structures of stigmatization or coordinated informational manipulation campaigns.
For your analysis, you must strictly follow these three quality-control rules:
1. Do not render political or emotional value judgments. Your analysis must be based strictly on the presence or absence of specific linguistic markers, psychological manipulation tactics, and amplification patterns.
2. Every conclusion or category of attack you identify must be accompanied, without exception, by an exact quote from the original document in quotation marks.
3. Present the results directly in an organized structure with clear subheadings and tables, omitting introductory summaries.
As an additional suggestion, you can take the list of keywords and hashtags that the AI identifies in that analysis and set them up in your social media monitoring dashboards: this allows you to anticipate waves of coordinated attacks or spikes in manipulation content before they go viral on a large scale.
It is important to note that not all of these AI tools are available without restrictions in every country. In countries where access to certain digital services is blocked or restricted, it is common for some of these tools to be inaccessible or to display errors; in such cases, using a VPN is usually the quickest solution. Furthermore, artificial intelligence tools serve as a support to streamline journalistic work but do not replace judgment, rigor, or the final product.
