Summary of Key Points
Artificial intelligence (AI) has almost eliminated the cost of content creation, but it has significantly increased the “cost of trust”—readers no longer readily believe that the content produced by companies or public figures is the result of genuine thought. To address this issue, the industry has introduced AI watermarks (such as the new version of Claude from Anthropic) and high-precision detection tools (like Pangram). At the same time, corporate content is being forced to return to its “true value”: real human involvement in thinking, verifiable research, and exclusive insights are now required to rebuild trust.
Why Is AI-Generated Content Concerning? The Core Issue Is Trust
When companies issue press releases or open letters, they are essentially making promises to the public—such as about their strategic plans or social responsibility commitments. Readers assume these are the result of careful consideration by real people. However, AI-generated content often lacks sincerity:
- Journalists Can See Through It: 53% of journalists oppose the use of AI-generated content by companies (according to a Cision report), as they are familiar with AI’s writing patterns and recognize its repetitive details and indirect language (e.g., “We do not pursue short-term profits but focus on long-term value”). These are typical outputs of AI-based predictions that require additional effort from readers to understand.
- The Sense of Neglect: Even if the content is only lightly polished by AI, readers may suspect a lack of genuine thought when they detect AI influences. For example, the founder of Zomato was criticized for using AI to write a response regarding the gig economy because the message lacked human warmth and specific details.
Technology to the Rescue: Making AI Content Transparent
To make AI-generated content traceable, the industry has adopted two approaches:
- AI Watermarks: The new version of Claude from Anthropic adds embedded watermarks to text, images, and code (available globally and open for testing), in compliance with the EU’s Artificial Intelligence Act requirements. This serves as a label indicating that the content was created by AI.
- High-Precision Detection: Pangram, a company funded by Menlo Ventures, uses detectors trained with “synthetic images and contrastive learning,” with an error rate of only 0.01%, ensuring that AI is not wrongly attributed to human authors. Deedy Das, a partner at the fund, stated, “AI has reduced content costs to zero, but it has also increased the cost of trust; detection tools are necessary to restore trust.”
How Can Corporate Content Stand Out in the AI Era?
Content that encourages readers to spend time reading shares several common characteristics:
1. Endorsement by Trusted Platforms: For example, papers by DeepMind’s team published in top academic conferences have credibility due to the conference’s rigorous review process.
2. Authentic Research: Articles written by AI for The Economist are the result of editors testing multiple large language models (LLMs) and include concrete experimental data.
3. Human Responsibility: The founder of Pangram personally participated in the communication and writing of articles for The Atlantic Monthly, giving them a sense of authenticity. Although AI is used for research and data organization, the final decisions and responsibility lie with humans—AI serves as a tool to enhance efficiency, not a substitute for human thought.
How Should AI Be Used?
We can learn from the “smart use” of 19th-century painters. For instance, painter Sargent used photographic sketches as a foundation but completed his works in an impressionist style, leveraging technology while preserving human creativity. In today’s context, companies should use AI for:
- Menial and Laborious Tasks: Collecting data, organizing materials, and creating preliminary drafts.
- Human-Centric Tasks: For critical thinking, verification, writing, and making commitments.
Conclusion
AI is not an enemy of corporate content; rather, it acts as a filter, eliminating superficial and formulaic material and encouraging companies to engage in genuine thought and sincere communication. In the future, trustworthy content will be the result of AI assistance combined with human responsibility.