AI in PR: Data Governance Shakes AI Hallucinations

Published on October 22, 2025

Build a process to make GenAI content accurate, consistent and dependable. 

In public relations, the use of artificial intelligence comes with uncertainty and fear. By borrowing the technology best practices around data governance, marketers can safely leverage AI to provide trustworthy information. 

Businesses are eager to tap the power of AI to craft and personalize messages, identify trends and streamline operations. But concerns over AI quality are still unresolved, especially the potential for errors or “hallucinations.” The prospect of an AI-generated press release amplifying fabricated statistics, or a social media campaign built on inaccurate data, threatens the authority of a brand voice and the consumer’s trust.

Assuring that level of quality takes a clear set of processes to collect, store and use information, what the tech world calls data governance. The marketing ecosystem needs ways to manage the risks and fully leverage the abilities of generative AI. 

Avoiding the “garbage in, garbage out” cycle is just the beginning. Marketers need to know more than whether AI content is accurate, consistent and dependable. They must get a handle on whether they are dealing with intellectual property a client can control, AI-generated plagiarism, or some gray area locked into AI’s gray matter.  

Artificial Intelligence in Marketing

• Data Governance Shakes AI Hallucinations
• Super-Snippet SEO: Generative Engine Optimization for Google AI Overviews
• Best AI Tools for PR Tasks
• Top 10 Generative AI Marketing Examples

First Steps: Crack Open AI’s ‘Black Box’

Generative AI has been in development for years, yet its algorithms still produce factually incorrect or misleading results. GenAI is a prediction machine: It doesn’t know the answer, it only calculates the most likely one, without any real-world grounding beyond what it has absorbed from its content library. When its database is short on known facts, odds are that GenAI engines will simply produce plausibly factual, truthy statements.

Like search engine algorithms, a GenAI model is a “black box” that hides its workings. It generates the true and the nearly true the same way. GenAI can take cues from the writing style, level of detail or reputation of the source, but only predicts accuracy. Validation is up to the user. 

Data governance enters the picture by setting up a process to use GenAI consistently and reliably. The first step in data governance is to get a sense of what could possibly go wrong:

• Limited Data. No AI model has complete, real-time information. When GenAI stretches beyond its static knowledge base, results can be hit or miss.
• Unchecked Bias. By following patterns in language, GenAI widens the gaps in its knowledge. Free speech is its raw material, which can be uninformed, opinionated, fanciful, or just wrong. 
• Hallucination. Designed to generalize, chatbots can misstate facts, sources and figures with complete confidence.
• Interpretation. The bot doesn’t ask follow-up questions. Instead, it’s apt to infer too much from a vague prompt or get lost in a complex or nuanced statement.

As marketers, we can verify what AI asserts, applying our research skills and subject matter expertise. Yet we also play the prediction game, just like AI. Does the model draw from primary and reputable sources? Is the information recent and relevant? Even human-generated copy can fail these tests. But good editors have trained themselves to spot BS. Often, prediction is good enough.

So, data governance starts with finding low risk, high reward AI uses for which no fact-checking is necessary. The low-hanging fruit among AI use cases capitalize on efficiency, not accuracy. They resolve pain points in a longer process with plenty of human intervention: 

• At the Beginning: Marketers can make AI their starting point for brainstorming marketing slogans, storyboards or pitches. 
• In the Middle: Chatbots can draft internal emails, generate meeting transcripts, list next steps and perform other routine work for human review. 
• At the End: PR practitioners can also use AI as a final flight check, proofreading or tweaking copy to match the client’s voice and tone. 

Technology PR Case Studies

• Aviation Services Branding Takes Flight
• Birthing an International Life Sciences Center
• Establishing a New Retail Technology Leader
• Social Media Framework Sets Stage for Project Management Success

Next Steps: Set Rules for Safe AI Use

Beyond these efficiency hacks, marketers can speed their research and content tasks if they lay down a few rules: 

• Equip the Toolbox: Decide which AI platforms to use, and start paying for them. As the saying goes, you are the product of a free service. In this case, whatever the chatbot reads goes in its library, maybe to be pulled out and shared competitors. Subscribe to full-featured services for the team and bar going rogue with less-secure alternatives.
• Lock It Down. Anyone buying GenAI tools for content development should check settings and review the terms of service (or have AI summarize them) to make sure proprietary client or company data remains confidential. Whenever possible, turn off features that retain information for future use or “to train our models.” Keep style guides and prompts elsewhere, in text files or spreadsheets, and upload them as needed.
• Strip It Out. It’s only a matter of time till the AI gets hacked. Some things simply should not be exposed to that risk, which is one reason why companies rule out personal targeting using GenAI. Before uploading any files, remove sensitive data:
  • Personally Identifiable Information: No names, emails, phone numbers or addresses of clients or employees.
  • Client Confidential Data: No client strategies, trade secrets, non-public financial data, or internal communications.
  • Proprietary Company Information: No internal strategies, source code, financial reports, or unreleased product details.
• Train the Team. Have a central place to store AI prompts, personas and brand style guides. Discuss what does and does not produce relevant, on-brand responses. 
• Make a Commitment. Create a simple policy to guide co-workers and inform clients. Disclose significant GenAI use and start documenting how the team reviews machine-generated content. 

The role of artificial intelligence in public relations is to optimize messages for human consumption–to be engaging, clear and compelling. Data governance frameworks show how far technology will take marketers toward those goals, and it gives them a better foundation for heading in purposeful new directions.