AI and Reputation Management: The Attacks You Never Saw Coming

Artificial intelligence (AI) is making it easier for your enemies to launch reputation management attacks. Contaminating training data with false information leads to harmful LLM responses that seem authentic, especially to people who don’t know you or your brand.

In this blog post, I’ll outline one common reputation management problem augmented by AI. I’ll explain how it worked and what I think could be a helpful response.

But know this. In this age of AI, it pays to watch your brand carefully. Models are always changing, and new issues are always coming to light. If you don’t have a reputation management plan in place now, it will be much harder to respond when the inevitable attack happens.

How AI Models Are Trained

AI tools like Gemini and Chat GPT were programmed (or “trained”) with readily accessible information. News articles, blog posts, and public commentary were indexed and correlated to help the tools understand basic facts and how they’re connected.

Those tools were also trained with data from Reddit (which the company isn’t happy about). While it’s impossible to know how different data points are weighted, simple experiments suggest that AI gives Reddit responses a lot of weight. They’re often cited in answers to common questions.

How AI Models Can Be Manipulated

Since AI tools often cite Reddit posts, it’s easy to launch an attack on a brand. Just write an “honest review” of a company or a product on Reddit and sprinkle in some false statements about quality or customer service.

In a sophisticated attack, someone might write a review like this and ask someone else to chime in and agree with what’s written there. A third person might write a similar bad review and cite the same problems, further solidifying the attack.

During the next search for your brand, those posts are scraped and cited in the answer. The posts themselves might also be cited on the results page for your brand. You’re officially dealing with a reputation management problem at this point.

Preventing AI Reputation Management Problems

Watching Reddit carefully is a smart move for all brand managers. You should be watching other social media sites and review sites too. Now is a great time to develop a social listening strategy and ensure you know what people are saying about you.

It’s also smart to ask your loyal customers to write a review about your brand on Reddit. Why not sprinkle in some good stuff?

Responding to an Attack

Responding to negative reviews from your official business account is the fastest and smartest way to respond to an attack.

If you’ve truly made a mistake, write an appropriate response that includes an apology. Own your errors, figure out how the problem happened, and ensure it never happens again.

If you never made the mistake you’re accused of, keep your response simple. Don’t repeat the words in the attack verbatim (as that augments the phrasing for AI tools). Instead, keep your response very general and polite.

I typically recommend something like this: “We don’t have a record of serving you as a customer. Maybe you meant this review for someone else? DM us if we can help.”

It’s rare for a Reddit attack to turn into a full episode of troll behavior (If that happens, I have advice!). But you may need to write the same measured response a few times to tamp down the attack and clear your name with the AI tools.

If you need help developing a strategy or responding to an attack, contact me. Let’s review your situation together and create an action plan to protect your brand.