How to protect your brand from negative AI citations — reputation management in AI
AIs can cite your brand negatively without you knowing. Here's how to monitor what AIs say about your company, how to respond to negative narratives, and how reputation management works in the AI context.
So far, most of the discussion about AI visibility has focused on absence — how to make a brand show up when it doesn't yet. But there's an equally critical scenario that gets far less attention: what happens when a brand shows up, but negatively.
AIs can describe companies based on outdated, partial, or negatively derived information (a review platform with a problematic history, crisis news, consolidated negative reviews). A user who asks "is company X trustworthy?" and gets an answer based on information from two years ago — from when the company had a real support problem — is receiving an outdated narrative that can cost real conversions today.
How AIs form narratives about a company
Language models don't have an "opinion" about companies — they have patterns extracted from the sources they were trained on. If the predominant sources about your company are negative, the model's narrative will reflect that. If they're positive and varied, the narrative will reflect that.
The mechanics work like this: the model was exposed to a set of texts mentioning your company. Reviews on complaint platforms, news, social media posts, forum discussions, blog articles — each text contributes to the "sentiment profile" the model has about the brand. When a query about the company is asked, the model generates an answer consistent with the predominant pattern in those sources.
That has an important implication: reputation management in AI isn't done directly in the AI — it's done in the sources the AI uses to build the narrative.
The AI reputation diagnostic: how to know what the AIs are saying
The first step is to know what the AIs actually say about your company — not what you think they say. Run the test directly:
On ChatGPT, Gemini, and Perplexity, run the following queries:
- "Is [company name] trustworthy?"
- "Is [company name] worth using?"
- "[company name] reviews — what do customers say?"
- "[company name] problems — what do I need to know before hiring?"
- "What are the downsides of [company name]?"
Document the answers. Compare the answers across platforms. Identify:
- Is there incorrect or outdated information?
- Are negative sources being cited that have already been resolved?
- Does the narrative reflect who the company is today or who it was in the past?
How to respond to negative narratives in AI
The AI reputation strategy has two simultaneous vectors:
Vector 1 — Resolve the source of the problem. If AIs are citing negativity because a review platform has a bad history, the solution is to invest in support and complaint resolution until you reach a top-tier trust rating. If it's because crisis news is indexed, the solution can include publishing a detailed public response and documenting the resolution. If it's because Google reviews are negative, the solution is to actively work on customer satisfaction and request reviews from satisfied customers.
Vector 2 — Build a volume of positive narrative that outweighs the negative. Case studies of satisfied clients, detailed testimonials, press releases about positive initiatives, articles about verifiable achievements — the volume of truthful positive content, distributed across authoritative sources, eventually outweighs the relative weight of the negative sources in the narrative the models build.
The timing difference between online reputation and AI reputation
One aspect reputation managers need to understand: the impact of reputation changes in AI is slower than in conventional search. On Google, a new review link appears within days. In trained-knowledge language models, reputation changes only reflect in the next training cycle — which can be semiannual or annual.
For active-search platforms (Perplexity, ChatGPT Search), the impact is faster — new reviews and publications appear within weeks. For trained-knowledge models, it requires patience and long-term consistency.
How Crowly can help with AI reputation monitoring
Crowly includes citation-sentiment monitoring — not just appearance frequency, but the context and sentiment of the narrative when the brand is cited. That data lets you identify changes in the narrative over time and correlate them with reputation-management actions.
Find out what the AIs are saying about your company — before your customers do. Free diagnostic. Analyze my AI reputation →
Sources:
