In the Google-search engine optimization (SEO) era, building a narrative about an organization took effort. It required clicking through sources, weighing credibility, deciding what was relevant and then piecing together a narrative. With AI, that synthesis is done for the consumer. That is the nuance many organizations misunderstand and fail to account for as they rush to optimize for generative engine optimization (GEO).
AI Search, Visibility and Reputation
The stakes are real. In 2025, 72% of S&P 500 companies disclosed AI as a material risk in their SEC filings, and reputation damage was the most commonly cited concern. But, so far, most of the conversation has focused on visibility: How often does our organization appear? Which sources are cited? Are we showing up favorably?
The more consequential question is about vulnerability: What does AI-driven search reveal when someone is actively looking for controversies or criticisms?
We tested several organizations to find out. Across five of the biggest AI platforms (ChatGPT, Claude, Gemini, PerplexityAI and Google AI Search), we asked adversarial questions—the kinds of questions that someone would ask if trying to drive a negative narrative. Every platform replied with a structured overview that synthesized years of scattered criticism from watchdog sites, ideological opposition groups and think tanks into neat categories. In one case, content spanning 60 years, including a congressional inquiry from the 1970s, a think tank analysis from 2003, a watchdog profile from last year and a Reddit thread from last month was assembled into a single, coherent five-part narrative that no individual source had ever produced.
AI constructed it from the pieces, presenting a hit piece and an investigation with equal weight and authority.
How to Re-Train the AI Narrative
Once AI models ingest a narrative and train on it, that narrative sticks. Each successive model training cycle embeds it deeper, and most executives don't know this is happening until it surfaces in a crisis. The good news is that this is a solvable challenge.
The Audit: Three Steps To Know Your Real Risk
-
First: Test with adversarial prompts across the five major platforms. Ask the questions a critic would ask about your industry's scandals and the organization's historical vulnerabilities. Document what comes back, especially the sources. What's the framing? What's the sentiment?
-
Second: Identify your source displacement targets. Which publishers, outlets, analysts or platforms is AI actually drawing from in your category?
-
We know AI systems are not only drawing from corporate websites, press releases or branded content. They are also pulling from what others have written: journalists, analysts, watchdog organizations, Wikipedia, industry experts and critics.
-
If AI is citing three specific news outlets as the primary authority on your industry, and those sources have documented negative narratives against you, those are your vulnerability nodes. That's where you need different sources to balance out an unfavorable narrative.
-
-
Third: Build a quarterly refresh cadence. AI models update constantly, and citation patterns shift. What you fix today may drift again in six months. The organizations that will stay ahead of AI-driven reputation management will audit quarterly to measure visibility and how the narrative about them is shifting across each major platform.
Displacing A Narrative
Once the vulnerabilities are identified, the strategy should be to secure favorable narrative in the sources AI draws on when forming its answer.
We tested this directly. Following the publication of an op-ed by an organization’s leadership on an issue important to them, we distributed a press release through a newswire service known to be indexed by major AI platforms. Within weeks, the team saw the results. On direct queries about the organization's position, all five platforms returned accurate, well-sourced responses, with several citing the press release URL directly and others quoting the op-ed in the organization's own language. A single earned media placement, structured and distributed correctly, demonstrably shifted how AI described the organization.
The Broader Lesson
AI platforms are not static. They index new content, update citations and shift narratives over time. This is a lever. Knowing which outlets AI trusts—and earning coverage there—is how organizations move the needle.
The gap between awareness and preparation is where vulnerability—and opportunity—lives. The organizations that move first will audit their AI narrative before it becomes a headline. A GEO strategy should not be about just search visibility. It should be about narrative vulnerability, because when you close those gaps, you're not just protecting a reputation. You're taking control of it.
Beth Kanter is an Executive Vice President at SKDK. Vishakha Mathur is a Vice President at SKDK.