Narrative Drift: Questions Executives Are Asking
Michael Etheredge is VP of Product Development at QuestionFuel, where he leads how the firm measures and corrects the way AI answer engines describe companies. He was featured by the Southern Economic Development Council on how AI is changing the way companies research communities and regions.
Narrative drift is the gap between how a company describes itself and how AI answers describe it. Once leadership teams understand the mechanism, the same six questions come up almost every time.
What follows are those questions and direct answers, written as a companion to our deeper explainer on narrative drift and AEO.
Highest risk: quiet periods, earnings, transactions, regulatory milestones.
Key takeaways
- Narrative drift is the gap between how a company describes itself and how AI answers describe it, and most companies have no process for checking it.
- The two causes are stale information the model still treats as current and competitor framing that goes unanswered.
- The same company can read as a category leader in one AI platform and a niche player in another, so one spot check is not enough.
- Quiet periods and regulatory silence are the highest exposure windows, because official communication stops while AI synthesis continues.
- Boards are beginning to treat AI visibility as a governance question rather than a marketing metric.
Quick answer: Narrative drift is the gap between how a company describes itself and how AI answers describe it. It happens for two ordinary reasons: stale public information a model still weights as fact, and competitor framing that no one has answered. The practical detection method is not a platform but a manual audit, running the 10 to 15 questions your buyers, investors, and journalists would actually ask across the major AI platforms and reading the answers as a stranger would. Below are the six questions executives raise most often, answered directly.
Is AI getting our brand wrong, and how would we even know?
Probably, at least in some conversations, and most companies have no way to find out. Nearly every organization has a process for monitoring social mentions, review sites, and press coverage. Almost none have a process for checking what ChatGPT, Gemini, Claude, or Perplexity say when a customer, investor, or journalist asks about them directly.
That gap is the risk. These conversations happen inside closed chat interfaces with no referral link, no alert, and no way to see them after the fact. A customer can form a wrong impression of your pricing, your leadership, or your market position in a single exchange, and your team will never know it happened.
The starting point isn't a monitoring platform. It's a simple manual audit. Run the 10 to 15 questions your buyers, investors, and journalists would actually ask across the major AI platforms, and read the answers as if you were hearing them for the first time.
What actually causes narrative drift?
Two forces, and neither one is malicious. The first is stale information: a product description that hasn't been updated, an old news story that never got a follow-up, a leadership bio that's a role behind. AI models don't know it's outdated. They weight it as fact.
The second is competitor framing. When a competitor publishes comparison content, category definitions, or "best of" lists that mention you, they're shaping how AI describes you, on their terms. If you're not producing that same kind of content yourself, you're ceding the framing entirely.
Both forces compound over time. A brand that goes quiet on a channel doesn't stay neutral in AI's eyes. It becomes whatever the loudest surrounding content says it is.
Do different AI platforms describe our brand differently?
Yes, and that's worth taking seriously rather than treating as noise. Each model, including ChatGPT, Gemini, Claude, and Perplexity, trains on different data, updates on a different schedule, and weighs authority and sentiment differently. The result is that the same company can come across as an established leader in one model and a niche also-ran in another.
Think of each platform as a separate focus group that's already formed an opinion of you, built entirely from whatever fragments of your story it happened to absorb. Some of those fragments are things you published. Many are not.
This is why a single spot-check on one platform isn't enough. A brand that looks well-represented in ChatGPT can be nearly invisible in Gemini, and neither result tells you the whole picture on its own.
What can leak into our AI narrative that we never meant to be public?
More than most companies expect. Old internal wikis, outdated sales decks, partner enablement documents, leaked memos, forum speculation, even a long-abandoned support thread. If it's accessible online, even buried several pages deep, it's fair game for a model to absorb and synthesize into an answer.
This is sometimes called shadow drift: the AI-generated version of your brand includes details you never intended to be part of the public record, mixed in with your official messaging as though it carries equal weight. A model doesn't distinguish between a press release and a stray comment on an old forum thread: both are just text it learned from.
The practical implication is that brand control now extends past your owned channels. It means periodically checking what's indexable that shouldn't be, not just what's published that should be.
What happens to our narrative during a quiet period or regulatory silence?
This is where narrative drift turns from an inconvenience into real exposure. When a company is legally or strategically constrained from speaking, whether ahead of an earnings call, during a clinical trial period, or amid pending litigation, official communication goes quiet. AI synthesis does not.
Into that silence flows whatever is already circulating: old coverage, speculative forum threads, analyst commentary, a stale product page. The AI answer someone gets during your quiet period may have nothing to do with what's actually happening inside the company, but it will sound just as confident as if it did.
For regulated industries especially, going quiet is a normal, often required business practice. But it also becomes a period of active reputational vulnerability that most companies aren't monitoring at all.
Is this a board-level governance risk?
Increasingly, yes. Boards are starting to treat AI visibility the way they treat cybersecurity or compliance: not a marketing metric, but a category of exposure that needs oversight. The moment that usually triggers this is an investor or executive searching the company name in ChatGPT or Gemini and getting an answer that's wrong, thin, or outdated.
When that happens, it doesn't register as an SEO problem. It registers as a governance question: did anyone know this was happening, and why wasn't it caught earlier. A model misstating your pricing, your market position, or your leadership isn't a ranking issue; it's information moving to decision-makers without anyone checking it first.
That's the case for treating this as a standing practice rather than a one-time fix: regular checks across platforms, a clear owner, and a way to correct the record when drift is found, before a board member finds it first.