This piece is Part 1 of a two-part series on narrative drift and the problem of AI-mediated brand meaning. It is possible both parts of the series are complete shit. I can’t quite decide, but I am certain the bigger problem I’m addressing is real. Part one explains what narrative drift is, why it matters in AI-mediated discovery. Part two lays out a practical framework for detecting and tracking narrative drift before it compounds into a brand equity problem.
I should call M. Night Shyamalan, because I have a horror movie script that’s scary af.
A wearable brand we are monitoring in Scrunch wants to position itself as premium. However, following a new product launch that generated a lot of coverage, they are being portrayed as a discount option by LLMs. The company screwed up their messaging, sent conflicting messages to ChatGPT et al, and now a brand equity nightmare — AI style — is unfolding. Correcting this is not going to be fast or simple.
In AI-mediated discovery, meaning is compressed, averaged, and replayed. This is a huge challenge for PR. Over time, that process can alter what the public thinks a brand stands for, even if the brand never changed its messaging.
That concept is called Narrative Drift.
Traditionally, it has a few meanings — it’s used in fiction writing and in marketing — but in the AI era it is worth refining. CMOs need to start building a framework for assessing it.
What is Narrative Drift?
Narrative Drift is the measurable gap between what a brand intends to be known for and how it is actually described by AI systems, media, reviews, and comparison content over time.
One question matters:
When AI answers a question that’s relevant to us, is our brand still explained the way we intend?
So, in other words, narrative drift is a qualitative GEO-presence issue.
Why it matters
AI systems preserve and reference patterns, not intent.
While LLMs are getting better at a range of tasks, multiple studies show they still struggle with basic factual accuracy.
- The Tow Center for Digital Journalism found generative search tools had major citation problems around news-related queries.
- A BBC / EBU study reported on by Reuters found that 45% of AI assistant answers contained at least one significant error, and 81% showed some form of inaccuracy or omission.
It is a leap, but a fair leap to assume that if AI systems struggle with facts, they will be even less reliable with brand meaning, which almost always involves implicit factors and interpretation.
How Narrative Drift happens
Drift happens over time and it compounds. Unlike many reputational crises, it’s sneaky.
A few common causes:
- Early coverage frames the brand narrowly and is never corrected
- Review language overwhelms brand language due to volume (meaning there is a balance problem between earned and owned media effectively)
- Core narrative pillars stop being reinforced
- AI summaries overweight whichever attributes repeat most
Once a framing takes hold, it tends to reinforce itself. AI systems don’t ask whether a description is strategically correct. They are pattern-matching monkeys.
In other words, narrative control is lost incrementally and then via accumulation. And while this is happening, you might be getting a high volume of media coverage that lulls you into thinking your PR function is doing its job.
Scary? For folks in PR, that’s a Nightmare on Narrative Street.
So how can you track this?
Stay tuned for Part deux of this semi-baked series to find out.