What Agencies Get Wrong About AI Content

Agencies have a larger role in AI discovery than producing more content.

The challenge is making brand meaning hold together across the public record.

Public relations and communications firms are increasingly expected to have a point of view on AI. Most conversations begin with production: faster copy, more formats, more content, and new ways to use the tools.

Those capabilities are useful. But they address a different problem from the one emerging in AI-driven discovery, where systems increasingly interpret information about brands and assemble their own descriptions, comparisons, and recommendations.

The mistake is treating AI as an output problem

There are two conversations about AI content that often get bundled together. The first concerns material created using generative AI: campaign drafts, social posts, email variations, creative concepts, and other work that agencies can produce more efficiently. The second concerns content intended to improve how companies appear in AI-generated search.

Both serve a purpose. Neither establishes whether an AI system understands a brand accurately.

A company might publish constantly, earn considerable press coverage, and maintain an active digital presence. Yet when someone asks an AI assistant to describe the business, explain its differentiation, or compare it with competitors, the answer can still be generic or inaccurate.

Publishing more won't necessarily fix that. The better starting point is to examine how the brand is already represented across public sources.

Consistency does not mean saying the same thing everywhere

Most agencies and marketing teams do not create confusion because their strategy is bad. They create it because normal communications decisions that work well in one channel do not always hold together across the full public record.

Campaign language changes. Positioning gets stretched to fit a launch. Social content is typically more playful than the website. Press materials emphasize the founding story while product pages outline features. These differences are not inherently problematic.

Consider a boutique hotel. Its travel coverage spotlights design and architecture, while a food publication showcases its restaurant. A regional newspaper might discuss the property's restoration of a historic building, and its own website highlights the guest experience.

Those stories are different, but they can reinforce the same underlying identity. Each adds another useful dimension to what the property offers and why someone might choose it.

The problem begins when the differences extend beyond expression into conflicting facts, categories, positioning, or claims.

A hotel whose website describes an intimate retreat may be presented by booking partners as a large business property. An agency might be promoting its new culinary direction while older profiles continue to identify the property primarily as a wellness destination.

Some of those descriptions may reflect legitimate changes. Others may be outdated or incomplete. Together, they create a potentially conflicting picture of the property.

This is why narrative governance matters. Agencies need to understand which elements of a brand's identity should remain stable and which can vary by audience, channel, or campaign.

Agency activity	What can vary	What should remain anchored
Campaign development	Creative language, imagery, emphasis, and cultural references can evolve.	The brand's fundamental positioning and product or service claims should remain accurate.
Earned media	Different publications can highlight different aspects of the company.	Core facts, expertise, category, and differentiation should remain supportable.
Social content	Voice, format, humor, and audience engagement can change by platform.	Creative expression should not unintentionally contradict the brand's central identity.
Website and product content	Pages can serve different audiences and stages of the decision journey.	Essential factual information and descriptions should be current, specific, and consistent.

Good communications strategy has always involved adapting a story to its audience. AI discovery introduces another reason to distinguish healthy variation from actual contradiction.

Where interpretive drift begins

Interpretive drift occurs when the brand meaning carried through AI-generated answers begins to diverge from its intended or current positioning. It can emerge from an accumulation of public information that no longer fits together.

Imagine a skincare company that originally built its reputation around natural ingredients but has since repositioned itself around clinical efficacy and dermatologist-developed formulations. Its website now reflects that shift. But older editorial coverage emphasizes clean beauty, retailer descriptions still highlight botanical ingredients, and influencer posts continue to associate the company with its original identity.

When an AI system is asked to describe the skincare line, some answers may accurately reflect its new direction. Others may revert to the earlier identity or produce a blended account that obscures the change.

That divergence is measurable. It does not establish exactly which source shaped an individual response, but it reveals where the brand's current meaning is not being carried through consistently.

For communications teams, a positioning refresh is not complete simply because the approved copy has been updated. Earlier versions of the story may continue to circulate across the wider public record.

Earned media can strengthen the evidence, not just the exposure

Earned media contributes more than visibility. A national feature may provide reach and prestige, while a trade publication documents a company's expertise. An interview can connect a founder's background to the business, and a specialist review may explain product qualities in greater detail.

For AI discovery, the semantic usefulness of a placement is worth considering alongside its traditional communications value:

• Does the article establish what the company does?

• Does it explain a meaningful point of differentiation?

• Does it connect expertise to the brand?

• Does it provide independent evidence for a claim the company makes about itself?

More coverage does not guarantee better AI visibility. Systems differ in what they retrieve, how they use sources, and how they construct answers. Even a strong placement may never surface in a particular response.

But agencies can improve the specificity and quality of the public evidence surrounding a client without assuming that every additional mention produces a corresponding improvement in AI performance.

Measurement should guide what agencies change

Before recommending changes to content or positioning, agencies need to understand how a company is already being represented.

A single AI response can reveal a problem, but it does not establish a pattern. A useful baseline examines how descriptions hold up across relevant prompts, systems, and decision contexts.

Consider three companies.

One is regularly recognized but described in generic terms that could apply to its competitors. Another has clear positioning on its website but rarely appears in unbranded recommendation requests. A third is accurately characterized in broad terms but repeatedly associated with outdated products or services.

Each observation points toward a different investigation.

The first may need stronger evidence of differentiation. The second requires an examination of category fit, competitive context, and where it disappears from consideration. The third may need to reconcile obsolete details across owned and third-party sources.

None of these situations automatically calls for publishing more.

That is the value of diagnosis before intervention. It helps distinguish a discovery problem from inaccurate representation or a weakness that emerges only during comparison and recommendation.

A baseline also identifies what is worth preserving. If particular descriptions, facts, and associations remain stable across models, changing them for novelty's sake may accomplish little. A fresh campaign does not require reinventing every element of the company's identity.

For agencies, the opportunity is to make recommendations based on observed behavior, then assess whether subsequent communications activity is associated with clearer and more consistent representation over time.

How agencies can talk about AI without overpromising

Agencies do not need to reinvent themselves as AI vendors.

The pressure to appear fluent in AI can encourage broad promises about automation, future-proofing, or helping clients win in AI search. Those claims may sound ambitious but often say little about the communications challenge in front of the client.

Public relations and brand teams already know how to shape narratives, clarify positioning, develop credible proof, and manage how companies present themselves across channels. Those capabilities remain valuable. What changes is the environment in which the results of that work are interpreted.

A more useful agency proposition is straightforward:

We help clients understand how their brands are represented in AI-generated discovery, identify where the public story is inconsistent or incomplete, and strengthen the communications signals that deserve attention.

That is an objective firms can explain without promising control over the systems themselves. Traditional communications measures still matter. Reach, media quality, share of voice, engagement, and reputation remain important indicators of market performance.

AI visibility adds another question: What happens to that accumulated work when a system assembles the available information into an answer?

Answering it builds on the strategic strengths agencies already possess: editorial judgment, evidence development, narrative stewardship, and an understanding of how reputation takes shape.

For agencies, the goal is not simply greater content volume. It is helping clients maintain an accurate, distinctive, and useful public account of their business, one that holds together even when someone else is telling the story.

Insights, Strategy and More

Most Recent in: Featured

Most Recent in: Strategy

Most Recent in: AI

Most Recent in: Mechanics

Founder’s Note