With apps like Perplexity receiving “a lot of inbound interest from the CMOs of large brands,” according to its head of advertising, and OpenAI quietly retooling ChatGPT into a shoppable experience, conversational marketing is evolving fast.

But while the promise is compelling, the reality is still taking shape, writes Esme Robinson, director of platform solutions at Epsilon.

Formats are shifting, expectations are high, and the right move depends on your readiness to test, learn and adapt.

Here are three key factors to weigh up before investing.

Intent is not the same as impact

AI chat apps have clear advantages. Their ability to surface content in context means recommendations can feel more relevant and less disruptive.

Perplexity’s ‘Sponsored Question’ format, for example, appears as a paid prompt within the chat flow, offering up suggestions tied to a user’s original query.

It’s a form of intent-led targeting that traditional search formats often struggle to replicate, especially when it comes to delivering a single, context-aware recommendation that aligns with what the user actually needs.

But prompting interest is only half the equation. Marketers are ultimately judged on outcomes like clicks, conversions and sales, not just engagement.

That’s where many chat formats still fall short.

To address this, some platforms are introducing merchant programmes that let brands surface real-time product, pricing and availability data. It’s a promising step, but for now still limited in scope.

Subscribe to Marketing Beat for free

Sign up here to get the latest agency-related news sent straight to your inbox each morning

You can’t optimise what you can’t see

A major challenge is connected identity. Many AI chat apps rely on temporary session data or limited signals, making it hard to build consistent, pseudonymised user profiles.

Without that foundation, brands can’t deliver the kind of tailored, people-based experiences that truly harmonise messages across devices and channels to build brand equity as well as increasing return on investment.

Measurement suffers for the same reasons. Fragmented data makes it harder to track performance or understand how spend links to outcomes.

Marketers are used to having clear attribution paths and closed-loop reporting, standards not yet embedded in most chat environments.

Until connected identity and measurement catch up, it’s difficult to assess what’s really working.

Others still set the standard

Elsewhere on the web, publishers and retailers have spent years building logged-in environments and direct customer relationships.

That means access to high-quality, first-party data and the ability to activate it with precision. For advertisers, it translates to better targeting, more control and clearer value.

Retailers, in particular, are well-positioned. With persistent customer accounts, they can power chat experiences that feel personal and genuinely useful.

A brand recommendation in a retailer-owned chat app can feel like a natural extension of the shopping journey, not a random interruption. And because these are environments brands already know and trust, there’s less guesswork involved.

By contrast, when chat apps rely too heavily on automation without the right identity layer, interactions can feel impersonal or intrusive. If users sense they’re being nudged by an algorithm rather than a conversation, engagement drops and conversions often go with it.

For AI chat app advertising to earn a place on the media plan, it needs to offer more than novelty.

Connected identity, integration and transparency will be the difference between a clever prompt and a meaningful customer connection.

For now, these platforms remain a valuable testing ground, but real success will depend on how well brands can connect the dots across data, journey and outcome.