The AI Trust Gap

What 55 Studies (and One Very Creepy Filter Ad) Reveal About Customer Engagement 

• Trust, not features, decides whether AI-driven personalization turns into engagement.
• Competence without disclosure reads as creepy, and creepiness makes travelers defer the booking.
• Only five of the 55 reviewed studies looked at the organizations deploying AI. The customer side is well mapped; the operator side is not.
• In hospitality specifically, personality may matter as much as precision.

A consumer recently described standing in the hardware aisle at Walmart, talking out loud about a water filter. The next day, an ad for that exact filter, same brand, landed in their inbox.

Coincidence? Sure. Probably. Almost certainly.

But it still felt like something was listening, and that feeling now has its own academic literature. A recent paper in Psychology & Marketing gives that feeling a clinical name: "creepiness." The paper argues it's one of the most underexamined variables in whether AI-driven marketing works (Petrova 2026).

That tension (AI powerful enough to know exactly what you want, and unsettling for exactly that reason) is the real story hiding under all the hype about AI and customer engagement. A systematic review that I recently came across gives the clearest picture yet of where that story stands.

The Review: 55 Studies, One Industry Left Behind

A team of researchers in Poland worked through nearly 20,000 academic articles on AI, customer engagement, and social media marketing, then screened that mountain down, using PRISMA (the standard protocol for systematic reviews), to 55 studies that actually ran the numbers, all of them peer-reviewed (Żyminkowska & Zachurzok-Srebrny 2025).

Their focus: tourism and hospitality, an industry the authors flag as having unusually low digital readiness next to banking or finance, even though it's stuffed with more customer-facing AI than almost any other.

That's the paradox baked into the whole field right now.

Hotels, airlines, and travel platforms are pouring money into chatbots, voice assistants, and recommendation engines. At the same time, the underlying service experience, especially at the moment of delivery, still comes down to whether a human remembered to fix the ice machine. Deloitte's enterprise AI survey series keeps finding leaders planning to raise, not cut, generative AI budgets. What does the research show?

Four Findings the Research Agrees On

Strip away the buzzwords, and the 55 studies cluster around four consistent findings:

1. AI reshapes the front door of customer service. Chatbots, voice assistants, and generative tools like ChatGPT are consistently linked to smoother, faster interactions. But the studies keep circling back to the softer stuff, ease of use and emotional response (yes, researchers really do measure "awe"), as the real driver of whether that smoothness converts into loyalty, not the technology itself.
2. AI is becoming a relationship, not just a tool. Some of the most interesting work looks at things like virtual influencers, gamified chatbots, and Metaverse experiences, where the AI isn't just answering questions. It's becoming part of how people express identity and connect with a brand.
3. Trust is the actual currency. Study after study in the review points to the same variable: trust, not features, is what determines whether AI-driven personalization turns into engagement or gets ignored.
4. Readiness varies wildly by culture and market. Studies from Vietnam, Ghana, and the Philippines, all inside the review's 55, show that AI engagement strategies that work in the U.S. or China don't automatically translate. Cultural norms, digital infrastructure, and plain old familiarity all shape whether AI-driven marketing lands or just sits there.

Trust shows up so often, in fact, that it's worth pulling apart on its own.

Why "More Personalized" Doesn't Automatically Mean "More Trusted"

Here's where the creepiness research becomes directly relevant. A separate 2026 study out of Texas A&M, published in the International Journal of Hospitality Management, tested how travelers respond to chatbots on hotel booking platforms and found that when a chatbot came across as unbelievable, inaccurate, or incompetent, users experienced it as creepy, and that creepiness pushed them to disengage or defer the booking (Akhtar et al. 2026).

The lead researcher's conclusion is blunt: "Competence without transparency doesn't solve the problem." 

Being good at guessing what you want isn't enough. The bot also must be honest about how it knew.

Marketing researchers have a name for this and have documented it for years: “the personalization-privacy paradox.” People say they want tailored experiences, and they simultaneously feel uneasy about the data collection that makes tailoring possible (Petrova 2026). AI doesn't resolve that tension. If anything, it sharpens it, because AI personalization is faster, more detailed, and far less visible than the junk-mail era it replaced.

So, what builds trust instead of eroding it? A few consistent threads run through the peer-reviewed literature:

• Disclosure, not just competence. Clear, upfront communication about how an AI system is drawing its recommendations reduces the "creepiness" effect, even when the underlying personalization stays the same (Akhtar et al. 2026).
• Warmth alongside capability. Research on digital voice assistants found that people respond with loyalty when the assistant comes across as both human-like and competent. One without the other underperforms (Maduku et al. 2024).
• Relational payoff, not just efficiency. A study on smart voice assistants found that pleasure and satisfaction drove repeat purchases and referrals. Researchers call this "relational cohesion." Everyone else calls it liking something enough to tell a friend about it (Hernández-Ortega et al. 2022).
• Trust must be earned before usage even starts. Work on chatbot adoption found that "initial trust" drives both the intention to use a chatbot and engagement with it, and that this trust is built by how compatible and easy a tool feels on first contact, not by how well people expect it to perform (Mostafa & Kasamani 2022).

One more wrinkle: some hospitality-specific research suggests that what researchers call anthropomorphism (i.e., making an AI cuter, more approachable, less obviously built to sell you something) increases customers' willingness to cooperate with it and co-create value, alongside its perceived service capacity and novelty (Tian et al. 2025). In other words, in travel and hospitality specifically, personality may matter as much as precision.

More Studies Needed to Evaluate How Well Organizations Deploy AI

Here's the part of the systematic review that should worry (or excite) anyone running a hospitality or travel brand: of the 55 studies reviewed, only five (about 9%) looked at the companies themselves. This included their AI readiness, internal capabilities, and how managers and employees adapt to the tools. The overwhelming majority studied consumers.

We have a rich understanding of how customers feel about AI. We know remarkably little, empirically, about whether the organizations deploying it are any good at it.

That's not a small gap. It means most AI strategy advice floating around right now, including plenty aimed at hotels, resorts, casinos, and venues, is built on customer-side research and then pointed at the business side with little direct evidence. The review's authors call this out explicitly as one of the field's most urgent open questions.

What to Consider If You're Building or Buying AI-Driven Engagement Tools

A few practical takeaways, stripped of the hype:

• Legibility over maximalism. Don't chase personalization for its own sake; chase the kind where the customer can tell, without having to ask, roughly how the system knows what it knows.
• Trust is a design requirement. Build it in from the start, not as an afterthought bolted onto a chatbot's tone of voice.
• Early adopters aren't your market. Watch the gap between what delights them and what holds up across a broader, less tech-comfortable customer base. The readiness gap between markets, and between individual customers, is real and well documented.
• Interrogate vendor claims. If you're evaluating an AI engagement pitch, ask what they've validated on the operational and organizational side, not just the customer-experience side. That's still mostly a research vacuum, which means a lot of vendor confidence is currently ahead of the evidence.

What I'm Watching Next

The next wave of research worth tracking won't be about whether AI can personalize a travel recommendation. That question is largely answered. It's going to be about whether the organizations deploying these systems can build the internal readiness and the disclosure habits to make that personalization feel like service instead of surveillance.

That's a harder problem than picking the right chatbot vendor. And right now, it's barely been studied.

Sources

• Akhtar, N., Gupta, P., Taheri, B., & Alghafes, R. (2026). Unveiling the role of chatbot conversational attributes of hospitality booking platforms in developing users' creepiness and paradoxical behaviors. International Journal of Hospitality Management, 133, 104428. https://doi.org/10.1016/j.ijhm.2025.104428

• Hernández-Ortega, B., Aldas-Manzano, J., & Ferreira, I. (2022). Relational cohesion between users and smart voice assistants. Journal of Services Marketing, 36(5), 725–740. https://doi.org/10.1108/JSM-07-2020-0286

• Maduku, D. K., et al. (2024). Do AI-powered digital assistants influence customer emotions, engagement and loyalty? An empirical investigation. Asia Pacific Journal of Marketing and Logistics, 36(11), 2849–2868. https://doi.org/10.1108/APJML-09-2023-0935

• Mostafa, R. B., & Kasamani, T. (2022). Antecedents and consequences of chatbot initial trust. European Journal of Marketing, 56(6), 1748–1771. https://doi.org/10.1108/EJM-02-2020-0084

• Petrova, A., Malär, L., Hoyer, W. D., & Krohmer, H. (2026). The phenomenon of creepiness in a digital marketing world. Psychology & Marketing, 43, 834–851. https://doi.org/10.1002/mar.70089

• Tian, M., Yan, J., & Li, X. (2025). Anthropomorphism of service-oriented AI and customers' propensity for value co-creation. Marketing Intelligence & Planning, 43(1), 50–72. https://doi.org/10.1108/MIP-08-2023-0388

• Żyminkowska, K., & Zachurzok-Srebrny, E. (2025). The role of artificial intelligence in customer engagement and social media marketing—Implications from a systematic review for the tourism and hospitality sectors. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), 184. https://doi.org/10.3390/jtaer20030184