AI Trust Gap: 60% of Consumers Use AI, Only 13% Trust It

New research from Klaviyo surveying nearly 8,000 consumers globally reveals a striking paradox at the heart of AI-driven marketing: most consumers are already using AI to shop and discover products, but an overwhelming majority don’t fully trust it. For marketers deploying AI at scale, that gap isn’t an abstract concern — it’s the single most important variable determining whether your AI investments convert or collapse. Understanding the exact shape of that gap, and which consumer segments it affects most, is now a prerequisite for any serious AI marketing strategy.

What Happened

On March 13, 2026, MarTech published findings from Klaviyo’s AI Persona Research — a survey of nearly 8,000 consumers globally — that laid bare the complicated reality of AI’s role in consumer purchasing decisions. The headline numbers tell the core story: 60% of consumers use AI tools at least weekly, yet only 13% say they completely trust AI. That 47-point chasm between usage and trust is the challenge every marketer has to solve right now.

The study goes deeper than a single binary trust question. Klaviyo segmented consumers into four distinct personas based on how they interact with AI in their shopping and discovery journeys.

AI Enthusiasts (approximately 26% of consumers) sit at the top of both the usage and trust spectrum — though even “relatively high trust” among this group falls well short of complete trust. According to the MarTech report, 89% of AI Enthusiasts used AI while shopping in the past six months, and 43% purchased multiple previously unknown products based solely on AI recommendations. These are the consumers converting through AI-first discovery channels right now, and they represent the most important beachhead segment for AI-forward marketers to understand and protect.

AI Evaluators use AI frequently but apply active skepticism to every recommendation. They use AI as a research starting point, not a decision-engine. They’ll surface an AI-suggested product, open three browser tabs, scan Reddit reviews, read a comparison article, then decide. The conversion strategy for reaching this group is fundamentally different from Enthusiasts — and collapsing them into the same campaign flow is a common and costly mistake.

AI Skeptics are aware of AI tools but approach them with wariness in shopping contexts. They may use AI for productivity tasks like writing or summarizing, but they actively resist AI-driven recommendations in their purchase journey, preferring peer reviews, editorial content, and human sales guidance. Reaching this segment via AI-forward messaging can actually deepen distrust rather than reduce it.

AI Holdouts (approximately 21% of consumers) rarely or never use AI for shopping decisions at all, per the Klaviyo research. They’re not necessarily technophobic — they just haven’t found AI valuable enough in commerce contexts to change established behavior. For marketers, this segment represents a retention risk if legacy, non-AI channels are de-invested in the rush to build AI-first stacks.

Beyond personas, the Klaviyo data surfaces behavioral patterns that fundamentally reshape how we should think about AI interactions. 78% of consumers include emotional or personal context in their AI prompts at least sometimes — they’re not typing “blue running shoes,” they’re typing “I need running shoes for a half marathon in April, I have flat feet, I run mostly on pavement, and my budget is around $150.” That’s rich, layered intent data. And 30% of users routinely use eight or more words in their AI prompts. This is conversational commerce, not keyword commerce. Most marketing stacks are still optimized for keyword-era query patterns. The consumers have already moved past that.

The study also documents the direct purchase influence of AI: 41% of consumers surveyed purchased a product recommended by AI within the past six months, and 27% say AI introduced them to a product they later independently researched before buying. AI is no longer merely a discovery tool — it’s a significant commerce layer, operating at scale, in a context of incomplete and highly segmented trust.

Why This Matters

The 47-point gap between AI usage (60%) and full trust (13%) would be less significant if AI were just a peripheral touchpoint in the customer journey. It isn’t. With 41% of consumers purchasing AI-recommended products and another 27% using AI as a research trigger, AI is now touching the majority of online shopping journeys — and doing so in a context where consumers are simultaneously engaged and skeptical.

This creates a new kind of conversion challenge. Traditional funnel thinking assumes trust builds as consumers move through stages — awareness, consideration, intent, purchase. AI scrambles this model. A consumer can go from zero product awareness to AI-prompted purchase in a single session — but only if trust holds at each micro-moment. When it breaks at any point, so does the conversion. And unlike a standard funnel where you can re-enter a consumer at a lower stage, AI-trust failures tend to produce hard exits: the consumer turns off the AI recommendation surface and reverts to a channel they’re already comfortable with.

The specific segment carrying the highest risk is, counter-intuitively, the AI Enthusiast group. You might assume these are your easiest audience — they’re high-usage, higher-trust relative to others, and already converting via AI channels. But the Klaviyo data reveals a trap: 40% of AI Enthusiasts notice low-quality or generic AI-generated marketing content multiple times weekly. These are your power users, and they carry the highest content quality bar of any consumer segment. Deploy a generic AI-written email campaign at them and you will lose the most valuable part of your AI-channel audience the fastest.

For agencies running AI-assisted content pipelines for clients, this is a direct operational challenge. The efficiency gains from AI content generation are real, but the audience most likely to convert via AI channels is simultaneously the audience most capable of detecting and penalizing generic AI output. You cannot have both efficiency and quality without actively managing for it. The era of using AI to manufacture content volume without quality controls has a shrinking shelf life among the most commercially active consumer segments.

The four-persona framework also demands a fundamental rethink of audience segmentation strategy. A 34-year-old tech professional in Austin might be an AI Enthusiast; her colleague with an identical demographic profile might be an AI Skeptic. Traditional firmographic or demographic segmentation won’t capture this distinction. The implication for CRM and CDP strategy is significant: you need behavioral signals — prompt length, AI tool engagement rates, AI-assisted session patterns, recommendation click-through versus bypass rates — to segment audiences meaningfully by their actual AI relationship, not a proxy for it.

There is also a structural disruption brewing in search and discovery. According to MarTech’s conversation with Bazaarvoice CMO Doug Straton, nearly 20% of consumers are now using AI-powered LLMs for all or most of their search needs, and more than half of shoppers use AI as a research supplement at some point in their purchase process. Consumer adoption of AI search is outpacing brand implementation of AI-optimized discovery content — the audience has partially migrated to a channel that most brands aren’t actively managing or even measuring.

The Salesforce State of the Connected Customer (based on 16,000+ consumers and business buyers globally) reinforces the underlying stakes: 61% of customers believe AI advancements make trustworthiness more critical, not less. Increased AI capability doesn’t automatically generate increased consumer trust — it raises the expectation bar for every trust signal a brand emits. The more AI a brand deploys, the more carefully it needs to manage the conditions under which that AI operates and the transparency it provides to customers about how it works.

The Data

The numbers from Klaviyo’s research, alongside Salesforce’s State of the Connected Customer and Bazaarvoice’s commerce data, paint a granular picture of where consumer AI adoption and trust actually stand — and what the risk/opportunity split looks like for marketers building AI strategies right now.

Consumer AI Persona Breakdown (Klaviyo AI Persona Research, ~8,000 Global Consumers)

Source: MarTech — Klaviyo AI Persona Research, March 2026. AI Evaluator and Skeptic share percentages estimated from overall totals; Enthusiast (~26%) and Holdout (~21%) figures confirmed in source.

Key Consumer AI Metrics: Usage, Trust, and Behavior

Sources: MarTech/Klaviyo, Salesforce State of the Connected Customer, MarTech/Bazaarvoice

One number in that table deserves particular attention: the gap between 41% who purchased via AI recommendation and 13% who fully trust AI means a large population is converting despite incomplete trust — motivated by convenience, product relevance, or enough baseline confidence to take a chance. That fragile trust state is easy to destroy with a single misfire: a conspicuously generic email, a recommendation that obviously missed the customer’s context, or an unexplained automated decision. The brands that structurally protect that fragile trust will retain the conversion; the brands that erode it will see AI-channel performance deteriorate even as total consumer AI adoption continues to rise.

Real-World Use Cases

Use Case 1: E-Commerce Brand — Persona-Segmented Product Discovery Campaigns

Scenario: A mid-market DTC brand selling outdoor gear wants to increase revenue from AI-driven product discovery without alienating skeptical customer segments that currently represent significant lifetime value.

Implementation: Use behavioral signals — session source, on-site AI search query length, engagement with AI recommendation widgets versus editorial content, bypass rates on recommended products — to classify visitors into the four Klaviyo persona buckets. AI Enthusiasts receive AI-curated recommendation emails with product descriptions that lead with use-case specificity and contextual framing (“For the technical hiker covering 10+ miles in variable shoulder-season conditions…”), since this group instantly detects and disengages from template copy. AI Evaluators receive recommendation emails that surface the rationale explicitly (“Based on your pack weight preference and three-season use cases, paired with 847 verified reviews…”) alongside links to comparison articles and editorial roundups. AI Skeptics and Holdouts receive human-written editorial newsletters with no AI attribution framing — the product recommendation is positioned as editorial curation, not algorithmic output.

Expected Outcome: Higher open and click-through rates from Enthusiasts receiving quality-differentiated content tuned to their standards, and lower unsubscribe rates from Skeptics who are not being pushed into an interaction mode that triggers distrust. Given that 40% of AI Enthusiasts notice generic AI content multiple times per week per the Klaviyo research, quality-differentiated content has a clear, measurable ROI impact on the highest-converting AI-channel segment — and segmenting the audience protects the full revenue base.

Use Case 2: B2B SaaS Company — AEO Strategy to Capture LLM-First Buyers

Scenario: A B2B SaaS company in the marketing automation space is seeing organic search traffic plateau while inbound pipeline attribution grows murky. A segment of prospects is arriving via “other” or direct — internal investigation suggests LLM-generated recommendations may be contributing to untracked inbound.

Implementation: Conduct a prompt-to-answer audit: run 40–50 high-intent buyer queries through ChatGPT, Perplexity, Claude, and Google AI Overviews. Document which competitors appear, what content those LLMs cite, and what answer formats they favor. Build out structured FAQ pages, use-case explainers, and authoritative comparison guides that directly answer the most common buyer prompts — in the format LLMs prefer: concise, citation-ready, structured with clear headers, factually precise. Per MarTech’s Bazaarvoice interview, nearly 20% of consumers use LLMs for most of their search needs, and B2B buyers follow the same behavioral shift. Schema markup, authoritative third-party citations, and consistently updated technical documentation all improve LLM citation frequency.

Expected Outcome: Measurable inclusion in AI-generated answers for target use-case queries within 60–90 days of AEO content deployment. Tracked via branded mention monitoring across AI platforms and emerging AI-referral traffic patterns in analytics. Early AEO movers in B2B SaaS categories hold a structural advantage: LLMs anchor on early, well-documented authoritative sources, making first-mover content positions more durable than equivalent SEO positions that can be displaced by a single algorithm update.

Use Case 3: Digital Marketing Agency — AI Content Quality Audit System

Scenario: A digital marketing agency has deployed AI-assisted content workflows across 12 clients to handle email copy, product descriptions, and social content at scale. Engagement metrics on AI-produced campaigns are trending lower than human-written benchmarks, but the root cause isn’t clear — channel, timing, and content quality are all candidate explanations.

Implementation: Establish a formal “Enthusiast Bar” quality review gate for all AI-generated output before client delivery. Every AI-produced piece goes through a human editor specifically checking for: generic phrasing that reads as template output rather than brand voice, absence of contextual or emotional specificity, missing differentiation from competitor messaging, and factual thin-ness. Use the 78% emotional-context figure from Klaviyo’s research as a brief-writing trigger — every AI content brief must include emotional or situational inputs reflecting the customer’s actual context, not just product attributes and CTAs. Build a quality scoring rubric and track rejection rates by client to pinpoint where AI workflows need the most human reinforcement.

Expected Outcome: Recovery of engagement metrics toward human-written benchmarks within two campaign cycles. Reduced client churn among brands whose audiences skew AI Enthusiast — the segment that detects and disengages from generic AI content most quickly. Agencies that can credibly demonstrate quality-differentiated AI content, not just AI content velocity, build a defensible competitive position against lower-cost automation shops that treat AI output as ready-to-ship by default.

Use Case 4: Retail Brand — Transparency Layers on AI Recommendation Surfaces

Scenario: A fashion retailer has integrated AI-powered product recommendations into its site and app. Conversion rates from AI-recommended products are running 18% lower than human-curated editorial picks despite AI recommendations being more dynamically personalized. Post-purchase survey data suggests shoppers are uncertain whether to trust the AI picks.

Implementation: Add explicit trust scaffolding to every AI recommendation surface. Show the recommendation rationale: “Recommended because you saved three similar silhouettes in navy this month.” Include UGC photo reviews directly in the recommendation card rather than requiring a click-through. Display a human-in-the-loop signal: “Suggested by Style AI, reviewed by our editorial team.” Add a one-click feedback mechanism on every recommendation card so consumers can signal mismatches — improving the model and demonstrating responsiveness simultaneously. The Salesforce research is clear: 72% of customers want transparency about when they’re interacting with AI agents — this applies to recommendation surfaces as directly as it does to chatbots. Hiding AI origins doesn’t reduce distrust; it creates a slow accumulation of mistrust when consumers eventually notice the opacity.

Expected Outcome: Lift in AI recommendation conversion rate toward editorial pick benchmark over two to three months. Measurable reduction in “add to cart, then abandon” rates on AI-recommended products — a strong leading indicator of recommendation-trust mismatch. Longer-term: building the AI recommendation → purchase → satisfaction → reinforced trust loop that progressively converts AI Evaluators into AI Enthusiasts and retains Enthusiasts who might otherwise defect after a single poor experience.

Use Case 5: Financial Services Brand — AI Friction Monitoring and Recovery Protocol

Scenario: A financial services company uses AI for customer service routing, fraud detection, and personalized offer delivery. NPS scores have dropped 11 points over two quarters despite overall satisfaction on human-handled interactions holding steady — the decline correlates precisely with the expansion of AI touchpoints across the customer journey.

Implementation: Deploy a structured AI friction monitoring protocol based on the framework outlined in MarTech’s “When AI Decisions Create Customer Friction”: establish mandatory human review for every high-impact AI decision (account flags, declined transactions, eligibility determinations), build customer-facing plain-language explanations for every automated action, conduct quarterly model drift audits, and create escalation paths that don’t require customers to restart their explanation from scratch when they reach a human agent. Track NPS by interaction type — human-handled versus AI-handled — to quantify the trust delta attributable to AI specifically. The article frames the business case directly: “friction erodes trust, trust influences renewal and renewal drives revenue.” Translate that chain into a financial model so AI experience quality has a revenue owner at the leadership level, not just a product manager.

Expected Outcome: NPS recovery toward pre-AI-expansion baseline within two quarters of implementing the monitoring and intervention protocol. Reduction in customer service escalation volume caused by unexplained automated decisions — a cost savings that frequently more than offsets the investment in human review layers for high-impact decisions. Most importantly: a repeatable, instrumented process for detecting AI-generated trust damage before it reaches the churn threshold rather than discovering it retrospectively in renewal data.

The Bigger Picture

The Klaviyo data doesn’t exist in isolation. It is one signal in a larger pattern reshaping the relationship between brand infrastructure, technology deployment, and consumer psychology — and the direction of that pattern has significant implications for how marketing strategy gets built over the next 18 months.

The Salesforce State of the Connected Customer, drawing on 16,000+ global respondents, documents a remarkable shift: 73% of customers now feel companies treat them as individuals, up from just 39% in 2023. That 34-point jump in two years is almost entirely attributable to AI-powered personalization reaching commercial scale. But the same survey shows 64% of those same customers think companies handle their data recklessly, and 71% say they are increasingly protective of their personal data. Consumers feel more personally known than ever, and more personally exposed than ever. AI is the engine of both experiences simultaneously.

This dual reality — impressive personalization alongside alarming data exposure — is what makes the trust gap structurally difficult to close. It is not a PR problem, a messaging problem, or a UX polish problem. It is a system design problem. The data collection, behavioral modeling, and cross-channel identity resolution that enable great personalization are the same capabilities consumers distrust when they think about their data being handled. The path to closing the trust gap runs through radical operational transparency, not just better marketing copy.

The shift in search and discovery behavior compounds the challenge on a different axis. According to MarTech’s conversation with Bazaarvoice CMO Doug Straton, more than half of shoppers now use AI as a research supplement, and nearly 20% have moved AI-powered LLMs to their primary or exclusive search channel. Consumer adoption of AI search is outrunning brand implementation of AI-optimized discovery content — the audience has partially migrated to a channel that most brands don’t actively manage or have analytics visibility into. AEO (Answer Engine Optimization) is emerging as a formal discipline precisely because traditional keyword SEO captures nothing from the LLM-search population. Unlike Google, which provides a ranked list that preserves click-through opportunities for multiple brands, LLM-generated answers typically surface one or a narrow set of recommendations with high consumer confidence already attached. The brands appearing in those answers hold a disproportionate, compounding advantage.

Meanwhile, AI friction events — misclassified fraud flags, inappropriate hyper-personalization, opaque automated decisions, recommendations that conspicuously miss context — are actively and quietly destroying trust in ways that are hard to diagnose and slow to recover from. As the MarTech friction article documents, the downstream revenue consequence of AI trust damage follows a reliable chain: friction erodes trust, trust influences renewal, renewal drives revenue. The brands that instrument this chain — that can see AI friction events in real time and intervene before they become churn — will manage AI-era customer relationships fundamentally better than those treating AI deployment as a configure-and-forget infrastructure decision.

The trajectory of the industry is unambiguous: AI usage among consumers will continue to increase, and AI’s role in the purchase journey will deepen. But trust will remain stubbornly low — and increasingly consequential — unless brands actively invest in trust infrastructure as a core requirement, not a post-launch enhancement. The Klaviyo research makes this actionable by defining exactly who the trust-gap segments are and what their distinct content, channel, and transparency expectations look like.

What Smart Marketers Should Do Now

1. Build AI-Affinity Segmentation into Your CRM and CDP Before Q3 2026

Stop segmenting audiences exclusively on demographics, purchase history, or firmographic data. Start capturing behavioral signals that indicate AI affinity: which customers engage with AI-powered features on your site or app, how long their AI search queries are, whether they convert from AI recommendations or routinely bypass them, how often they interact with an AI chatbot versus escalate to a human. The four Klaviyo personas are not hypothetical archetypes — they represent real audience segments with meaningfully different content expectations, channel preferences, and conversion triggers. If your CRM cannot distinguish between an AI Enthusiast and an AI Holdout, you are sending the same message to an audience with radically different trust states, and you’re leaving conversion on the table at both ends. Map behavioral proxies to persona classification and update your segmentation logic this quarter, before the next major campaign build.

2. Implement the Enthusiast Bar as a Standing AI Content Quality Gate

AI Enthusiasts — the segment with the highest AI-channel conversion rate — notice generic AI-generated content multiple times per week, per the Klaviyo research. Before deploying any AI-assisted campaign asset — email copy, product descriptions, ad creative, landing page text — run it through what I call the Enthusiast Bar: would someone who uses AI daily find this copy obviously templated? Does it carry the contextual specificity that comes from rich brief inputs, or does it read as a median of training data outputs? Does it reflect a genuine brand voice or a generic approximation of one? The 40% weekly detection rate among Enthusiasts means you have roughly a coin-flip chance of alienating your best AI-channel prospects with every unreviewed AI asset that reaches them. Add a human editing gate specifically tuned to generic-content detection — not just fact-checking or legal review — as a non-negotiable step in the AI content pipeline.

3. Launch an AEO Content Program Targeting Your Top 30 Buyer Queries

With nearly 20% of consumers already using LLMs as their primary search mechanism per MarTech’s Bazaarvoice interview, every week without an Answer Engine Optimization strategy is a week competitors may be capturing LLM-first buyers you’re invisible to. The tactical starting point is a prompt audit: identify the 30 questions your ideal buyer is most likely to ask an LLM when searching for a solution to the problem your product addresses. Run those prompts through ChatGPT, Perplexity, Claude, and Google AI Overviews. Document who appears in the answers and what content is being cited. If you’re not represented, build the content that fills those gaps — structured FAQ pages, detailed use-case explainers, authoritative comparison guides. LLMs favor high-authority, frequently-cited, clearly-structured sources, so traditional long-form SEO content structure often translates; what’s missing is deliberate optimization for conversational AI search query patterns.

4. Add Transparency Layers to Every AI-Powered Customer Touchpoint

The Salesforce data is explicit: 72% of consumers want to know when they are interacting with an AI agent. This principle extends beyond chatbot disclosure to every AI-driven recommendation widget, automated pricing decision, personalized content surface, and eligibility determination your systems produce. Implement “Why am I seeing this?” functionality on all recommendation modules. Label AI-powered features directly and matter-of-factly. When automated decisions affect customer experience — service routing, pricing tiers, content filtering, access decisions — provide explanations that a non-technical customer can act on. Transparency does not erode trust; opacity does. The brands that have most successfully narrowed the AI trust gap are the ones that surfaced their AI usage proactively and made the AI’s reasoning legible, rather than concealing automation behind a polished interface and hoping customers don’t notice.

5. Instrument AI Friction Points Before They Become Churn Events

The causal chain is direct and documented: AI friction → trust erosion → reduced renewal or repurchase. Start by mapping every AI touchpoint in your customer journey and identifying the ones with the highest friction potential — any that result in a denied action, an unexplained decision, or a recommendation that conspicuously misread the customer’s context. For each high-risk touchpoint, establish monitoring metrics: escalation rates, NPS delta by interaction type, churn correlation coefficients. Build human escalation paths that don’t require customers to restart their explanation. The MarTech friction framework recommends treating efficiency and customer experience as equal priorities in AI deployment — in operational terms, this means friction instrumentation is a launch requirement for any AI decision system, not a post-launch optimization. Do not ship an AI decision layer without a measurement framework for the trust damage it is capable of causing at scale.

What to Watch Next

AI Persona Segmentation as a Native CRM Feature (Q3–Q4 2026): Expect CRM and CDP vendors to begin surfacing AI-affinity scores as a native segmentation dimension in the second half of 2026. Klaviyo, having commissioned the foundational research on AI consumer personas, is well-positioned to build persona classification signals directly into its platform. Watch for comparable feature launches from Salesforce Marketing Cloud, HubSpot, and Braze. When AI affinity data is natively available in your existing marketing stack, the persona-segmented strategy outlined in this post scales without requiring custom behavioral modeling — early adopters of persona-aware segmentation will have a measurable head start when these tools ship.

LLM Advertising Formats Going Mainstream (Q2–Q3 2026): The Bazaarvoice / MarTech conversation flagged emerging ChatGPT advertising formats as a development to monitor. OpenAI has publicly moved toward monetizing AI-generated answer surfaces through sponsored placements, and Perplexity has launched its own advertising product. If 20% of consumers are already using LLMs as their primary search mechanism, paid placement within LLM-generated answers will become a significant paid acquisition channel within 12–18 months — and it currently has no established auction dynamics, no standard CPM benchmarks, and no established creative best practices. First movers will define category norms and capture early learning curves. Track product announcements from OpenAI, Perplexity, and Google AI Overview ad teams through mid-2026.

Regulatory AI Transparency Requirements (H2 2026 and Beyond): The Salesforce finding that 72% of consumers want disclosure when interacting with AI agents will eventually become a regulatory floor, not just a consumer preference. EU AI Act implementation timelines are advancing through 2026, and US state-level AI disclosure requirements are working through multiple legislative sessions simultaneously. Marketers should build AI transparency and disclosure practices now, treating regulatory requirements as already in effect rather than scrambling for compliance when mandates arrive.

AI Trust Score as a Marketing KPI: Watch for the emergence of “AI trust index” or “AI interaction confidence score” as a trackable, benchmarkable marketing metric — analogous to Net Promoter Score but specifically measuring consumer trust in a brand’s AI-powered interactions. As Salesforce research shows, 61% of consumers believe AI advancement makes trustworthiness more critical — the brands that develop reliable instruments for measuring AI-specific trust perception in Q2 and Q3 2026 will carry a meaningful strategic intelligence advantage into 2027 planning cycles.

Bottom Line

The Klaviyo research spanning nearly 8,000 global consumers is a data-backed mandate: AI adoption is real, accelerating, and commercially consequential, but it is running far ahead of consumer trust, and the 47-point gap between 60% weekly AI usage and 13% complete trust is precisely where marketing dollars either convert or evaporate. The consumers most actively converting via AI channels — the AI Enthusiast segment — are simultaneously the most sophisticated detectors of low-quality AI marketing output, which means the efficiency gains of AI content generation carry a concrete quality risk requiring active, systematic management. Salesforce’s broader research reinforces the structural challenge: as AI capability scales, consumer expectations for trustworthiness, transparency, and responsible data handling scale in parallel — not inversely. The marketers who close the AI trust gap first through persona-aware segmentation, quality-differentiated content pipelines, AEO strategy, and friction instrumentation will compound an advantage in the highest-value consumer relationship in digital commerce. The rest will spend the next two years debugging why their AI stack isn’t converting at the rates they projected.