Zappi: Defining the Future of Advertising in AI Assistants: Insights for Enterprise Leaders

The integration of advertising into Artificial Intelligence (AI) assistants is no longer a theoretical discussion; it is a present reality. As OpenAI rolls out ads in ChatGPT and major events like the Super Bowl feature AI commercials, the industry stands at a critical juncture. AI assistants are crossing Geoffrey Moore’s “chasm” into early majority adoption, and the norms established now will dictate consumer expectations for years to come. This report, Lines in the Sand, by Zappi Research, surveyed early adopters—the leading edge of consumer behavior—to provide essential guidance for senior marketing and CX leaders. Their findings offer concrete insights into where consumers will and will not accept advertising within these emerging platforms.

Consumer Perceptions: Trust, Annoyance, and Age Dynamics

Early adopters of AI assistants exhibit nuanced perceptions of advertising within these tools, shaped by inherited trust, age-dependent expectations, and distinct annoyance thresholds that differ from traditional digital channels.

Inherited Trust and Age-Dependent Nuances

AI assistant advertisements largely inherit the trust levels associated with established search advertising platforms. Eighty-two percent of respondents found AI assistant ads at least as trustworthy as Google Search ads, with 33% stating they would be more trustworthy. This indicates that the medium does not begin with a trust deficit. However, this trust is not uniform across all demographics. Younger users (18-34 years old) are three times more likely than older users (55-75 years old) to consider AI ads more trustworthy. Forty-two percent of 18-34-year-olds extend more trust to AI ads than to Google Search ads, suggesting they perceive AI output as more authoritative, akin to a recommendation rather than a mere placement. This distinction carries significant implications for creative strategy and regulatory compliance.

What this means: Enterprises can leverage the initial halo of trust, but must recognize the varying levels of implicit endorsement among user segments. Overly aggressive or personalized ad placements could quickly erode this foundational trust, particularly for older demographics who are less accustomed to conversational interfaces.

Annoyance Thresholds and Contextual Sensitivity

When it comes to annoyance, AI ads are perceived similarly to social media advertisements. Two-thirds of users reported that AI ads would feel no more annoying than ads on platforms such as Instagram, TikTok, or Facebook. This suggests that users are not applying a fundamentally higher standard to AI advertising regarding mere presence. The primary differentiator for success is not the volume of ads, but rather their context. While social media is often a high-volume advertising channel, the emotional zone for AI assistant ads aligns, indicating that over-engineering to prevent annoyance is a misallocation of resources. Instead, focus must be placed on ensuring ads appear in appropriate contexts.

Summary: Marketers should calibrate frequency caps and creative quality for AI assistant ads to align with social media norms, rather than attempting to eliminate all potential annoyance. The critical challenge is to meticulously manage where ads appear.

What to do:

• Leverage inherited trust: Begin with campaigns that capitalize on the baseline trust users have for AI assistants, similar to search.
• Segment creative strategy: Develop ad creative that accounts for age-dependent trust perceptions, particularly for younger audiences who may view AI ads as recommendations.
• Calibrate frequency: Apply social media-comparable frequency caps to manage user exposure, avoiding excessive ad volume.

What to avoid:

• Budgeting against a trust deficit: Do not assume AI advertising starts with inherent user distrust; instead, focus on converting inherited trust into brand equity.
• Implied endorsement: Avoid creative that could be misconstrued as an AI assistant’s direct recommendation, especially with younger users, to mitigate regulatory and reputational risks.
• Over-engineering for annoyance reduction: Focus energy on contextual relevance rather than striving for an ad-free experience, which is not users’ primary concern outside of sensitive contexts.

Strategic Imperatives: Mitigating Churn, Defining Contextual Guardrails, and Adapting to Personalization Shifts

The introduction of advertising into AI assistants presents strategic challenges related to user retention, brand safety, and the efficacy of targeting mechanisms. Enterprise leaders must plan for potential churn, establish stringent contextual exclusions, and adapt to evolving personalization capabilities.

Churn Risk: A Real and Concentrated Threat

A significant portion of early adopters, 40%, indicated they would either pay for an ad-free version or switch to a different AI assistant if ads were introduced. This figure is a ceiling, not a forecast, but it underscores a substantial behavioral risk. Among paying users, this inclination is even stronger, with 35% expressing a willingness to escape ads, more than double the 17% among free users. This suggests that the most valuable segment—premium subscribers—is also the most susceptible to churn due to advertising. The Super Bowl LX demonstrated this behaviorally, with Anthropic’s Claude, positioned as an “ad-free” alternative, experiencing an 11% jump in daily active users post-event, compared to Open AI’s ChatGPT (2.7%) and Google’s Gemini (1.4%).

What this means: Enterprises must plan for an actual churn rate of 15-25% as a working assumption. Reaching paying users through AI assistant inventory will likely compress effective reach quickly, emphasizing the need for early engagement.

Contextual Sensitivity: The Non-Negotiable “No-Fly Zones”

Users draw hard lines regarding ad placement within specific contexts, treating these as non-negotiable “no-fly zones.” Five distinct contexts account for over 70% of consumer concern, regardless of ad volume:

1. Confidential professional work (17%): Examples include internal strategy discussions in a B2B SaaS environment.
2. Mental health or struggles (16%): Such as users seeking support for stress or anxiety from a healthcare AI assistant.
3. Medical information or symptoms (14%): For instance, queries about prescription drug interactions or disease symptoms.
4. Children and students (13%): Educational use cases or content directed at minors.
5. Financial or legal matters (11%): Queries about investment advice, tax preparation, or legal documentation. Paying users are particularly sensitive in these high-stakes contexts, demonstrating higher rates of unacceptability compared to free users.

Summary: Contextual exclusions are paramount. Platforms must provide transparent, auditable context detection. Advertisers who proactively define and enforce these boundaries will set the standard for the industry.

Personalization: The Erosion of Behavioral Targeting

A significant challenge arises from user privacy preferences: half of all users indicated they would either turn off memory/personalization features or specifically opt out of ad personalization once ads appear in AI assistants. Specifically, 27% would disable all memory and personalization, while an additional 25% would maintain memory but opt out of ad personalization. Fifteen percent already have these features off. This response is a rational privacy decision, not hostility. The act of monetizing personalization fundamentally shrinks the available data for behavioral targeting, leading to a “noisier, sparser signal” for campaign measurement. Paying users are particularly surgical in this regard, with 31% specifically opting out of ad personalization.

Summary: Campaign measurement for AI assistant advertising will operate on a lower behavioral floor than current social and search benchmarks. Advertisers must rapidly build contextual targeting capabilities.

What to do:

• Plan for churn: Budget for a 15-25% churn rate among AI assistant users upon ad introduction. Factor this into reach and cost-per-acquisition models.
• Prioritize context exclusions: Treat the five sensitive contexts (confidential professional work, mental health, medical, children/students, financial/legal matters) as non-negotiable brand safety exclusions.
• Demand auditable detection: Push AI assistant platforms for transparent, auditable context detection tooling before committing ad inventory.
• Invest in contextual targeting: Develop and implement semantic, intent-based, and query-derived contextual targeting strategies. Example: A telecom company could target users asking about home internet speeds with ads for fiber optic services, regardless of personal browsing history.
• Monitor brand sentiment: Complement traditional ad performance metrics with sentiment analysis and direct feedback channels to detect early signs of user dissatisfaction.

What to avoid:

• Delaying market entry: The market is forgiving now, but the window to shape consumer expectations is closing. Delaying allows competitors to define the norms.
• Ignoring paying user sensitivity: Underestimating the churn risk among premium subscribers will severely impact effective reach and platform viability.
• Over-reliance on personalization: Do not assume current levels of personalization data will persist. Behavioral targeting alone will be insufficient.
• Compromising on brand safety: Allowing ads in sensitive contexts will lead to significant reputational damage and user exodus.

Operational Frameworks for AI Advertising Success

As AI advertising matures, establishing robust operational frameworks will be critical for managing risk, ensuring brand safety, and achieving measurable outcomes. This includes defining clear governance, adapting measurement strategies, and prioritizing immediate actions.

Operating Model and Roles

The introduction of AI assistant advertising necessitates an evolution in marketing and CX operating models. This involves defining new roles and responsibilities, establishing guardrails, and setting performance thresholds.

• Governance and Risk Controls: Implement a dedicated AI advertising governance committee, comprising legal, marketing, CX, and data privacy leads. This committee must establish explicit context exclusion policies, identifying “no-fly zones” for ad placement (e.g., healthcare provider AI assistant interactions for patient data; financial advisor AI assistants discussing sensitive portfolio details). All ad placements must undergo an approval process that includes checks for these exclusions, potentially using automated content analysis tools with human oversight. Thresholds for acceptable risk should be clearly defined, with zero tolerance for ads in critical sensitive areas.
• Data Readiness and Integration: Shift emphasis from individual user profiles to contextual data signals. Marketing teams must collaborate with data engineering to develop pipelines that extract and categorize semantic intent from AI assistant interactions without relying on PII or granular behavioral histories. This involves integrating with content classification systems and developing robust function calling capabilities to determine the nature of user queries (e.g., “planning a business trip” vs. “discussing personal finances”).
• Roles: A “Contextual Strategist” role might emerge, focusing on developing sophisticated semantic targeting logic. “AI Brand Safety Officers” would be responsible for configuring and auditing exclusion lists, working closely with platform providers to ensure compliance. CX teams must establish direct feedback loops within the AI assistant interface (e.g., “Was this ad relevant?”).

Measurement and Expected Outcomes

Traditional ad performance metrics will require re-evaluation in the AI assistant environment.

• Metrics: Beyond standard metrics like click-through rates (CTR) and conversion rates (CVR), focus on:
• Contextual Relevance Score: An internal metric evaluating how well an ad aligns with the immediate query context.
• Ad Recall and Brand Lift: Measuring changes in brand favorability and recall through post-exposure surveys.
• Churn Rate related to Ads: Monitoring user churn from AI assistant services, with attribution analysis for ad-related dissatisfaction. Aim for churn rates below 15-25%.
• Complaint Rate: Tracking explicit user complaints about ads, aiming for rates below 0.5% for high-stakes contexts.
• Customer Effort Score (CES): Evaluating the effort users expend to dismiss or ignore irrelevant ads.
• Expected Ranges: Initial campaigns should prioritize learning and optimization. A high Contextual Relevance Score (e.g., >70%) should be an early target. Maintaining or improving CSAT/NPS scores for the AI assistant is paramount, even with ads present.

Immediate Priorities (First 90 Days)

1. Secure Early Inventory: Negotiate preferential access to AI assistant ad inventory. The market is forgiving now, but the window to shape consumer expectations is closing rapidly. Early entrants will establish the reference points for future advertising.
2. Define Non-Negotiable Context Exclusions: Explicitly document the five sensitive contexts identified by Zappi Research (confidential professional work, mental health, medical, children/students, financial/legal matters) as non-negotiable brand safety exclusions. Push AI assistant platforms for transparent, auditable tooling to detect and enforce these exclusions. The absence of clear exclusion capabilities should be a deal-breaker.
3. Initiate Contextual Targeting Development: Start building semantic, intent-based, and query-derived contextual targeting strategies. Since personalization data will likely diminish, parallel development of robust contextual targeting is essential. This proactive approach will mitigate risks associated with a “noisier, sparser signal” for campaign measurement.

What “Good” Looks Like:

• Flawless Contextual Adherence: No ads appearing in any of the defined sensitive “no-fly zones.”
• Positive User Sentiment: Ad feedback (explicit or implicit) consistently points to relevance rather than intrusiveness, maintaining high CES/CSAT for the AI assistant.
• Measurable Brand Uplift: Campaigns demonstrate clear, attributable increases in brand perception metrics (e.g., brand awareness, favorability, purchase intent) without significant churn from the AI assistant platform.
• Seamless Integration: Ad delivery mechanisms are integrated with existing brand safety tools and customer feedback systems (CRM, ticketing platforms) to enable rapid response and policy adjustments.

Summary

The integration of advertising into AI assistants represents a pivotal moment for enterprise marketing and CX leaders. The Zappi Research report, “Lines in the Sand,” provides a clear directive: early adopters are already setting the norms, and proactive, principled engagement is non-negotiable. Success hinges on a deep understanding of consumer trust dynamics, particularly their sensitivity to contextual relevance and their willingness to opt out of personalization for privacy.

To thrive, organizations must secure early inventory, rigorously enforce contextual exclusions, and pivot towards robust contextual targeting strategies. Establishing clear governance, adapting measurement frameworks, and fostering cross-functional collaboration will enable brands to navigate this nascent channel effectively. The opportunity exists not just to advertise, but to help define the ethical and effective standards for AI-powered brand engagement, ensuring long-term trust and customer value in this transformative landscape.

Reference: Zappi. (2026, May). Lines in the sand: Where consumers will, and won’t, accept ads within AI assistants. Zappi Research.