AI Marketing Research: Gen Z Trust & Creative Deskilling
Three questions are quietly reshaping marketing strategy right now: what actually persuades Gen Z on AI-driven social, whether large language models can do the horizon-scanning work agencies used to charge a fortune for, and whether generative AI is hollowing out the creative teams it's supposed to help. The latest round of AI marketing research offers partial answers to all three — and a few useful warnings.
This briefing summarizes three peer-reviewed papers covered on the AI & Marketing Research Radar. One is a survey of 487 Gen Z consumers in Chennai, India. One is a system-design study for LLM-powered trend detection. The third is a conceptual paper on how generative AI changes the meaningfulness of creative work.
If you lead brand, growth, or creative teams — or advise clients who do — the through-line matters more than any single finding. AI is changing what to measure, how to scan for trends, and what creative work is even for. Here's what the evidence supports, and what it doesn't.
Quick Takeaway
- For Gen Z, voluntary brand advocacy predicts purchase intent more strongly than reach or impressions.
- AI personalization works indirectly — it builds trust and participation, which then drive buying intent.
- Multi-LLM consensus pipelines can auto-map social media trends, but marketing validation is still thin.
- Generative AI risks 'deskilling' creative teams when it replaces core creative decisions, not just grunt work.
- All three papers are early-stage evidence — directional, not definitive.
What This Research Means for Marketers
The practical shift is a measurement shift. If your AI-powered social spend is optimized for impressions and reach, you're likely measuring the wrong signal for Gen Z audiences — participation and advocacy do the actual selling. Meanwhile, the studies on LLM trend detection and creative deskilling point to operational choices most teams haven't made deliberately yet: which parts of research and creative work you automate, and which parts you keep human on purpose.
Papers Covered
Paper 1: AI-Driven Social Media Marketing and Purchase Intention: The Roles of Brand Trust, Consumer Citizenship Behaviour, and Digital Participation among Generation Z
- Source / venue: International Review of Management and Marketing (2026)
- Link: https://doi.org/10.32479/irmm.22943
- Source type: Peer-reviewed journal article
- Method: Quantitative survey analyzed with Partial Least Squares Structural Equation Modelling (PLS-SEM) to test mediation and moderation among AI marketing, brand trust, digital participation, consumer citizenship behaviour, and purchase intention.
- Sample: 487 Gen Z consumers in Chennai, India, recruited via structured questionnaire.
- Main finding: AI-driven social media marketing increases Gen Z purchase intent indirectly — through brand trust, digital participation, and especially consumer citizenship behaviour (voluntary sharing, recommending, defending the brand), which was the strongest predictor in the model.
- Evidence strength: Peer-reviewed, moderate sample, but single-city, self-report, cross-sectional — correlational only.
- Limitation: Findings come from one city in India, rely on self-reported intent rather than observed purchases, and don't isolate which AI tactic (chatbots vs. recommendations vs. personalized content) drives results.
- Practical implication: Track shares, comments, and UGC as primary Gen Z KPIs — not impressions. Design AI-personalized content to earn trust and prompt advocacy, then let advocacy do the selling.
Paper 2: Automated Semantic Ontology Construction for Foresight Studies Using Large Language Models
- Source / venue: System Research and Information Technologies (2026)
- Link: https://doi.org/10.20535/srit.2308-8893.2026.2.09
- Source type: Peer-reviewed journal article
- Method: System design and validation study. Pipeline scrapes social media text, runs multiple LLM configurations to extract concepts, applies a consensus mechanism across models, embeds outputs in vector space, and iteratively clusters them using cosine similarity.
- Sample: Social media text data from unspecified volume and time period; sources include a Ukrainian Telegram channel ('Victory Drones'). No human participants.
- Main finding: The multi-model consensus pipeline produced stable, converging concept clusters and reduced hallucination rates compared to single-model extraction — suggesting a viable, lower-cost alternative to expert-panel foresight work.
- Evidence strength: Peer-reviewed but in a regional Ukrainian journal with limited international visibility; no benchmark comparison against expert-produced ontologies; domain (drone warfare Telegram) is far from commercial marketing use cases.
- Limitation: No head-to-head cost comparison with traditional foresight methods, no external validation against expert output, and no evidence yet that the pipeline generalizes to consumer or brand data.
- Practical implication: The multi-LLM consensus technique — asking several models the same question and keeping only what they agree on — is a portable tactic for reducing errors in AI-assisted marketing research today.
Paper 3: The Impacts of Generative AI on the Meaningfulness of Creative Work
- Source / venue: Journal of Business Ethics (2026)
- Link: https://doi.org/10.1007/s10551-026-06342-4
- Source type: Peer-reviewed journal article (conceptual/theoretical)
- Method: Theoretical analysis integrating Amabile's componential model of creativity with a holistic framework of meaningful work. No empirical data collected.
- Sample: Not applicable — conceptual paper.
- Main finding: Generative AI can enhance creative work by automating repetitive tasks, but risks 'deskilling' when it takes over core creative decisions. The paper also identifies a 'penalty for AI use' — social and professional stigma attached to visibly AI-heavy work.
- Evidence strength: Peer-reviewed in a reputable journal, but conceptual only — no empirical measurement of deskilling, stigma, or meaningfulness effects.
- Limitation: Arguments depend on the specific creativity and meaningful-work frameworks chosen; individual variation (some creatives find AI liberating) is acknowledged but not resolved; no data on the size or dynamics of the 'AI penalty'.
- Practical implication: Keep human judgment central to creative decisions, not just AI output review. Think carefully about how you frame AI's role to clients — 'AI-assisted' and 'AI-generated' may land very differently.
Plain-English Payoff
AI is changing how brands reach Gen Z, how strategists scan for trends, and how creative teams do their jobs — and in each case, the upside is real but conditional. For Gen Z, stop optimizing for eyeballs and start designing content worth sharing. For trend work, multi-model AI pipelines are becoming plausible, but not yet proven for marketing. For creative teams, use AI to eliminate drudgery, not judgment.
Money Move
Package a Gen Z engagement audit for brands still optimizing social spend for impressions. Pull participation metrics — shares, comments, UGC, brand defense — separately from reach, benchmark them against category peers, and prescribe AI personalization changes that shift the mix toward advocacy. The Chennai study gives you a defensible business case, and most brands marketing to this cohort still haven't rebuilt their measurement around it.
Evidence Check
- All three papers were reviewed at the full-text level, but only the Gen Z paper offers empirical marketing data.
- The Gen Z findings are correlational and self-reported — don't overclaim causality or global generalizability from one city.
- The LLM foresight paper is published in a regional journal with limited international replication; treat it as a blueprint, not proof.
- The creative deskilling paper is conceptual — a well-argued framework, but no measured effect sizes.
- Sample sizes, effect magnitudes, and cost comparisons should not be inferred beyond what the papers report.
What to Test Next
- Action step. Run a two-week audit of a Gen Z social campaign that separates participation metrics (shares, comments, UGC) from reach and impressions, and compare which correlates more tightly with conversions in your own data.
- Action step. Pilot a multi-LLM consensus workflow for one recurring research task — competitor scans or trend briefs — by running the same prompt across three models and keeping only overlapping outputs. Track time saved and error rate versus your current process.
- Action step. Map your creative workflow into 'automate,' 'assist,' and 'human-only' zones. Identify at least two creative decisions currently delegated to AI that should move back to a human owner to protect long-term team skill.
- Action step. Draft internal and client-facing language for how your team describes AI's role in creative work, and test which framings preserve trust with the clients or audiences you care about.
How This Connects to AI Marketing Strategy
Across these three papers, one pattern keeps surfacing: AI's biggest marketing effects are indirect. It doesn't sell Gen Z directly — it earns trust that fuels advocacy. It doesn't replace strategists directly — it reshapes what desk research looks like. It doesn't replace creatives directly — it changes what creative work means and how it's valued. Marketers who treat AI as a direct output engine will keep measuring the wrong things.
FAQ
What does AI marketing research say about reaching Gen Z?
One peer-reviewed 2026 survey of 487 Gen Z consumers in Chennai found that AI-driven social media marketing increases purchase intent indirectly, through brand trust, digital participation, and voluntary brand advocacy. Advocacy was the strongest predictor — suggesting shares and recommendations matter more than impressions for this cohort. Findings are correlational and single-city, so treat them as directional.
Is 'consumer citizenship behaviour' just engagement by another name?
Not quite. Engagement typically includes any interaction, including passive ones. Consumer citizenship behaviour refers specifically to voluntary, pro-brand acts — recommending, defending, sharing without being asked. The Chennai study found this specific behaviour predicted purchase intent more strongly than general digital participation.
Can large language models really do trend detection for marketers?
The system-design paper in this briefing shows a multi-LLM consensus pipeline can produce stable topic maps from social media text with reduced hallucination rates. But it was tested on non-marketing data (including a Ukrainian military Telegram channel) and hasn't been benchmarked against expert output. Treat it as a promising blueprint that needs marketing-domain validation before you replace human analysts.
Is generative AI hurting creative teams?
The Journal of Business Ethics paper argues it can — when AI takes over core creative decisions rather than repetitive tasks, workers lose opportunities to develop skill and find the work less meaningful. The paper also flags a growing 'penalty for AI use' — reputational stigma around visibly AI-heavy work. These are conceptual arguments, not measured effects, but they're worth taking seriously in team design.
How should small businesses apply this AI marketing research?
Focus first on the Gen Z measurement shift: if you sell to younger audiences, track participation and advocacy metrics alongside reach. Use AI personalization tools to build trust and prompt sharing rather than push volume. Keep human judgment on the creative decisions that shape your brand voice — and use AI to remove drudgery around it.
What is a multi-LLM consensus workflow?
It's a technique where you run the same query through several large language models and keep only the outputs where the models agree. The foresight paper found this approach reduced hallucinations and produced more stable results than any single model. It's a low-cost way to add error checking to AI-assisted research tasks.
Should agencies disclose AI use to clients?
The creative work paper suggests framing matters. There's an emerging reputational penalty for work perceived as AI-heavy, even when quality holds up. This doesn't mean hiding AI use — it means being deliberate about how you describe AI's role, and keeping humans clearly accountable for creative judgment and final output.
Listen to the Episode
Sources and Further Reading
- AI-Driven Social Media Marketing and Purchase Intention: The Roles of Brand Trust, Consumer Citizenship Behaviour, and Digital Participation among Generation Z
- Automated Semantic Ontology Construction for Foresight Studies Using Large Language Models
- The Impacts of Generative AI on the Meaningfulness of Creative Work
Related Big Plans Media:
About Big Plans Media
Big Plans Media helps marketers, educators, entrepreneurs, consultants, and business leaders translate AI marketing research into practical strategy. AI & Marketing Research Radar is produced by Big Plans Media and hosted by Evita, an AI-generated research briefing avatar trained on Dr. Eva Wolf's research framework.
