AI influencer credibility alone does not raise purchase confidence among young Ghanaian consumers, a 27 August 2026 study of 817 social media users has found. Trust is the only mechanism that converts perceived expertise into confident buying decisions.

The study, published in Frontiers in Communication by Ebenezer Arthur Duncan, W. A. Inkumsah, and R. Opare-Boateng, is the first dedicated look at AI influencer credibility among youth in sub-Saharan Africa's beauty and fashion sector.

What did the researchers measure?

The team surveyed Ghanaian social media users aged 18-45 who had bought or considered buying a beauty or fashion product in the past six months. Quotas tracked platform usage: 367 from Facebook, 341 from YouTube, 109 from Instagram. Respondents also had to confirm prior exposure to AI-driven personas such as Aitana Lopez or Imma.

Sample breakdown:

• Age 18-24: 37.0% (302)
• Age 25-34: 34.0% (278)
• Age 35-44: 19.2% (157)
• Age 45+: 9.8% (80)
• Women: 57.3%; Men: 42.7%
• First-degree holders: 49.1%

The survey used 24 five-point Likert items. It tested four core constructs: AI-driven influencer credibility (AIDI), cultural values (CV), trust in AI influencers (TAIDI), and decision confidence (DC). Influencer attachment (IA) was modeled as a separate outcome.

What did the data show?

All seven hypotheses held. The structural model explained 50.4% of the variance in decision confidence and 59.0% of the variance in trust. Influencer attachment sat lower at 17.7%.

Key path coefficients:

• AIDI → TAIDI: β = 0.768, t = 40.12, p < 0.001 (f² = 1.436)
• TAIDI → DC: β = 0.733, t = 12.61, p < 0.001 (f² = 0.417)
• CV → AIDI: β = 0.383, t = 7.60, p < 0.001
• CV → DC: β = 0.123, t = 2.51, p = 0.012
• AIDI → DC: β = -0.115, t = 1.79, p = 0.073 (not significant)

Why does trust matter so much?

The most direct result from the structural model: AI-driven influencer credibility had a negative, non-significant direct effect on decision confidence. Without trust in the AI source, exposure to even a credible-looking virtual persona does not move purchase certainty — and may even dent it.

The indirect effect through trust was positive and significant (β = 0.563, t = 11.55, p < 0.001). This is the textbook definition of indirect-only mediation. Credibility builds trust, and trust is what then drives decision confidence.

"For AI agents, consumers may perceive competence yet withhold decision confidence until trust is established, due to concerns about autonomy and lack of human accountability," the authors write.

Does culture change the math?

Cultural values did double duty. They raised perceived credibility of AI influencers (β = 0.383) and pushed decision confidence directly (β = 0.123). They also strengthened emotional attachment to the influencer (β = 0.272, p < 0.001).

A robustness check of an interaction term — CV × AIDI — came up empty. The relationships did not change at different levels of cultural values.

What does this mean for brands targeting Ghanaian youth?

The authors frame three operational takeaways.

• Cultural alignment drives credibility. Content using local languages, aesthetics, and values lifts perceived expertise. Cultural congruence also has a direct line to decision confidence, independent of the trust pathway.
• Trust is the bottleneck. Since trust is the only bridge from credibility to confident purchases, transparency and consistency are the levers that matter. Clear disclosure of AI-generated content and cross-platform coherence are practical starting points.
• Attachment is a parallel track. Both cultural relevance and credibility build emotional bonds. Storytelling that builds a sense of connection can sustain engagement even without a human presenter.

What doesn't the study cover?

Four caveats from the authors. The data are cross-sectional, so the model is associational, not causal. The sample is Ghana-only and ages 18-45. Cultural values were measured with a three-item short form focused on collectivism and power distance. Common method variance was only checked with Harman's single-factor test, which the authors flag as insufficient.

The study also focused exclusively on beauty and fashion. The authors call for replication in finance, health, and other product categories where the trust logic might break down.