Do AI-generated presenters communicate cultural heritage as well as people?
In a randomized 601-person experiment, one human presenter was rated clearer and more immediate than one AI-assisted digital presenter. Information-quality pathways looked similar, but the test covered a single one-minute video and measured intentions, not real sharing.
The short version
- What happened
- Researchers randomized 601 mainland-Chinese adults to watch a roughly one-minute Hakka tulou explainer with either one human narrator or one AI-assisted digital narrator.
- Who it affects
- Cultural practitioners choosing automated presenters need evidence about trust, clarity and learning—not just production cost or novelty.
- What remains unproven
- Convenience sample concentrated among ages 25–38 in mainland China
Keep in mind
Convenience sample concentrated among ages 25–38 in mainland China
One cultural topic and one narrator of each type
Voice, appearance, motion and narrator identity were not separated
Only about 53% correctly identified narrator type
Self-reported ratings and sharing intention, not comprehension or observed sharing
Small indirect effects
Source analysis · 1 linked source. Primary analysed source date: 26 October 2026.
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Read the full report · 4 min The evidence, methods, implications and limitations
One script, two complete presenter packages
A human presenter outscored an AI-assisted digital presenter on two viewer ratings in a randomized experiment about Hakka tulou heritage. The 601 participants, recruited online in mainland China and aged 25 to 38 on average, watched one of two smartphone videos. Both versions used the same 252-character script, background footage, music, subtitles, editing pace and near-identical one-minute duration.
The human version scored 3.745 versus 3.397 for nonverbal immediacy, a 0.348-point difference on a five-point scale (95% confidence interval 0.230 to 0.466; Cohen's d 0.474). Perceived clarity was 3.725 versus 3.355, a 0.370-point difference (95% CI 0.251 to 0.489; d 0.497). Those are moderate differences for these particular videos—not a general verdict on every digital presenter.[1]
The source pathway differed more than the information pathway
The researchers modelled two routes to reported inspiration and intention to share: nonverbal immediacy through narrator credibility, and clarity through perceived information quality. Links from immediacy to credibility and from credibility to inspiration were stronger in the human condition. By contrast, the differences between conditions for clarity-to-information-quality and information-quality-to-inspiration were not statistically significant.
That distinction matters for museums, educators and cultural organisations. A digital presenter may be less effective when the communicator's warmth and credibility are central, while careful scripting and clear information can still support the content route. Yet the mediation links are associations measured after exposure. The experiment randomized the video condition, not credibility or inspiration themselves, and the indirect effects were generally small.[1]
The distinction that matters
What differed—and what did not
The randomized comparison supports a claim about the two finished videos, not a single isolated AI feature.
Human condition
One human face, voice and performance
Higher immediacy and clarity ratings
Digital condition
One AI-assisted face, voice and animation
Lower source-related ratings
Unresolved
Identity and production cues varied together
No universal human-versus-AI effect
Narrator type was not cleanly isolated
The test used one human and one digital presenter. Their faces, voices, gestures, intonation and animation quality all varied together, so the study cannot say that 'being AI' caused the rating gap. Only 53.33% of people in the human arm and 53.49% in the digital arm identified the narrator type correctly, making the manipulation check unusually weak even though assigned-condition results remain valid for the two videos shown.
Participants reported impressions and electronic word-of-mouth intention; researchers did not observe sharing, cultural knowledge retention or later behaviour. The authors also disclosed using GPT Image 2 for the digital narrator and related images and ChatGPT to polish the manuscript. The useful next test would cross several human and digital presenters, hold voice and motion cues separately, preregister outcomes and measure comprehension and actual sharing across ages and cultures.[1]
Watch the background, then check the evidence
Illustrative photo · not a video frame
The Royal Society · 44 min 28 sec · Background explainer
Is this the AI future we want? | The Royal Society
YouTube loads only when you choose; playback starts when you press play.
A Royal Society discussion with Madhumita Murgia and biomedical engineer Alison Noble on AI’s potential in healthcare, science and culture, recorded at the 30 June 2026 event.
Evidence check: A discussion of opportunities and risks, including future possibilities. It is not evidence that every application described is available or clinically proven.
Published 2026-07-01. Source checked 14 October 2026. Check captions on the publisher’s video page.
Check the healthcare evidencePhoto: Daniel Sone / National Cancer Institute / Unsplash · Licence
Continue the story
Read next
Continue exploring cultural heritage: compare the finding and its evidence limits.
What can Goa's new Konkani AI do—and what still needs testing?
Goa and India's BHASHINI programme demonstrated two-way speech translation and launched a Konkani language model. The official account gives no model card, dataset size, error rate or results across Konkani's different scripts.
Read the next reportExplore the full source and evidence record
What this means for people
- Cultural practitioners choosing automated presenters need evidence about trust, clarity and learning—not just production cost or novelty.
Global context
Cultural expectations about presenters, synthetic media and heritage representation vary. Results from one Chinese-language Hakka tulou video need replication across communities and platforms.
What the evidence does not yet show
- Convenience sample concentrated among ages 25–38 in mainland China
- One cultural topic and one narrator of each type
- Voice, appearance, motion and narrator identity were not separated
- Only about 53% correctly identified narrator type
- Self-reported ratings and sharing intention, not comprehension or observed sharing
- Small indirect effects
What to watch next
- Factorial tests separating face, voice, motion and disclosure
- Multiple presenters and heritage topics
- Retention, accuracy and observed sharing
- Community review of representation and consent
Living evidence record
Impact record IAI-1SVS37Y
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Present
Record status
Monitoring
Last checked
26 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what Frontiers in Psychology published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 26 October 2026
Research question, topics and regions
How viewers evaluate a human narrator and an AI-assisted digital narrator in matched cultural-heritage videos
Topics: Digital humans · Cultural heritage · Short video · Trust
Regions: China · Asia
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.