What a Robot Guide Can and Cannot Do: Evidence from 2026 Public Culture-and-Tourism Projects
The phrase robot guide suggests the robot is the point. Look at the culture-and-tourism projects that actually shipped in 2026, and the real drivers are three other things: a vertical knowledge base, voice and vision interaction, and location awareness. The robot body is only the shell that carries them to the visitor.
Capabilities already in production
| Capability | Public project |
|---|---|
| Automatic commentary at points of interest, object photo recognition, multi-turn free Q and A, group relay guiding | Guangdong Museum of Folk Craft smart AI-glasses guide, built with Rokid and Yuancheng Technology, combining an on-device lightweight museum-specific model with speech recognition and waveguide AI glasses |
| Multilingual venue commentary, immersive cloud tours, smart itinerary planning | Chaoshan intangible cultural heritage digital platform by Hanshan Normal University (large models, voice cloning, digital humans, 3D modelling and VR, AIGC) |
| Library search and recommendation, automated stack scheduling, drone inventory, AI library assistant | Shenzhen Library (multi-model collaboration, knowledge graph, embodied robots, hybrid cloud-compute plus local-knowledge-base plus AI-middleware architecture) and Futian District Library's Fuxi Zhitu system with assistant Fululu |
| One-stop service for inbound visitors: booking, entry, guiding, payment, enquiries | Guangdong Museum's Changxing Yuebo smart tourism project (large models, RAG, agents, computer vision, multimodal data) |
| Localised personalised recommendation and consumption guidance | Wanli Data's ChaoShan check-in AI platform (millions of local user behaviour records plus a local merchant knowledge base, with OCR receipt recognition) |
| Differentiated content and IP-based operation | EngineAI URKL global humanoid robot free-combat league, using robot combat as a differentiated content format |
Four boundaries not yet solved
- Weak at complex narrative and emotional delivery: standardised commentary works, historical context and emotional pacing still need people
- Degraded recognition in noisy environments: accuracy drops noticeably in open scenic areas and peak-hour venues
- Knowledge updates remain manual: an unmaintained knowledge base goes stale, so content governance is recurring cost rather than one-off delivery
- No cross-scenario transfer: a museum knowledge base is useless in a scenic area, so vertical knowledge must be built per scenario
Three prerequisites before deployment
- A vertical knowledge base: the Guangdong Provincial Cultural Centre built a six-library-plus-knowledge-graph dataset covering programmes, exhibits, literature, events, talent and awards, turning scattered resources into searchable structured knowledge
- Compute and network: Shenzhen Library uses a hybrid architecture of cloud compute, local knowledge base and AI middleware to balance response speed with data localisation
- An operating mechanism: content updates, fallback handling and on-site staffing determine whether the system survives its first holiday peak
Three questions before choosing a solution
- Is the problem insufficient commentary or insufficient reception capacity? The first prioritises knowledge base and voice, the second multilingual support and group relay guiding
- Do you need a physical robot at all? If the goal is commentary and Q and A, AI glasses, audio guides and mini-programs are usually cheaper than a humanoid robot; the robot body matters when brand presence and interactive atmosphere do
- One-off investment or long-term operation? Museum-enterprise collaboration can achieve zero public spending, while owned equipment requires budget for content refresh and maintenance
One-line judgement: a robot guide is not buying a machine, it is building a knowledge base, an interaction layer and an operating routine. The body can be rented, bought or replaced by AI glasses and guide terminals, but the knowledge base and the operations cannot be skipped.
Cases are drawn from publicly released Guangdong 2026 AI-plus-culture-and-tourism projects, including the Guangdong Museum of Folk Craft AI-glasses guide, the Guangdong Museum's Changxing Yuebo project, Shenzhen Library and Futian District Library's Fuxi Zhitu, the Guangdong Provincial Cultural Centre's AI assistant, Hanshan Normal University's Chaoshan heritage platform, Wanli Data's ChaoShan platform and the EngineAI URKL robot combat league. In the same period, Guangdong reported RMB 12.55 billion of tourism spending during the Mid-Autumn holiday, with visitor numbers up 28.9% year on year.
FAQ
- What can a robot guide do today?
- Verified capabilities include location-triggered automatic commentary, object photo recognition, multi-turn free Q and A and group relay guiding (an AI-glasses project at the Guangdong Museum of Folk Craft), multilingual venue commentary and immersive tours (a Chaoshan intangible-heritage platform), library search, automated stack scheduling and drone inventory, and one-stop service for inbound visitors covering booking, entry, guiding, payment and enquiries.
- Can robot guides replace human guides?
- Not yet; they supplement. They excel at high-frequency, standardised, multilingual repeat commentary and enquiries, but not at complex historical narrative, emotional delivery or deep interaction. The common pattern is AI for standard content plus human backup for highlights.
- What must be ready before deployment?
- Three things: a vertical knowledge base, as in the six-library-plus-knowledge-graph dataset used by the Guangdong Provincial Cultural Centre; compute and network, with a hybrid cloud-compute plus local-knowledge-base plus AI-middleware architecture; and an operating mechanism covering content updates, fallback and on-site staff.
- What budget levels apply?
- Public cases rarely disclose single-project figures, but two models recur: zero-public-spend museum-enterprise collaboration, and direct purchase or rental where public daily rates run RMB 2,500–5,000 for humanoid robots.