What citizens are asking their cities about culture
City-deployed AI chatbots, fed by Encore. Each conversation is a real citizen asking a real question about their city's cultural programme — we aggregate, classify, and report.
June — from probe to exploration.
226 cultural conversations across the Capital of Culture network — up 65% on May and a third consecutive monthly record. 113 distinct sessions across 25 active days of 30. Five cities, five languages handled natively — English, Spanish, Czech, Dutch, and (returning) Serbian. Average response length 1,249 characters; 86% of replies are substantive long answers.
Five cities live: Brno, Cáceres 2031, Larnaka 2030, LOV2030, Novi Sad. Cohort mean Q/session 2.00 (up from 1.76) — the exploration threshold, crossed. 6Q+ deep sessions tripled (2 → 6); one-shot ≤3-word queries fell from 35% to 25% of traffic.
- 54%LOV2030Belgium
Third consecutive record month (121 messages, +70%) and a first-ever depth breakthrough: 4 deep sessions including a 12-question dive. The final week alone carried 69 interactions around the After Hours summit evening. The agent also met its most serious moment yet — urgent safety questions about a children's event during the 37°C heatwave.
- 25%Cáceres 2031Spain
Volume nearly doubled (56 messages) while keeping the cohort's deepest engagement: 2.24 Q/session and the longest answers (1,620 chars avg, 98% long-form). Includes the month's best story: a canonical-answer gap flagged in May's reporting was stress-tested by the city team on 8–9 June and verified fixed by the 11th — the network's first closed flag → fix loop.
- 15%LarnakaCyprus
Volume tripled (35 messages, reversing two months of decline) and depth improved: average Q/session up from 1.09 to 1.59. A cluster of near-identical strategy questions arrived study-group-style — someone out there is teaching the bid book.
- 4%Novi SadSerbia
Small but its most substantive month: 2.5 Q/session, its first 6Q+ deep session, and Serbian back as the primary language. A resident moved from concert-programme questions to a festival-finance transparency challenge — and got calm, factual, non-defensive answers throughout.
- 2%BrnoCzech Republic
A quiet month (4 one-question sessions). Notably, an identical Cultural Forum cost query asked twice received the identical figure both times — a small consistency win worth having.
| City | Msgs | Sess | Days | Avg Q | Max Q | Deep (6Q+) | Avg resp (ch) | Long resp |
|---|---|---|---|---|---|---|---|---|
| LOV2030 | 121 | 58 | 20 | 2.09 | 12 | 4 | 1,201 | 81% |
| Cáceres 2031 | 56 | 25 | 12 | 2.24 | 7 | 1 | 1,620 | 98% |
| Larnaka | 35 | 22 | 13 | 1.59 | 5 | — | 1,015 | 91% |
| Novi Sad | 10 | 4 | 3 | 2.50 | 6 | 1 | 938 | 70% |
| Brno | 4 | 4 | 4 | 1.00 | 1 | — | 343 | 50% |
| Cohort total | 226 | 113 | 25 | 2.00 | 12 | 6 | 1,249 | 86% |
How many questions a session contained.
- 1Q64 sessions (57%)
- 2Q26 sessions (23%)
- 3Q7 sessions (6%)
- 4Q5 sessions (4%)
- 5Q5 sessions (4%)
- 6Q+6 sessions (5%)
86% of cohort sessions are 1–2 questions. The cohort is largely a probe-and-go audience.
25 of 30 days had cohort activity.
- Peak: 23–25 June — 66 messages over three days, the LOV2030 After Hours / Sing the City event window plus the heatwave. Time-critical content demand at its clearest.
- Secondary peak: 9 June (17 messages) — concentrated in Cáceres; largely the city team stress-testing the agent after that day's award news, i.e. internal QA rather than citizen traffic. Flagged because honest peaks matter.
- Larnaka's growth was spread evenly (13 active days) rather than event-spiked.
- LOV2030
English/Dutch mix, fluent Dutch replies confirmed throughout (including the emotionally loaded heatwave exchanges).
- Cáceres 2031
~98% Spanish; the single English query (a donor request) is under technical review.
- Larnaka
English only this month; no Greek-script queries (25% in May).
- Novi Sad
Serbian back as the primary language, handled natively.
- Brno
Czech + English split.
- 24%Strategy & bidbook53 hits (23.5%). Sharply up from 9.5% in May and still the cohort's engagement spine; Larnaka's study-group cluster and Cáceres' post-award stress-test both land here.
- 9%Events & dates21 hits (9.3%). The biggest engagement days in every growing city were event-driven — and event logistics (registration, agendas, "do I need to sign up?") were also the month's most repeated fallback.
- 4%Volunteer & participation10 hits (4.4%). May's "next frontier" arrived: a 7-question volunteer application session in Cáceres ended with the agent drafting the applicant's email. Doubled share month-over-month.
- 4%Contact & staff8 hits (3.5%). Still the most reliable bottom-of-funnel signal, and still the biggest unconverted ask — one user's exit line: "Then I'll go look on Google."
- 2%Open calls & submissions5 hits (2.2%). Including a summer-stage applicant chasing a response to an open-call submission — a warm lead, and a case for automatic submission confirmations.
- 2%Safety & transparency5 hits (2.2%). New this month: heatwave safety concerns (LOV2030) and a festival-budget transparency challenge (Novi Sad). Citizens are treating the agents as trust and escalation channels.
- 1%Vacancies & jobs3 hits (1.3%). Smaller share, same structural gap — the vacancies content miss is now five months running.
- 1%Brand assets & logo3 hits (1.3%). Quieter than May's cluster.
- 1%Partnership2 hits (0.9%). Cross-city and inter-org partnership questions persist modestly.
The case for an 18th SDG on Culture
Look at what people are actually asking these chatbots about — and you start to see something interesting. Mapped against the 17 UN Sustainable Development Goals, the conversations don't scatter randomly. They cluster. Culture isn't a sidebar to sustainable development. It's the thread running through almost all of it.
- SDG 11
Sustainable Cities & Communities
Event access, venue and programme questions led volume again; the heatwave-safety exchange is SDG 11 in its most literal form — a citizen asking whether a public cultural space is safe.
- SDG 16
Peace, Justice & Strong Institutions
The month's defining theme. A resident challenging festival finances and getting factual, non-defensive answers; a parent's safety concern taken seriously — cultural agents functioning as responsive-institution infrastructure.
- SDG 17
Partnerships for the Goals
Partnership and collaboration queries persisted; a cross-border volunteer application (Cuba → Cáceres) is the month's best emblem.
- SDG 4
Quality Education
Larnaka's study-group-shaped cluster suggests the bid book is being taught somewhere.
- SDG 10
Reduced Inequalities
Five languages handled natively, including Serbian's return; one empty-response incident (English donor query, Cáceres) logged and under review — access equity is measured at the failure points, not just the successes.
- SDG 8
Decent Work & Economic Growth
Vacancies and jobs queries persist against a five-month content gap — demand without supply.
SDG 18: Culture as a Driver of Sustainable Development
Every conversation in this dataset touches at least one SDG. Many touch four or five. The people asking these questions aren't thinking in silos — they're asking about belonging, about identity, about what their city stands for and how they can be part of it. Culture isn't the means to the other 17 goals. For many of these users, it isthe goal. The evidence for a dedicated SDG on culture isn't in a policy paper. It's in 715 conversations — and counting.
- Working well
The depth threshold was crossed
May's page set a marker: above 2.0 Q/session would signal a move from probe to exploration. June hit 2.00 exactly, with deep sessions tripled and one-shot queries down ten points. The audience the cohort was built for — citizens who stay and dig — is measurably arriving.
- Growing trend
Agents as trust infrastructure
The month's two most serious conversations weren't information lookups: an urgent safety question during a 37°C heatwave and a festival-finance transparency challenge. Both agents held a calm, factual register. What neither had was the thing that ends such conversations well — a live fact or a monitored human contact. Escalation contacts are the next content layer.
- The opportunity
The loop closes when cities engage
Cáceres turned a May report line into a shipped fix in under a week, then verified it through the agent itself. Meanwhile the cohort's most repeated miss (event registration details) has an equally simple habit-fix: whenever a city promotes a named event, the registration/agenda entry lands in the knowledge base the same week.
Aggregated from 5 live city widgets · feeds the next Tuesday digest