From the institutional website to the intelligent citizen service
Municipal websites were created to publish information. Citizens, however, need guidance and results. With the arrival of AI-powered browsers (like Atlas), the relationship changes: residents no longer "search" through internal menus, ask and expects a clear, verifiable, and official response. This article explains how to move from an informational portal to a guiding service, with an AIO/Atlas approach that benefits both people and AI systems.
- The current problem: websites designed for departments
- What changes with AI-powered browsers
- How AI “thinks” when reading a municipal website
- Citizen-centered (and AI) design principles
- Technical checklist: structure, schema and natural language (and in AI)
- 90-day improvement plan for a public portal
- Governance, security and reliability: the new duties of management
- Closing: From Information to Citizen Trust
1. The current problem: websites designed for departments, not for people
Most institutional websites reflect the internal structure of the city council: areas, units, headings. This map is useful for working internally, but not for addressing external needs. A citizen doesn't think of "Urban Planning > Subsidies"; they think of... “What assistance is available to renovate my home?”. When the web replicates the organizational chart, the search becomes an obstacle course: broken links, PDFs without context, technical jargon, and pages without a clear path to action.
- Extensive menus that mix procedures, news, and regulations.
- Administrative language (“correction”, “procedure”) without citizen translation.
- PDF forms, scattered requirements, and generic contact phone numbers.
- Excessive time to find a procedure or an appointment.
- Low completion rate for online transactions.
- Mistrust and an increase in repetitive inquiries about the same issues.
| Dimension | Typical portal | Citizen service |
|---|---|---|
| Structure | Internal organizational chart | Intentions and life situations |
| Language | Technical and legal | Natural, with examples and glossary |
| Interaction | Static pages + PDFs | Step-by-step guided workflows |
| Success metric | Page views | Helpful answers and completed procedures |
If the citizen doesn't understand the website, the website doesn't serve the citizen—or the institution.
2. From the institutional website to the intelligent service: what changes with AI browsers
AI-powered browsers do not list pages; they solve tasks. A resident asks a question (“How do I apply for rental assistance if I’m young?”) and the system summarizes the answer, citing reliable sources. In this scenario, the official municipal source can—and should—be the primary reference, but only if the information is clear, up-to-date, and structured for reuse.
- Immediate and understandable answers, without internal jargon.
- Clear eligibility: “Do I meet the requirements?”, “What documents do I need?”
- Guided tours by profile (young people, families, self-employed, seniors).
- Clear titles, short paragraphs and native tables for requirements/deadlines.
- FAQ in natural language (how, where, when, how much).
- Data with source, date and responsible party (verifiability).
The opportunity is enormous: a municipal website understandable by AI acts as source of truth that other attendees (public or private) reuse, multiplying the reach of the service without multiplying the support.
Turn your portal into the "official source" that AI wants to cite.
3. How AI “thinks” when reading a municipal portal
An AI system interprets content types (services, procedures, news, datasets) and looks for consistent patterns: title hierarchies, lists of steps, tables of requirements, actual questions and answers, and metadata indicating who publishes, when it was reviewed, and which regulation or call for proposals it corresponds to. It doesn't evaluate institutional power; it evaluates clarity, coherence and verifiability.
- H1 unique for the service and H2 by intention (what it is, who can, how to apply).
- Tables with “requirements / documents / deadlines / where to submit”.
- FAQ with genuine questions from citizens and customer service staff.
- Text embedded in PDFs without an HTML summary.
- Contradictions between pages or outdated dates.
- The same procedure described in several URLs without a canonical code.
| Pattern | How to implement it | Result |
|---|---|---|
| Question and answer | FAQ in natural language (how, where, when, how much does it cost) | Quotable and consistent fragments in attendees |
| Table of requirements | Native HTML with clear headings and notes | Reusable comparisons and checklists |
| Metadata | Date, responsible party, version, link to standard | Greater trust and traceability |
AI does not interpret hierarchies; it interprets structure and evidence.
4. Citizen-centered (and AI) design principles
Designing for AI is, at its core, designing for people who shouldn't need to know the organizational chart. The initial question isn't "Where do I post this information?", but rather "What does the citizen need to know to achieve their goal with minimal effort?" These principles translate that approach into a public portal.
- Organize by life situations: living, moving, undertaking, caring, participating.
- Create routes by profile (young people, families, self-employed, seniors, organizations).
- Avoid mixing news with service: they are different intentions.
- Titles that describe tasks (“How to register if…”) instead of headings.
- Glossary and "easy reading" for unavoidable terms.
- Examples and notes on “frequent cases” to reduce doubts.
- Tables of requirements/documents/deadlines with notes.
- FAQs per service with 6–8 real questions.
- "What to expect" blocks (times, costs, response).
Mini-guide to rewriting (before/after)
Previously (institutional): “Procedure for applying for a minor works permit. Deadline for correcting deficiencies in accordance with ordinance X.”
Then (focused on the citizen):
- What is it: License for simple renovations (painting, changing flooring, kitchen without structural changes).
- Who can apply: owner/authorized person.
- Documents: ID card, sketch, budget, community consent if applicable.
- Estimated timeframe: 7–15 business days.
- How to present: electronic headquarters (steps and link) or in-person registration by appointment.
Fewer clicks, fewer repeat calls, and more procedures completed the first time.
5. Technical checklist: structure, schema and natural language
This checklist guides the work of web, communications, and systems teams. The goal is not to "have everything," but activate only what is essential so that the information is understandable by people and understandable by AI.
| Element | Concrete action | Result | Where to validate it |
|---|---|---|---|
| Schema types | Use Government Organization (cover and contact), Service (service sheets),
FAQ Page (real questions), Breadcrumb List and WebPage with dateModified. | Interpretability and citability by AI. | Rich Results Test + Search Console → Improvements. |
| Architecture | Separate services/procedures of news and rules. One intent per URL; canonical when there is duplication. | Less ambiguity and clear routes. | XML site map + editorial review. |
| Natural language | Titles that answer the “how/where/when/what I need”, paragraphs ≤90 words, glossary and examples. | Human understanding and more accurate AI synthesis. | Editorial QA + tests with citizens. |
| Native tables | Requirements, documents, deadlines, costs and channels in HTML (not image/PDF). | Reusable fragments by AI and accessible reading. | DOM inspection + screen readers. |
| Metadata | Date of review, person responsible for the content, version and link to standard or call for proposals. | Greater trust and traceability. | Visible footers/sidebars. |
| Accessibility | Alternative text, adequate contrast, keyboard navigation, easy reading and translation into co-official languages. | Inclusion and legal compliance. | WAVE / Lighthouse / WCAG audit. |
- Publish only one PDF with the call for applications without HTML summary.
- Duplicate FAQ (plugin + custom block) in the same URL.
- Service records without a revision date or responsible party.
Each service card is ready to be cited by AI and useful for the citizen.
6. 90-day improvement plan for a public portal
One quarter is enough to transform an informational portal into an intelligible service. This plan is implemented in three phases with measurable deliverables and shared responsibility between communications, systems, and service areas.
- Identify 10 priority services (impact/volume/complaints).
- Map intent by service (what question each URL answers).
- Audit current structure (H1/H2/H3), tables and FAQs.
- Rewrite fact sheets with TL;DR, tables, and 6–8 real FAQs.
- It includes metadata (date, responsible party, version, standard).
- Glossary and easy-to-read version of required terms.
- Implement schema (Organization/Service/FAQ/Breadcrumb).
- Optimize CWV (LCP, INP, CLS) on mobile and accessibility review.
- Clean indexing (sitemap, canonicals, 3xx redirects).
Indicators of success
| KPI | How to measure | Threshold/target |
|---|---|---|
| Useful response time | Task test with 10 citizens | < 30 seconds to find “what I need” |
| Impressions at GSC | Performance per rewritten URL | +25–40% vs. baseline |
| CTR in long-tail queries | GSC by query/URL | ≥ 4–6% sustained |
| Completion of online procedures | Electronic headquarters metrics | ↑ compared to the previous 3 months |
It's not about "expanding" the portal, but about make understandable The 10 services that most affect citizens.
7. Governance, security and reliability: the new duties of management
GDPR · Data Protection ENS · Security eIDAS · Digital Identity Law 39/2015 · Procedure ENI · Interoperability
If citizens use AI to stay informed or even to pre-fill procedures, institutions must guarantee reliability of the source, data protection, interoperability and traceability.These are the practical keys with official references.
7.1 Source reliability
- Visible versioning (date, responsible party, "last revision") on each record.
- Publish HTML summaries of calls for proposals, not just the PDF.
- Present a feed/sitemap specialized for AI (URLs, schema types,
dateModified).
7.2 Data protection (GDPR)
Avoid sending personal data to external AI services that are neither contracted nor audited. Apply the principle of data minimization and appropriate disclosure clauses.
- GDPR – Regulation (EU) 2016/679
- Explicit information and consent clauses in offices and forms.
- Impact assessments (DPIA) where appropriate.
7.3 Security (ENS)
The headquarters, back-office systems, and APIs must be aligned with the National Security Scheme.
- ENS – RD 311/2022 (organizational, operational and protective measures).
- Robust authentication (Cl@ve, certificates, institutional OAuth), encryption, event logging, and environment segregation.
7.4 Electronic identity and signature (eIDAS / eIDAS2)
In legal proceedings and relationships, identification and signature must comply with European regulations: the regulation eIDAS and its update eIDAS2 (European Digital Identity Framework).
- Regulation (EU) No 910/2014 (eIDAS)
- EUDI Wallet Framework (eIDAS2) – European digital identity and verifiable credentials.
7.5 Electronic procedure and relationship (Law 39/2015)
- Law 39/2015 – common administrative procedure and validity of acts.
- AI can guide and help with completion, but does not replace the will or signature of the interested party.
7.6 Interoperability (ENI)
- ENI – RD 4/2010 – principles and standards of interoperability.
- Publish data in open formats (JSON-LD, CSV, RDF) and catalogs (CKAN) for use by AI.
7.7 Transparency and ethics
- Public policy on the use of AI: objectives, roles, audits and logs.
- It explains when an assistant responds and how it is supervised.
- Clear channels for complaints and human review.
Innovate, yes, but with guarantees: traceability, consent, security and effective human control.
8. Closing: From Information to Citizen Trust
The public portal of the future will not be a collection of pages, but a trusted platform that engages with citizens and AI assistants. The key is not in “having more content,” but in structure, clarify and verify what people really need to live, work, start businesses and participate in the city.
From Direction & Results We help municipalities and public entities to transform their websites into
smart citizen servicesAIO diagnostics, intent-based rewriting, schema, accessibility, security, and measurement.
We start with the 10 most impactful services and leave a system that your team can maintain.
Want to evaluate your portal using AIO/Atlas criteria and a 90-day plan? Schedule a no-obligation exploratory meeting.
Quick answers to implement without losing focus AIO/Atlas.
Can an AI assistant fill out an official form for me?
Can help you to understand requirements and prepare the documentation, but the identification/signature must conform to eIDAS (and eIDAS2) already the Law 39/2015.The consent of the interested party and a valid signature are essential.
What happens if AI gives an incorrect answer about a municipal service?
Reduce that risk with clear fact sheets, visible dates and responsible parties, and structured data. Publish a feed for AI and correct outdated sources. verifiability and the update They are the best defense.
How do we measure the success of the portal's AIO adaptation?
Key indicators: time to useful response (<30s), impressions and CTR per URL in GSC, transaction completion, and reduction of repeat calls. Reviewed quarterly.
Can we use external AI models with personal data?
Only with legal bases and GDPR guarantees (Regulation 2016/679), commissioned contracts, impact assessment and ENS measures (RD 311/2022). Alternative: deploy your own instances or approved providers.
How do we start if the portal is very large and fragmented?
Attack first the 10 services with the greatest impact. Rewrite with TL;DR, tables, and FAQs; implement schema and measure. Scale afterward.
👉 In Direction & Results We help organizations simplify, integrate, and govern their digital ecosystem.
Our support combines technical analysis and business strategy so that digitization stops being a project and becomes a real competitive advantage.
Would you like us to notify you when we publish new content?
We value your time. We'll only send you articles, guides, or tools that help you improve, make better decisions, or take better action.
