Artificial intelligence in Church communication: 7 decisions before you start.

The question is no longer whether parishes, dioceses, congregations and Catholic projects will use artificial intelligence. Many already use it to summarize documents, look for ideas, draft texts, translate or adapt a post to different channels. The important question is a different one: how can we use AI to communicate better without losing truth, charism and our relationship with people?

This was one of the central issues of the 7th International Seminar on Church Communications, held in Santiago de Chile on 27 and 28 July 2026. Its discussions point to one practical conclusion: AI can expand capabilities, but the value of Church communication will keep depending on discernment, trust and encounter.

So this is not about adopting a fashionable tool. It is about making good decisions. These seven can serve as a starting point.

Before opening an AI application, it helps to answer three questions:

1. Which person are we trying to help?
2. What specific need do they have?
3. What should they be able to do after receiving our message?

A parish may need to explain how to request a baptism. A congregation, to make its charism known to young people discerning a vocation. A foundation, to mobilize donors. A diocese, to offer clear information in a sensitive situation. These are different problems, and they should not be solved with the same digital pastoral strategy nor measured with the same indicators.

At the seminar, Fr. Antonio Spadaro suggested looking at artificial intelligence beyond its rules and technical possibilities: asking what it produces inside a person, in their desires and in the way they relate to others. Applied to communication, this adds a fourth question: what kind of relationship is this content fostering?

If the goal is not clear, AI will only help us produce — faster — content that perhaps nobody needs.

Not every task carries the same human value. A simple way to organize the work is to separate two groups.

In Diego Errázuriz's workshop, AI was presented as a structure that brings scale, organization, translation and segmentation, while the person brings vision, meaning and accompaniment. It is a useful distinction: automate the logistics; do not automate the relationship.

Good use of AI should not fill the team's calendar with even more posts. It should free up time to listen, answer and accompany better.

Matthew Harvey Sanders explained two common limits of general-purpose assistants: they can invent plausible information and they do not always show where an answer comes from. His experience with Magisterium AI points to a clear principle: first locate the relevant sources, then build the answer from them.

This has a direct consequence for any Catholic institution: if official information is not published, updated and well organized, it will be harder for people — and also for search engines and AI assistants — to find a correct answer.

Instead of limiting the website to institutional news, it is worth creating content that answers real questions:

• What do I need in order to have my child baptized?
• How can I collaborate with this work?
• What does a religious community do and how is a vocation discerned?
• Where can I ask for help or accompaniment?
• What is the institution's position on a specific issue and what sources support it?
• How do I sign up for an activity?

Every page should include a direct answer, clear steps, the date it was updated, the geographic scope when relevant, a contact person or channel, and links to official sources. This is the proper work of web design and development and of campaigns and SEO.

This serves people, improves search rankings and increases the chance that an AI-generated answer will find reliable information from the institution itself. SEO thus stops being a technique for attracting visits and becomes a service: being present with a clear answer when someone needs it.

The fluency of an answer does not guarantee that it is true. No text should be published simply because it "sounds good". That is the difference between generating text and doing content production with judgment.

A safe workflow can look like this:

1. Select the sources. Official documents, your own data, interviews, authorized testimonies and up-to-date materials.
2. Ask AI for a draft. Give it context: audience, purpose, tone, length and what must not be said.
3. Verify everything that can be verified. Names, dates, figures, quotations, doctrine and links.
4. Rewrite in your own voice. Recover the charism, the vocabulary and the closeness of the institution.
5. Have it approved by a responsible person. Someone must sign off on what is published.

When it comes to doctrine, minors, personal testimonies, crises, mental health, abuse, private data or pastoral decisions, review must be especially rigorous. AI has no conscience, no benevolent intention and no moral responsibility. It may be competent at producing an answer; it cannot will the good of the person receiving it.

Juan Narbona pointed to an opportunity for Church communicators: moving from a paradigm centred on producing and distributing content to one centred on relationship. AI can take on part of the mechanical work, but the time gained should be spent listening to the context, understanding people and creating closeness. That is the heart of work in social media and storytelling.

This changes the way we plan.

Instead of only asking "what do we publish this week?", a team can ask:

• Which doubts keep coming up in the messages we receive?
• Who are we failing to help?
• Which conversation needs a personal reply?
• Which content should lead to a meeting, a call, a sign-up or a community?
• Do we answer as quickly as the need requires?

Social media is conversational. Asking for comments and not replying to them creates frustration. Publishing a phone number nobody answers or a form with no follow-up also communicates something. Closeness does not depend only on the tone of a text: it is proven in the quality and speed of the response.

AI can classify enquiries, identify frequent topics or prepare an initial draft. The sensitive reply, the judgment and the bond belong to people.

In the panel on media and public scrutiny it was recalled that credibility is built slowly and can be lost quickly. It was also clear that a crisis is not resolved by multiplying messages or by treating journalists or the audience as adversaries.

For a Catholic organization the demand is greater, because people expect coherence between what it proclaims and what it does. The communicative response must begin with the truth of the facts, the acknowledgement of responsibilities and a real willingness to repair. This is the terrain of strategic communication and media relations.

Before a crisis, it helps to have:

• identified internal sources of information;
• data and documents in order;
• clear approval responsibilities;
• a prepared spokesperson;
• a protocol for sensitive questions;
• active listening channels;
• an informed community, not one used as a shield.

During a crisis, the metrics workshop proposed a strikingly simple rule: one voice, one clear and coherent message. Artificial intelligence can help gather questions, classify reactions or detect topics, but it must not decide the institutional position or manufacture an automatic reply for someone who has been harmed.

Transparency does not mean telling everything without judgment. It means not hiding what matters, distinguishing facts from interpretations, acknowledging what is not yet known and updating information when it changes.

Guillermo Bustamante insisted that you cannot properly measure a piece of content without first knowing what it was meant to achieve. Followers, impressions or likes only make sense within a goal.

A basic table can help:

• Make it known

  Are we reaching the right people?

  Qualified reach, visits from relevant searches, new users

• Build community

  Do people take part and come back?

  Useful comments, shares, subscriptions, returning users

• Serve

  Do they find and complete what they need?

  Contact clicks, completed forms, sign-ups, response time

• Engage in dialogue

  Is there a real conversation?

  Questions answered, response speed, recurring topics, quality of exchange

• Rebuild trust

  Is the change understood and is uncertainty decreasing?

  Questions resolved, evolution of topics and sentiment, fewer repeated mistakes

Quantitative measurement must be completed with qualitative listening. Knowing that a page received a thousand visits does not explain whether it resolved the question. Reviewing enquiries, comments and conversations reveals what is missing, what is misunderstood and what new content is worth creating.

There is no need to transform all your communication at once. It is better to start with one concrete problem and a process that can be reviewed.

Week 1: listen and choose

• Gather the questions received over the last three months.
• Group them by topic and type of person.
• Choose a need that is frequent, important and narrow enough.
• Define the result you expect and how you will measure it.

Week 2: organize sources and create the answer

• Collect the official documentation and check that it is current.
• Interview the person responsible for the topic.
• Write a web page that answers directly and practically.
• Add steps, contact details, the update date and the sources.

Week 3: adapt and converse

• Use AI to propose versions of the content for email, social media, short video or FAQs.
• Review each version and restore the institution's own voice.
• Publish on the channels where your chosen audience actually is.
• Set aside time and named people to reply.

Week 4: measure and learn

• Check whether the expected audience arrived.
• See what actions they took.
• Read and classify the new questions.
• Correct the content and document what you learned.
• Decide whether to extend the pilot to another topic.

• It answers a real need of a specific person.
• It has a purpose and an expected action.
• It was produced from identified, up-to-date sources.
• Names, data, quotations, doctrine and links have been verified.
• It respects privacy, dignity and image or testimony permissions.
• It reflects the charism and tone of the organization.
• A responsible person has reviewed and approved it.
• There is capacity to handle the responses it generates.
• It will be measured with indicators tied to its purpose.

Artificial intelligence offers a real opportunity for teams with little time and many fronts: better research, organized information, repurposed content, translation, segmentation and testing. But adopting it will only make sense if it helps us communicate with more truth, more clarity and more closeness.

The goal is not to look more technological. The goal is that a person finds the answer they need, recognizes an authentic voice and can take a step towards a real community.

At Catholic Connect we help parishes, dioceses, congregations and Christian projects turn their mission into a concrete communication strategy: from listening and diagnosis to the website, content, social media, campaigns and measurement.