Most communications teams already use artificial intelligence (AI). To draft content, summarise reports, adapt messages for different channels, or speed up processes that used to take days. What very few have is a clear AI editorial policy for communications teams: a set of shared rules on what is permitted, who reviews what, and how that use is disclosed.
That absence is not neutral. It is a decision, and it has consequences.
Because the question that few organisations have yet considered is not whether to use AI, but who is accountable when something goes wrong. Who decides what the AI can produce on behalf of the organisation? Who gives the final approval before it goes out? And under what criteria are those decisions made, beyond the instinct of each person on the team on that particular day?
That is organisational responsibility, and in corporate communications, its absence becomes apparent sooner than you might think.
The dominant conversation about AI in communications tends to focus on capability: what it can write, translate, summarise or generate. But the more difficult conversation is about what happens when something goes wrong, and no one has decided on the rules.
When a communication team works without a shared policy on AI use, predictable things happen. Each member uses it differently, without a common standard for review or approval. Content is published that implicitly carries the institutional voice, without anyone having deliberately authorised it. And when an error occurs (a hallucination, incorrect data, a tone that does not represent the organisation well), there is no clear framework of accountability to fall back on.
An analysis of AI policies across 52 newsrooms found that only 8% of the documents studied specified how they were to be implemented. The rest set out principles without explaining who verifies them or what happens when they are not met. A policy without implementation is a statement of intent, not a governance tool.
This is not a problem exclusive to journalism. It is also an issue in institutional communication. And it is not theoretical: the transparency obligations under the EU’s AI Regulation, concerning AI-generated content published for the purpose of informing the public, come into force in August 2026. For organisations that communicate publicly, such as NGOs, public institutions, international bodies or advocacy organisations, this is a regulatory deadline, not a distant consideration.
It doesn’t look like chaos. It looks like this: every team member makes their own decisions, without a shared framework, without a paper trail, and without an agreed line between what the AI drafts and what a person claims as their own. The risk isn’t a single, dramatic failure. It is the slow erosion of something harder to recover than a mistake: institutional credibility.
A useful AI editorial policy doesn’t have to be a lengthy document. It needs to answer four practical questions clearly.
It is not about banning tools. It is about being explicit. Which uses are approved? (drafting, summarising, translation, research support). Which require additional review? (public-facing content, sensitive topics, content attributed to specific voices). Which are ruled out? (content that cannot include unverified AI-generated data, content involving personal data).
Organisations that have developed effective AI policies typically divide usage into three categories: audience-facing, internal operational, and administrative or business, each with its own level of scrutiny. The distinction matters because the risks are not the same in all three cases.
Human review is not a bureaucratic formality. It is the mechanism that maintains institutional accountability. The question is not whether to have human review, but at what stage and by whom.
A minimum viable approach: any AI-assisted content published under the organisation’s name requires the approval of a specific individual who takes responsibility for its accuracy and tone. That person is not approving the tool; they are approving the result as if it were their own work. The principle that runs through all organisations that have developed AI guidelines is the same: the chain of thought and decision-making begins and ends with people, with AI serving as a supporting tool.
Traceability does not mean declaring the use of AI in every social media post. It means keeping an internal record: which tool was used, for what purpose, who reviewed it and when. This protects the organisation in two ways: externally, if questions arise about a specific piece of content; internally, if the policy needs to be reviewed based on actual usage patterns.
Article 50 of the EU AI Regulation specifically requires that text published for the purpose of informing the public on matters of general interest be declared as AI-generated or AI-assisted where applicable. Beyond the legal obligation, there is a strategic reason to declare it: audiences increasingly notice the absence of a declaration more than the declaration itself.
There is a version of AI transparency that is purely defensive: a legal notice buried in a footer, a policy document sitting on a server that nobody opens. That version meets the formal requirement, but does almost nothing for trust.
The version that works is different. It treats transparency as information the audience needs to understand your message, not as a legal safeguard for your organisation.
That means being specific rather than vague. “This content has been drafted with the assistance of AI and reviewed by [name/title]” tells the reader something useful. “We may use AI tools in our communications” tells them nothing.
The Poynter Institute, which published its first AI ethics guidelines for newsrooms in 2024 and updated them significantly in 2025, has noted that demand for practical tools has grown substantially as AI becomes integrated into daily workflows. The leap from principles to practice is exactly where most organisations get stuck.
Before publishing any AI-assisted communication, ask yourself: if the audience knew exactly how this was produced, would it make it less credible? If the answer is yes, the problem isn’t the statement: it’s the production process. Transparency is diagnostic, not cosmetic. It forces you to answer the question your policy should have already resolved.
The most common reason why communications teams lack an AI policy is not ideological. It is a question of capacity. Developing a meaningful AI policy can take months and requires cross-departmental collaboration – a challenge for teams operating with limited staff and competing priorities.
The answer is not to wait until you have the capacity for a comprehensive process. It is to start with a minimum viable policy that your team can actually follow, and build from there.
Decision 1: Permitted uses. Draw up a list of what is explicitly approved for AI use in your team’s work. Keep it short and specific. If something isn’t on the list, it requires a discussion before proceeding.
Decision 2: The review rule. Define who reviews AI-assisted content before it goes out, and what that review entails. Not a cursory read-through: a check for accuracy, tone and consistency with the institutional voice.
Decision 3: The disclosure standard. Agree on what you disclose, in what format and where. Make it specific enough that anyone on the team can apply it without having to ask every time.
An AI policy written in 2025 will need to be reviewed in 2026. Tools change. The regulatory context changes: the transparency obligations of the EU AI Regulation alone will require updates by August 2026. Around 63% of the AI policy documents examined in large communications organisations specified that they would be updated at some point, but only 6% committed to a specific interval. Setting a review date at the time of adoption is the simplest way to prevent a policy from becoming obsolete without anyone noticing.
A policy tells people what to do. But in institutional communications, the situations that matter are rarely the obvious ones: they are the borderline cases, the ambiguous formats, the moments when “is this AI-generated content?” is genuinely uncertain.
The UNESCO Recommendation on the Ethics of AI, adopted by 194 member states, frames human oversight not as a safeguard against AI, but as an expression of organisational values. That perspective deserves to be internalised: the aim is to have a team that exercises judgement on what their organisation should communicate and how.
That judgment cannot be outsourced to a policy document. It is developed through practice, conversation and the occasional awkward situation. A policy creates the conditions for that development.
The question facing communications teams is not whether to use AI, but how to use it. Most already do, and there are good reasons for this. The question is whether that use is governed: whether there are shared rules, clear accountability, genuine human review and honest transparency about how content is produced.
Without that governance, the risk isn’t a single, visible failure. It’s the gradual erosion of something that takes years to build and is very difficult to recover: the institutional trust that makes communication matter.
If you’re working on an AI policy for your communications team or are wondering where to start, I’d love to hear where you’re at on LinkedIn. What’s the hardest part: getting internal buy-in, the wording of the policy, the actual implementation, or simply finding the time? I look forward to hearing from you.
Institutions keep designing messages for public feeds while their audiences make decisions in WhatsApp groups no institutional algorithm can reach. The problem is not technical. It is a question about how trust works when public visibility disappears and what remains are relationships.
When a deepfake involving your organisation surfaces, the first instinct is to treat it as a technical emergency. But what determines the real damage is not the fake content itself — it is what your organisation does, or fails to do, in the hours that follow. That is a communication decision.
A story can work and still cause harm. This practical guide helps you gather and publish testimonies with real consent, context and dignity, without turning people into “communication material”.