AI-assisted political writing is increasingly becoming part of the campaign communications stack, not as a replacement for speechwriters, but as a way to multiply their output across channels, candidates and languages.

How European Political Campaigns Are Using AI

Some of that movement has been thoughtful. In 2026 the European Democratic Party in Brussels unveiled what it called the world's first political advert built entirely with artificial intelligence, a short film of digitally generated actors calmly reciting lines like "I don't vote, I'm not real" and "workers' rights mean nothing to me." 

The point was not to persuade anyone of anything specific. It was to unsettle people just enough to think about what it means when a convincing face says something no real person actually believes. Used that way, AI generated content becomes a way of interrogating the technology in public rather than quietly hiding what it is.

When AI-Generated Political Content Goes Wrong

Not every experiment has landed as cleanly. Malta's Partit Nazzjonalista launched an AI chatbot specifically to walk voters through its manifesto, a genuinely useful idea on paper, since most people will never read a manifesto cover to cover but might well ask a direct question about housing or health policy. 

In practice, the bot generated inaccurate visuals and misleading content, and the party had to switch off its image generation function to stop the damage. The chatbot was not lying on purpose. It was doing what these systems do when nobody has checked their output closely enough, which is produce something fluent, confident and wrong.

That gap between fluent and correct is the real story here, more than any grand claim about AI replacing speechwriters. A hallucinated statistic in a printed leaflet is embarrassing and gets fixed in the next print run. 

The same hallucinated statistic delivered by a chatbot, in a warm conversational tone, to a voter who asked a direct question and got what felt like a direct answer, carries an authority the leaflet never had. People do not fact check a conversation the way they fact check a document, because a conversation does not feel like a document. It feels like being told something by someone who knows.

There is a larger trend running underneath both examples, which is that European parties are turning to AI partly because they have no choice. Party membership has been declining across most of the continent for two decades, which means fewer volunteer hands available to write the fifteenth localised version of a policy explainer nobody particularly wanted to write in the first place. 

AI drafting tools are genuinely useful for that unglamorous, repetitive layer of political writing, the shorter version for a text message, the one pitched at a specific age group or region, the routine reply to a routine constituent question. Used there, the technology behaves less like a threat to anyone's job and more like a very fast junior staffer.

What Political Campaigns Should and Shouldn’t: Automate

The risk is not that AI writes political content. It is that the range of unglamorous, repetitive writing AI can plausibly handle quietly expands to include material that actually needed a person's judgement, a specific factual claim, a policy position with real consequences, a line that will be read back to a candidate in a debate six weeks later, without anyone deciding that expansion on purpose. It happens by drift rather than by decision, one convenient shortcut at a time, until an organisation's most sensitive public statements are running through the same drafting pipeline as its least sensitive ones.

In practice the line worth holding usually falls in the same few places. Anything with a number in it that could be checked. Anything that commits a candidate to a specific promise rather than a general value. Anything likely to be quoted back at them later, in a debate, an interview or a fact check. And anything going out under a real person's name rather than an organisation's collective voice. Draft everything else with AI and barely anyone will notice or mind. Draft any of those four categories without a human reading it first, and sooner or later somebody will.

That tension, between what AI can draft and what it should be trusted to say unsupervised, is exactly the ground covered by a session on AI assisted speechwriting and manifesto writing at the first ever Political Communication Summit this October bringing together more than 500 professionals from campaigns, institutions, media and civil society for this inaugural edition, held under the patronage of Vienna's Deputy Mayor.

If your communications team is already experimenting with AI, the question is where automation should stop. The Political Communication Summit brings the people making those decisions together on 7 October at Vienna City Hall including a dedicated session on AI-assisted speechwriting and manifesto writing.