Accuracy first, then speed
A press release with a wrong figure is corrected in public and remembered. AI drafts well from verified inputs (the approved fact sheet, the boilerplate, the spokesperson's notes) and badly from memory, where it fills gaps with plausible inventions. Draft from source documents only, and check every number, name, title and date against them. The same caution applies to coverage summaries: a generated media digest can misattribute a quote or merge two articles, so check anything an executive will act on against the original piece. Quotes stay human. Words attributed to a named person must be theirs, or approved by them before anyone publishes.
Press kits and media assets
A good press kit answers a journalist's questions without a phone call: the release, a fact sheet, executive biographies, approved headshots, product images with captions and usage terms, logos in vector formats, and B-roll where video matters. Generation helps with adaptation: crops for different outlets, regional versions, captions in several languages. It should not produce executive portraits or imagery that looks like documentary photography of real events. Many newsrooms will not run generated images as news photographs, and one press image later found to be generated casts doubt on everything else in the kit. Label any illustrative image as an illustration.
Crisis speed with approvals intact
Crises reward preparation. Holding statements for the likeliest scenarios, approved in advance by legal and leadership, can be adapted in minutes, and a dark site (a crisis page built beforehand and kept offline) can go live with them. AI helps summarise coverage and social conversation as it develops, and turns the approved statement into updates for each market and language, so every office says the same thing at the same time. The approval chain should get shorter in a crisis, not disappear. Name the two or three people who sign off, and their deputies, before anything happens.
Deepfakes: awareness and response
Communications teams now field fakes as well as avoid making them: cloned executive voices used in payment fraud, fabricated statements, doctored product images. Prepare a response plan with a monitoring routine, a verification step before reacting, a named spokesperson and a template denial. Make genuine material easy to verify by publishing official statements in one place and attaching C2PA Content Credentials to official images and video. The EU AI Act adds duties from 2 August 2026: deepfakes an organisation publishes must be disclosed, and so must AI-generated text published to inform the public on matters of public interest, unless it has had human editorial review under someone's editorial responsibility.
Updated 25 September 2026