For PR and Communications Managers
Cognitive Sovereignty Checklist for PR Managers
AI tools like ChatGPT and Cision AI can write press releases that are accurate and grammatically clean. They cannot read the moment. They cannot sense when a journalist is fishing for a different story than the one you are offering. They cannot decide whether silence or response is the right move in a crisis. Your job is to keep your own instinct sharp while these tools do the labour.
These are suggestions. Take what fits, leave the rest.
0 / 19 applicable
Tap once to check, again to mark N/A, again to reset.
Press Releases and Narrative Strategy
Read the AI draft aloud before it goes to journalistsbeginner
AI-written press releases often lack cadence. Journalistss feel this flatness even when they cannot name it. Reading aloud will show you where the narrative loses energy or where a claim sounds defensive rather than confident.
Ask which journalists will actually care about this storybeginner
Before you send any press release, name three specific journalists and explain why each one would run this. If you cannot do this, the story angle is not clear enough. AI does not have journalist relationships and cannot answer this question for you.
Identify what competing story the AI draft ignoresintermediate
Every press release is an argument against some other narrative. The AI tool will not see this. Ask yourself what story your competitors or critics will tell about this announcement. Your release should address it without appearing defensive.
Change the lead if it matches the AI template too closelyintermediate
Most AI tools follow the inverted pyramid. Sometimes the news is not in the first paragraph. If you can predict the opening sentence from other releases in your industry, rewrite it. Journalistss notice.
Test the release for tone mismatch with your brand voiceintermediate
AI tools trained on thousands of corporate releases will write you a corporate release. This may not match how your organisation actually sounds in conversation or in previous communications. Audit your last five successful press releases. Find the specific words or sentence patterns that made them yours.
Keep a list of which outlets have misinterpreted your previous releasesadvanced
These outlets need different information architecture. AI will not know this. When you customise releases for these journalists, you are not adding extra work. You are preventing your own miscommunication.
Write the headline yourself, then let AI draft the bodyadvanced
Your headline commitment forces you to decide what the story actually is before you generate text. This prevents AI from writing a release that is defensible but off-target.
Media Monitoring and Crisis Signal Detection
Record what you noticed in Meltwater summaries that were wrongbeginner
When an AI summary misses tone, write down the original quote and what the summary said. Over time you will see the patterns AI cannot catch. This becomes your early warning system.
Read the full article, not just the AI summary, if a journalist covers you negativelybeginner
AI summaries compress criticism into bullets. They lose the writer's uncertainty, the counterarguments they included, and the damage they could have done but did not. You need to read the original to judge whether you should respond.
Track which monitoring alerts actually turned into problemsbeginner
Most alerts will not. Over six months, mark which ones mattered. You will see that AI catches volume but not impact. Use your own record to tune which alerts you actually read.
Notice when sentiment analysis contradicts your own readingintermediate
When Cision labels a mention positive but it felt backhanded to you, trust yourself. Sarcasm, embedded criticism, and social pressure are not sentiment buckets. Your instinct here is real data.
Listen for what the journalist is asking you to explain before you explain itintermediate
Media monitoring AI tells you what was said about you. It does not tell you what a reporter is investigating. Phone calls and email threads contain pressure points. Spend your time reading those carefully.
Compare AI summaries to the conversation you had with that journalist last monthadvanced
If a reporter's angle has shifted, that shift matters. AI summaries treat each mention as new. You remember the story they were pursuing. This memory is part of your job.
Crisis Response and Moment Judgement
Do not use the AI draft for the first response in a crisisbeginner
The first communication sets the emotional tone. AI will give you the safest option. Safe is often silent. Your first statement should show your organisation understands what went wrong before explaining what you will do.
Ask whether the crisis is about what happened or about what it meansintermediate
AI drafts address facts. Crises are often about meaning. A safety incident means the organisation does not care about people. A data breach means the organisation was careless with trust. Address the meaning first.
Write a response that would make you uncomfortable if you had to stand behind it on a news broadcastintermediate
If the language feels fine in a press release but sounds hollow when you imagine saying it to a journalist on camera, rewrite it. Your discomfort is a signal that the response is not authentic.
Decide whether you are responding to what was said or what you fear will be said nextintermediate
Crisis AI drafts often defend against the worst possible interpretation. This makes the response sound paranoid. Respond to what has actually been said. If you need to get ahead of a narrative, do that in a separate statement.
Compare the AI draft to your organisation's actual values statementadvanced
Crisis responses that contradict your published values will be caught. AI does not read your values carefully. Before you send any crisis statement, check it against what you have publicly committed to doing.
Identify which stakeholder the AI response is written for and whether that is the right oneadvanced
An AI crisis draft might be written for lawyers, for the general public, or for employees. You need to send different statements to different audiences at different times. Know whose trust you are trying to keep first.
Five things worth remembering
- Keep a file of your best press releases from the last two years. When AI generates a draft, compare it sentence by sentence to your own work. Your past writing is your benchmark.
- When a journalist calls with a question, answer it first yourself before you check what an AI monitoring tool said they were looking for. Your read of the conversation is more accurate than sentiment analysis.
- In a crisis, write your own statement first. Then show the AI draft to yourself and ask which one you would defend in a meeting with your CEO. The answer tells you what is missing.
- Track the names of journalists who have run your stories twice or more. These relationships are your real asset. No press release matters if the journalist will not read it. AI cannot build these relationships for you.
- Set a rule that you read the full article or transcript before you act on any media mention. Five minutes of your time prevents hours of crisis response to a misunderstanding.
Related reads
Common questions
Should pr managers read the ai draft aloud before it goes to journalists?
AI-written press releases often lack cadence. Journalistss feel this flatness even when they cannot name it. Reading aloud will show you where the narrative loses energy or where a claim sounds defensive rather than confident.
Should pr managers ask which journalists will actually care about this story?
Before you send any press release, name three specific journalists and explain why each one would run this. If you cannot do this, the story angle is not clear enough. AI does not have journalist relationships and cannot answer this question for you.
Should pr managers identify what competing story the ai draft ignores?
Every press release is an argument against some other narrative. The AI tool will not see this. Ask yourself what story your competitors or critics will tell about this announcement. Your release should address it without appearing defensive.