30 Practical Ideas for PR Managers to Stay Cognitively Sovereign
AI can draft a technically perfect press release that no journalist will read. It can summarise media coverage with accuracy but miss the journalist's tone that signals whether they are building a story or killing one. The risk is not that AI does PR work poorly. The risk is that you stop doing the work that made you good at PR in the first place.
These are suggestions. Take what fits, leave the rest.
Read three full journalist bylines every Monday morning before opening your AI drafting toolbeginner
Pick journalists who cover your sector. Read their last three pieces without skimming. Notice what angle they chose, which quotes they led with, what they buried, what surprised them. This trains the eye that AI summaries cannot replace.
Write one press release entirely by hand before using AI to draft the nextbeginner
Handwrite or type without prompts. Feel the difference between what you naturally emphasise and what the AI model prioritises. You will notice what you lose when you skip this step.
After AI generates a press release, identify the one sentence a journalist would never usebeginner
Read the AI draft. Find the line that sounds like a press release rather than news. Rewrite it from the reporter's perspective. This is where human judgement lives.
Call three journalists each quarter to ask what made them actually cover a story from your sectorintermediate
Not to pitch. To listen. Ask what surprised them, what felt overcooked, what angle they were waiting for someone to give them. Store this as your real media monitoring system.
Create a file of press releases that failed and analyse whyintermediate
Keep releases that journalists ignored despite good news, good quotes, good timing. Mark where the narrative went wrong. Train yourself to spot what your AI tools cannot: the moment when a release stops being news.
Test AI press releases with one trusted journalist before distributionintermediate
Send the AI draft to a contact you trust. Ask if it reads like news or like someone trying to sell them something. Use their feedback to train your eye on what the tool misses.
Write a two sentence brief on the narrative strategy before prompting AI to write anythingintermediate
Do not tell AI what story to tell. Tell yourself first. What is the actual news here. Why would someone care who has no stake in your organisation. Write this down. Then check if the AI draft matches your strategy or drifts from it.
Keep a media sentiment file separate from your AI monitoring tooladvanced
When you notice a tone shift in coverage before it becomes obvious, write it down with the date and publication. Track your own instinct against what AI flagged as neutral or positive.
Spend thirty minutes each month reading coverage of your competitors' announcementsintermediate
See which competitor releases got coverage and which did not. See which angles worked. This cannot come from an AI summary. You have to read the story yourself to feel what made it land.
Document one case each month where your narrative instinct contradicted what the AI recommendedundefined
Note what you chose instead. Track whether it worked. This is how you build evidence of what your judgement catches that the tool does not.
Staying Sharp on Media Relationships and Signals
Schedule a face to face coffee with one reporter each monthbeginner
Not a media meeting. A conversation. You will pick up tone and interest that video calls and email miss. You will know whether a story is coming before it appears.
Read the comments section and social replies on coverage of your sectorbeginner
AI monitoring tools miss what readers are saying about stories. Read them. This tells you whether a narrative landed or fell flat with the actual audience.
When AI flags a mention as neutral, read the full article before accepting that ratingbeginner
Read the context around your mention. Is it in a positive section or a criticism. AI scores sentiment. You read direction and damage.
Create a personal list of ten journalists who cover your space and track their recent anglesintermediate
Not from AI. Update it monthly by reading their bylines. Know what story each one is building. Know when to pitch to them and when to wait.
When a story about your organisation or sector breaks, read at least five versions before formulating a responseintermediate
Different outlets frame the same story differently. Read them all. Feel the momentum. This is what tells you whether crisis response should be aggressive or measured.
Track which journalists follow your organisation on social media and which ignore itbeginner
This tells you who is interested in what you do versus who views you as off their beat. Adjust your outreach accordingly. AI cannot make this distinction.
Ask your journalist contacts directly what they wish you would stop doing in press releasesintermediate
Ask what format works. Ask what timing works. Ask what angle they are tired of. Use this to brief the AI tool on what this particular beat actually wants.
When a competitor gets significant coverage, reach out to that journalist to ask what angle workedadvanced
Not to pitch yourself. To learn. What made them choose that story. What would make them choose your story. Build your own media intelligence.
Keep a running file of which publications and journalists cover which types of your announcementsintermediate
Do not rely on AI to figure out your media list. Build it yourself over time. Know that Product News goes to Tech Reporter and Culture News goes to Features Reporter.
Before a major announcement, talk to three relevant journalists off the record to test the angleadvanced
Do not pitch the story. Ask if the angle matters to them. Ask what they would want to know. Shape your announcement based on what you hear, not what AI predicts will work.
Defending Judgement in Crisis and High Stakes Moments
Write a crisis response plan that documents your instinct before AI is allowed to draft anythingbeginner
Before you ever use AI in crisis, document what your organisation's values require you to say. What cannot be compromised. What timing matters. This becomes your guardrail against AI taking shortcuts.
In any crisis situation, spend fifteen minutes reading raw social media and news before asking AI to summarisebeginner
Get the unfiltered version first. Feel the actual mood. Then check whether AI's summary of the situation matches what you saw with your own eyes.
When AI drafts a crisis statement, rewrite the opening sentence from scratchintermediate
The first line determines whether people believe you or dismiss you. Do not trust AI to get this right. Make it reflect your actual position, not a templated response.
For any crisis response, identify what the opposing side will say next and prepare for that before publishingintermediate
AI drafts what you want to say. You need to think three moves ahead about what critics will say in response. This requires human judgment about how arguments move.
Call your legal and leadership teams before using an AI drafted crisis responsebeginner
Do not automate this conversation. Speak to the people who understand organisational risk. AI cannot weigh what your organisation actually can and cannot do.
In a crisis, assign one team member to monitor live coverage every hour instead of relying on AI alertsintermediate
Coverage moves fast. AI summaries come on delay. You need a human reading the story as it develops to spot what angle is winning and what is dying.
After you issue a crisis response, track coverage for three days by hand reading, not AI reportsintermediate
See whether your statement moved the conversation. See which parts journalists quoted and which they ignored. This tells you what worked beyond what AI metrics show.
Create a decision matrix for crisis moments that does not include AI recommendationsadvanced
Document: what triggers a holding statement, what triggers a full response, what triggers silence, what requires a named spokesperson. This is your human judgment baseline.
For major crises, have one human draft a response without consulting AI, then compare it to the AI versionadvanced
You will see immediately what the tool adds and what it removes. This trains you to spot where AI makes you sound distant when you need to sound human.
After any significant crisis response, conduct a debrief that asks what human judgment prevented the AI version from being wrongadvanced
Document the specific moments where a human choice changed the outcome. Build institutional memory of why your instinct mattered. Share this with your team.
Five things worth remembering
Your relationship with journalists is your real media monitoring system. AI tools are supplements to it, not replacements. Invest in the relationships and treat the tools as research assistants.
Press releases written by AI sound like press releases because they optimise for comprehensiveness, not for the single angle that makes a story newsworthy. Always start with narrative strategy, not with the tool.
In crisis situations, the first human response usually outperforms the carefully drafted AI version. Draft fast with human judgment. Polish it with AI if you have time. Never do it the other way around.
Train your team to finish every AI output by asking: would this get coverage if we sent it tomorrow. If the answer is no, the AI did its job correctly by showing you the gap between what you want to say and what journalists actually want to hear.
Keep a crisis war room notebook separate from all AI tools. Handwrite the key decisions and who made them. This becomes your evidence later of where human judgment prevented mistakes.