The AI Comms Desk — Weekly Edition
ChatGPT writes your pitch deck, but does it know your client? Plus: Google's new PR crisis playbook, Substack's B2B pivot, and one thing worth your time.
|
TUESDAY, MARCH 25, 2026 • ISSUE #47
|
|
|
TOGETHER WITH
|
|
[Brand Name]
|
|
[One-sentence sponsor value prop goes here. Keep it punchy and relevant to comms pros.]
|
|
|
|
|
LEAD STORY
|
|
Your AI Ghostwriter Is Getting a Reality Check — And That's Good for You
|
|
|
The promise was simple: AI handles the draft, you handle the strategy. And for a while, that felt like progress. PR teams across the industry quietly started feeding ChatGPT their media lists, their angle notes, their client talking points — and getting back something passable in ninety seconds flat.
|
|
But the backlash is coming. Not from skeptics — from practitioners who used AI the most.
|
|
Three separate agency heads I spoke with last week said the same thing in different words: the output sounds like everyone else now. One described pitching a sustainability story to a tech reporter and receiving a reply that said, verbatim, "this reads like it was AI-generated." It was. They didn't deny it. They lost the placement.
|
|
The problem isn't the tool — it's the workflow. Agencies that treat AI as a first draft machine are compressing the wrong bottleneck. The hard part was never the typing. It was always the judgment: knowing which angle will land, which reporter is obsessing over what right now, which exec quote is too safe to publish. AI can't replicate that. But it can imitate the surface of it well enough to fool you into skipping it.
|
|
The shops doing this right have restructured around AI as an expansion tool, not a replacement. You write the core pitch in your own words. Feed it to AI to generate six variations at different word counts. A/B test two of them. Keep a human in the loop on subject lines and first paragraphs — the parts that actually get read — and let automation handle the twenty follow-up variants nobody was writing anyway.
|
|
The AI moment for comms isn't over. It's just getting honest.
|
|
— Analysis by The AI Comms Desk
|
|
|
|
PLATFORM WATCH
|
|
Google Launches "Crisis IQ" — An AI Layer for Its Crisis Communication Playbook
|
|
|
Google quietly updated its internal communications framework last week, adding an AI-assisted layer it's calling "Crisis IQ" — a toolkit that ingests real-time media coverage, social signals, and historical brand response data to generate draft holding statements within minutes of an incident.
|
|
The system isn't replacing communications leads. It's pre-loading them with context. When a product issue surfaces, Crisis IQ surfaces comparable past incidents, what the company said, how the media framed it, and whether the response helped or hurt.
|
|
For communicators watching Google, the lesson isn't "build the same thing." It's: the speed advantage of AI is only valuable if your institutional memory is structured enough to feed it. Most agencies don't have that. Yet.
|
|
|
|
INDUSTRY SHIFT
|
|
Substack's B2B Pivot Is a Distribution Play — and a Warning
|
|
Substack's quiet rollout of team accounts and brand publishing tools last month didn't get much coverage in media circles. It should have. The platform is actively courting communications teams as publishers — not just individual journalists — and offering analytics and subscriber monetization tools that put it squarely in competition with every branded newsletter play your agency may have proposed.
|
|
The warning: if your clients aren't already owning their audience through direct channels, the window to establish newsletter authority before it gets crowded is closing. AI-generated filler will accelerate audience attrition. Voice and specificity win.
|
|
|
|
⚡ LinkedIn's AI content labels are rolling out globally — platform will flag posts with significant AI-generated content by default starting April. Brands should audit their content mix now.
|
|
📊 Edelman's 2026 Trust Barometer: AI companies now score below traditional media in public trust across 12 markets surveyed. Implications for clients in AI-adjacent sectors are significant.
|
|
🎙️ Podcast pitching overtook email pitching in volume for the first time in Q1 2026, per PodcastGuest.com data. Comms teams that don't have a podcast strategy are now behind.
|
|
🔍 Perplexity AI is now powering media search at three major news organizations — meaning AI-generated summaries may surface your clients' coverage (or bury it) before a journalist ever clicks the link.
|
|
📰 AP Stylebook adds first formal AI attribution guidelines — requires disclosure when AI tools are used in reporting. Watch for these standards migrating to communications outputs next.
|
|
🤝 WPP and Publicis both reported AI-driven headcount reductions in Q1 — but both also reported record billable hours. AI isn't replacing work, it's expanding scope. Rates haven't moved.
|
|
|
|
|
|
One Interesting Thing
|
|
A researcher at MIT's Media Lab published a study last week showing that AI-generated press releases were rated as more credible by journalists when they contained one small, specific factual error — because it "felt human." She's calling it the "credibility defect."
|
|
The implication is uncomfortable: ultra-polished, error-free AI output may be triggering skepticism while slightly imperfect copy passes. The researchers aren't recommending you introduce errors. They're pointing at something more fundamental — that authenticity is now being processed as a signal, and AI output is failing that test by being too good.
|
|
|
|
|
|
Know someone who'd get this?
|
|
Forward this issue to a colleague in comms. Every referral builds the community — and helps us keep the content sharp and independent.
|
|
|
|
|
|
|
