The Roundup

INTELLIGENCE

ProPublica Union Authorizes First U.S. Newsroom Strike Over AI Protections

ProPublica's editorial staff voted to authorize the first U.S. newsroom strike specifically over AI protections. The ProPublica Guild, representing about 150 staff at 92% participation, is demanding a ban on AI-related layoffs and "just cause" firing protections before management implements AI tools. For comms pros who pitch investigative outlets: expect turbulence. Watch whether this contract language becomes a template for newsroom AI governance nationwide.

INTELLIGENCE

White House Releases AI Legislative Framework: Sector-Specific, No New Regulators

The White House released a seven-point AI policy framework on March 20, recommending Congress not create a new federal AI regulatory body. The framework favors sector-specific oversight via existing agencies and affirms that AI training on copyrighted internet content is not a copyright violation. No new agency means no single point of contact for AI policy comms. Clients in regulated industries will need messaging that reflects their sector's existing regulator.

FROM THE FIELD

PRSA's March briefing flagged a shift worth attention: AI is now an audience, not just a tool. Comms professionals are being advised to write press releases, bios, and boilerplate so clearly that AI systems cannot misinterpret them. Clear attribution. Unambiguous claims. Verifiable facts.

Try it: Paste your most recent client press release into ChatGPT and ask: "What does this company do and why does it matter?" Compare the answer to your messaging brief. The gaps are your GEO audit.

FROM THE FIELD

Meta's Ads Manager now has an AI agent that reviews performance data and recommends campaign changes. It launched fast, within two months of the Manus acquisition. Practical implication: paid social and earned media are converging in Meta's ecosystem. Expect clients to start asking why your earned results do not show up the way their ad analytics do.

Try it: Ask your next client contact what Meta's Ads Manager AI is recommending. Use the answer to sharpen your earned media messaging alignment.

THE WORKFLOW

How to Audit Your Brand's Presence in AI Search (In Under an Hour)

AI search is now a coverage beat. Google AI Overviews appear in more than 25% of all Google searches, up from 13% a year ago. ChatGPT processes roughly 2 billion queries a day. Perplexity handles 30 million daily active queries. If you have not checked how your brand shows up in these answers, someone else has.

This audit takes 45 to 60 minutes. Do it once a quarter.

What you need:

A browser (incognito mode). ChatGPT, Google Gemini, and Perplexity (free tiers work). A spreadsheet to record findings. 15 to 20 seed questions about your brand, industry, and competitors.

Step 1: Build your query list (10 min)

Write 15 to 20 questions an informed buyer or journalist might ask. "[Your brand]: what do they do?" "Best [your category] tools." "[Competitor] vs [your brand]." Avoid vanity queries. Think: what would someone type if they were evaluating vendors or writing a story?

Step 2: Run queries across all three platforms (20 min)

Open ChatGPT, Gemini, and Perplexity in separate tabs. Paste each query. Record: Does your brand appear? In what context? What sources are cited? What competitors appear when you do not? Note the exact language AI uses to describe you. If it is wrong, it is the new version of a bad Google snippet, except it travels further.

Step 3: Flag the gaps (10 min)

Three categories. Missing entirely: your brand does not appear for queries where it should. Content gap. Inaccurate: AI describes you incorrectly or with outdated info. Source problem. Weak positioning: you appear, but buried or vaguely described. Visibility problem.

Step 4: Identify which sources AI is citing (10 min)

For every mention of a competitor or industry term, look at what sources the AI pulls from. You will typically see Wikipedia, industry publications, well-structured company pages, and recent earned media. These are your targets.

Step 5: Build a fix list

For each gap: one content action. Missing entirely? Write a clear, factual "what we do" page with structured data. Inaccurate? Pitch a correction or profile to the publications being cited. Weak positioning? Build a briefing document and get it in front of journalists who cover your space.

Watch out for:

Running queries only once (AI outputs vary by session, run each at least twice). Assuming your website is being read (AI pulls from external sources more than your .com). Treating this as a one-time task (set a quarterly calendar reminder).

TOOL SPOTLIGHT

Muck Rack's Media List Agent: AI-Powered Pitch Targeting

What it claims: Muck Rack launched Media List Agent in beta on February 2. It takes a pitch (or pitch description) as input and recommends journalists to add or remove, with explanations grounded in Muck Rack's editorially verified database. Up to 50 journalist recommendations per prompt. It connects directly to Muck Rack's Pitch Placement Impact Widgets so teams can track which pitches led to actual coverage.

Performance data: Beta customers saw an average 10% increase in pitch open and click rates versus pitches built without the agent.

Who it is for: PR and comms teams already using Muck Rack. The agent lives inside the platform, not as a standalone tool.

Pricing: Not publicly listed. Muck Rack's advanced tier is "Premier" with custom pricing. Free demo/trial available.

The honest take: This is a real, measurable improvement over manual media list building. The 10% boost in open/click rates is credible for beta data. The limitation: English-language pitches only. The agent recommends journalists but does not send outreach autonomously. Human approval required for every action. If you are already on Muck Rack, activate it. If not, request a demo and test it against your current process.

THE ANGLE

The Real Story of ChatGPT Ads Is Not the Ads

OpenAI charging $60 CPM for ChatGPT ads is a headline. But the real story is what happens around the ads.

When a user asks ChatGPT about a product category, the answer it generates comes from earned media, Wikipedia, industry publications, and authoritative third-party content. Not from the brand's own website. Not from paid placements. From the work that communications professionals produce.

For years, the PR industry has fought to prove that earned media drives business outcomes. Now there is a $200K-minimum ad product literally sitting next to AI-generated answers that are shaped by earned coverage. That is the proof of value the industry has been asking for. Earned media is not just a trust signal anymore. It is the source material for the most expensive new ad channel on the internet.

The comms teams that will win this moment are not the ones scrambling to buy ChatGPT ads. They are the ones making sure that when ChatGPT answers a question about their client, the answer is correct, current, and favorable. That is a PR job. It always has been.

PROMPT LAB

Stress-Test Your Pitch Angle Before You Send It

You have a pitch. It feels solid. But "feels solid" is not a reporting criteria. Before you hit send, run this prompt.

The prompt:

You are a skeptical [beat] journalist. I'm about to pitch you a story. Your job is to push back hard. Here is my pitch angle: [paste 2-3 sentences] Tell me: 1. What's not new about this? 2. What's the obvious counter-argument? 3. Who else is saying the exact same thing? 4. What would make this actually worth covering? Be direct. Don't soften it.

Why it works:

Most pitches die because they do not anticipate the journalist's first objection. This prompt forces you to hear that objection before you send the email. It simulates the mental model of a beat reporter who has covered your space for years.

Before:

After running the prompt:

Now you know what to add before you hit send.

THE READ

Axios just laid off 11 staffers to automate tasks and is doubling down on subject matter experts, while signing content licensing deals with OpenAI. For PR pros, this is a signal: the outlets you are pitching are restructuring around expertise and AI, which means your pitches need to match that standard or they will not land.

Know someone who should read this?

Forward this issue to a colleague in comms. Every referral builds the community and keeps the content sharp.