This week, beehiiv launched something that sounds deeply nerdy but is actually a pretty big deal for anyone who publishes a newsletter: an MCP.

If you've been avoiding AI in your day-to-day work, MCP stands for Model Context Protocol, and if you haven't encountered the term yet, you will. It's an open standard that lets AI tools like Claude, ChatGPT, and Gemini connect directly to the apps you already use. So you can essentially let your AI buddy of choice see your email, your project management tool, your Slack, your newsletter platform, your diary, etc. Instead of copying data into a chat window and asking an AI to analyze it, the AI just talks to the tool. Presto. That means Claude or ChatGPT can synthesize data and narratives across all of the places where you show up, but also - increasingly - it's leading to one's AI buddy becoming the homepage of your internet, your productivity, your life, your grip on reality. (Or maybe that's just me.)

Anyway, back to beehiiv. If you're a paid beehiiv newsletter writer (I am and, full disclosure, I serve on an advisory board for the beehiiv Media Collective), you can opt in to a pilot program, link up to Claude and start asking questions like: Who are my most engaged free subscribers who haven't converted to paid, and what do they have in common? And it will actually look at your subscriber data and tell you. Not only do you not have to export Excel spreadsheets or CSVs, you also don't really need to have a deep understanding of data, analytics, or audience strategy.

The MCP is only as good as the prompts you feed it. I sat in on an eye-opening session with beehiiv's Ryan Gilbert earlier this week. Ryan was querying his own newsletter data at a very high level. He got a ton of useful insights and action steps out of it that are beyond my level of understanding. But that's okay, because for me, it's more useful to type in simpler queries like: What are the red flags that you see in my audience data?, or, Based on looking at all of the audience data available, what are five things that I can do right now to increase my engagement and retention? The MCP and the advice that Claude gives me will meet me on my level and even allow me to level up a bit.

Tyler Denk, beehiiv's CEO, wrote in his newsletter that he expects most beehiiv creators will eventually manage their newsletters through AI rather than the app itself. What a wild thing for a tech company CEO to say: you won't use our interface.

The beehiiv MCP is a signal. Things are changing significantly and quickly. And it arrives in a week where the conversation about journalists and AI got a lot more interesting…

On Wednesday, Wired's Maxwell Zeff published a piece profiling tech reporters who are using AI throughout their work. Alex Heath, who left The Verge last year to go independent on Substack, now speaks his scoops into a microphone using a voice-to-text tool, feeds them to Claude, and lets it write his first draft. He's built a custom set of instructions – his "10 commandments" of writing like Alex Heath – so the AI matches his voice. He still revises everything. But he says this workflow saves him 30 to 40 percent of his writing time, and it frees him up to do more of the thing that actually makes his newsletter valuable: reporting.

Casey Newton, who writes Platformer, said there's a difference between newsletters where the value is in the information and newsletters where the value is in the voice. If it's information, he thinks readers won't care much that AI helped write it. If it's voice and opinion and analysis, using AI to generate the whole thing feels cheap. Newton isn't using AI to write Platformer (yet), but he's experimenting with a Claude-based editor modeled on his own past work, and says the feedback rivals what he's gotten from human editors.

Taylor Lorenz, meanwhile, uses AI to run the business side of User Mag (SEO descriptions, data analysis, YouTube optimization) but won't let it anywhere near her actual writing. She doesn't trust it with sensitive reporting materials, and she loves the craft too much to hand it over.

What I find useful about the Wired piece is that it models something we all need to start doing: talking openly about how we use these tools. Because a lot of people are using them and not saying so. 👀

The same day the Wired piece dropped, The Atlantic published Vauhini Vara's investigation into undisclosed AI use at major newspapers. The piece centers on a New York Times "Modern Love" column that an AI researcher flagged as likely more than 60 percent AI-generated. The author told The Atlantic she'd used AI for "inspiration and guidance and correction" but hadn't copied and pasted from a model.

Vara cites a research team that ran thousands of articles from the Times, the Wall Street Journal, and the Washington Post through AI detection tools and found that opinion pieces were six times more likely to contain AI-generated content than newsroom articles. That's a pattern, so we need to start talking about it.

Here's where I think this lands for us over here in creator journalism land: there are two ways to handle AI, and we're watching both play out in real time. One is to use it quietly, hope nobody notices, and risk getting caught in exactly the kind of credibility crisis that's now engulfing legacy media. The other is to be upfront about it and to tell your readers what you use AI for, what you don't, and where the human judgment lives.

I'm going with option two. So here's my disclosure:

I use Claude EXTENSIVELY in my workflow. Every Thursday morning, it assembles the research brief I use to build this Friday post. It sweeps my Gmail, my Slack channels, my meeting notes, and a handful of industry sources, then organizes everything into essay pitches, links, community spotlights, and programming notes. I still write every word you read. I still make every editorial call. Wht I don’t do is spend half my Thursdays on a scavenger hunt – scrolling through Slack, skimming email newsletters, running Google Alerts, looking for the links and stories worth sharing. Now Claude does that searching and puts it all in one place. And it did take me a while to train Claude to do the search correctly. But taking the time to train Claude means I actually have the time to read and digest articles, watch the videos, think about what matters and why. The searching was never the valuable part of the work. The reading and thinking was. I just couldn't get to enough of it before and stressed. A lot.

I also use AI to help me think through drafts of essays and longer reported pieces. It doesn’t write them, but my new AI editor does push back on my arguments, suggest cuts, and flag when I'm being unclear. It's not a replacement for a human editor (I wish I could afford one). But for a one-person operation, it's the difference between publishing something I've pressure-tested and publishing something I just hope holds up. (If anyone's interested, reach out to me and I'll share the prompt with you that I use to design my AI editor.)

For independent creator journalists, this is the moment when you really need to be thinking about how to integrate AI into your workflows, because despite all of the inherent risks that come with ceding some control to robots, the gains for the kind of one-person and small-team operations that make up most of the creator journalism world are considerable. Not only are we doing the content creation - the part we want and need to preserve as human-generated – but we're also managing budgets, social media accounts, marketing plans, subscriber drives, ad sales, and on and on and on. Much of that work can be done by the robots now, and it's a fair trade to reclaim some of that time for the work that actually matters: the journalism.

The beehiiv MCP isn't going to fix everything overnight and Claude doesn’t suddenly mean we don’t need editors or fact checkers or peers to review our work. But between beehiiv connecting your newsletter to AI, Wired profiling the reporters who are already doing this openly, and The Atlantic documenting what happens when people do it secretly – the message of this week is pretty clear: AI is already in the workflow. The only question is whether you're going to be transparent about it or not.

I know which side I'm on.

(Ryan Gilbert, who leads beehiiv's creator partnerships and has been running our Building with beehiiv series, led a session specifically on MCP integration this week for the Project C Community. If you're a member, look for the replay in the Slack replay folder.)

📌 YouTube revamps its Creator Partnerships portal at NewFronts. The old BrandConnect was a "ghost town." The new version uses Gemini to match creators with campaigns — and unlike Agentio (which takes 20%), YouTube isn't charging a platform fee.

📌 The Creators Guild of America launches Mosaic. Every campaign and collaboration gets verified, documented, and owned by you. Covered in The Hollywood Reporter, Rolling Stone, and Forbes this week.

📌 The 2026 journalism layoff wave is already worse than last year (and it's only March). CBS News Radio shut down after nearly a century. The Washington Post, Politico, Vox Media, CNBC, and the Wall Street Journal have all cut staff. CBS alone laid off 6% of its newsforce.

📌 Rick Ellis's Too Much TV newsletter sparked a Verge investigation into whether Kalshi and Polymarket's journalism partnerships raise ethical red flags. The piece centers the Project C member’s argument that prediction market partnerships undermine journalist credibility – and ends by noting people subscribe because they trust him.

📌 A new nonprofit called "Deinfluence" is going after undisclosed influencer campaigns with litigation. Founded by a lawyer who pursued fraud in the travel influencer space. The timing is right as Republicans in Congress are now investigating whether a left-leaning digital group violated campaign finance disclosure rules through influencers.

🎉 Seamus Hughes (Court Watch News) is officially a Goldsmith Prize finalist for investigative reporting — one of the most prestigious awards in the field. Congratulations, Seamus!

🎉 Jillian Melero published the March issue of Connect Puerto Rico, pulling together reporting from Canary Media, IEEFA, 9 Millones, and TIME on the archipelago's energy crisis — plus a new recurring section, Energy Workforce Watch.

🎉 Jason Mundok's local lifestyle podcast So Good Lancaster hit 3,500 downloads across 21 episodes. Steady, consistent growth!

🎉 Zineb Haddaji published her first AND second TikTok videos — and the community has been cheering her on with feedback and encouragement.

🎉 Alicia Hastey, a former CBS News producer, joined the community this week with her project thefourthestate.co — aiming to bring together creators in a single space and help them grow audiences as watchdog journalists. Welcome, Alicia!

🎉 Huma Razvi joined the community and is building a newsletter called The Fifth — recently moved from Ghost to Substack after a year of "crickets." Welcome, Huma!