Your Fact Receipt, Ready to Install

Happy Wednesday! Something a little different up top this week.

If you've ever thought about starting a newsletter of your own — for a consultancy, a practice, or just to stay in front of people between projects — I'm hosting a free one-hour session next Monday, November 16, at 3 PM ET: Kick Off Your Newsletter. We'll cover how to find your angle in one sentence, what your first issue should look like, which platform fits you, and a weekly routine you can actually keep. There's time for questions at the end.

And if you'd rather read than watch, I wrote it all down, too. How to Build Your Own Newsletter is a 23-page guide drawn from 16 years of writing Now I Know every weekday. Comms Stack readers get $10 off with the code COMMSSTACK.

Let’s get started!

The notes I got back on last week's piece about fact receipts were really kind — thank you. And they pointed me toward something I should have seen coming. The idea landed; the setup is where people get stuck. "Make it a standing rule in your AI" is easy to say and a little harder to do when you're staring at a settings screen you've never opened.

So I wanted to make building your own process as easy as I could. This week I'm giving away the fact receipt as a skill you can install — free.

It's a three-page guide, and there's nothing else to download or unzip or anything like that. (My original version had a big zip file — I figured out a way to make it easier.) Page 1 shows what a finished receipt looks like. Page 3 is the skill itself — a page of plain-English instructions you copy and paste. Page 2 shows exactly where to paste it in Claude, ChatGPT or Copilot, plus a route that works in even the most locked-down work AI. I've tested it in all three, and fixed what broke along the way, but of course, when those platforms change, this — being a PDF — can’t adapt. (But you can always ask your AI of choice for help there!)

Once it's installed, every draft meant for other people — an email, a post, talking points, a release — ends with a table. Every fact, number, name and quote gets its own row, with where it came from and what the AI found when it looked. The rows with no source at all go to the top, so you can't miss them, and the last line tells you exactly which claims still need you. (Tip: If it gets annoying, you can always tell your AI to stop doing this automatically and only provide it when prompted.)

The part I care about most is the one rule the AI can't break: it never marks its own work as checked. It can tell you "read it, matches." It can tell you "the draft says 'nearly half' and the survey says 41%." But only you can turn a line green, and only by telling it you checked. That's the grading-your-own-homework problem from last week, written right into the instructions.

Here's the thing I'd most like you to take from this, though. The skill is plain English. There's no code in it — it's a page of instructions, the same kind I gave my own AI chief of staff. Which means you can open it up and change it. Maybe your team needs a column for legal review, or you want every quote matched against a transcript, or you'd rather skip the receipt on internal notes. Edit the text and the behavior changes with it.

That's the bigger idea behind giving it away. A prompt you copy is something you use. A process you can read, test and adjust is something you built — even if someone else handed you the first draft.

Why this matters for communicators: We own the accuracy of what goes out under our names, and AI hasn't changed that. What it has changed is how easy it is for a wrong number to slip past in a sentence that reads perfectly. A standing check that runs every time — and that leaves the final call with you — is a small setup with a lot of protection behind it.

Your next steps:

He has a newsletter on beehiiv. So do Codie Sanchez, Jay Shetty, David Begnaud, Colin & Samir, and Joanna Stern.

They could publish just about anywhere. They chose beehiiv to build a direct connection with the people who want to hear from them most. And you can do the same, whether you’re world-famous or just at the start of your plan for world domination.

Get started for free.

Up top, I wrote about making your AI show its receipts on your own drafts. This week, Elvis Sun built something that does the same for the AI answers about you.

The idea started with Gab Ferree. In her recap of Muck Rack's Generative Pulse Summit, she passed along a point from Muck Rack's Matt Dzugan: "AI sources are a food chain." The page an AI answer cites often isn't where the fact started. Matt's example was an answer that cited Wikipedia, which cited a niche trade outlet, which cited a wire press release. Gab's advice was to dig past the first citation and find what's feeding it.

Elvis, who builds Medialyst with Carly Martinetti, took that literally. His free skill, /citation-foodchain, runs in Claude or any AI agent. You paste in an AI answer (or a ChatGPT, Google or Perplexity share link), and it checks whether each cited page actually says what the AI claims. Then it hops upstream, page by page, with a word-for-word quote at every step, until it hits the root: a study, a press release, a Reddit thread — or nothing. His first test was the old line that our attention span is eight seconds, shorter than a goldfish's, "according to a Microsoft study." ChatGPT cited Time, Time linked to a Microsoft report, the report's chart credited Statistic Brain, and Statistic Brain named the AP and the National Library of Medicine. When the BBC asked both of them back in 2017, neither had any record of it. Four hops, and nothing at the bottom.

Elvis's takeaway is the one I'd underline: the roots of the chain are things comms teams already own — the press release, the study, the expert quote, the story a trade outlet picked up. A lot of AI-visibility money goes after whatever pages AI happened to cite this month. The food chain points further down, to the sources we've been shaping all along. That's one more reason I keep saying AI visibility is a comms job.

Why it matters: Last week, Sarah Evans showed how AI can get your brand wrong. This shows you where the wrong fact might be coming from, and it's often further upstream than anything a citation report will show you.

Your next step: Grab the skill and run it on one AI answer about your organization or your industry. If the chain ends at your own press release, good. If it ends at nothing, you've found your gap.

I spent 15+ years running digital and AI-adoption strategy inside global brands — the kind of enterprise-scale channel work most "AI consultants" have never actually done from inside a company that size. I bring that same playbook to comms teams directly now: workflow audits, executive AI systems, the intranet-and-channel rebuilds I used to run in-house.

A handful of clients at a time. Reply and tell me what you're stuck on.

Real-time speech recognition is harder than batch transcription. You don't get to wait for the end of the sentence. Wispr Flow built its own real-time model, Canto.

On a challenge set of messy audio (music, traffic, nearby voices, whispering, far-field speech), it had the lowest word error rate among the real-time models tested. It also tied for the lowest error rate on LibriSpeech. On 10 hours of real dictations from 2,300+ speakers, it beat every model tested, including Google, OpenAI, AssemblyAI and Deepgram.

Talk to Cursor, Claude Code or Slack, and get clean, formatted text as you go. Mac, Windows, iPhone and Android.

See the benchmarks

Looking for a role at the intersection of communications and AI? Here are some opportunities to check out:

Was this forwarded to you by a friend or colleague? Want to get this in your inbox next week? Click here to subscribe!

Quick question: how can I help?

What workflows are you struggling with? Where does AI still feel mysterious or overwhelming? What has worked that you’d like to share with others?

I’m a builder, and I’d love to help you and the rest of The Comms Stack community find great new ways to use AI.

Reply and tell me.
I read every response.

Even a one-sentence reply helps. For example:
“I wish AI could help me with ______.”

Until next Wednesday,

Dan

Advertise | Subscribe