Happy Wednesday, and welcome to The Comms Stack! This week:

Let’s get started!

You know advice is great when it seems obvious in retrospect. I was blown away by this LinkedIn post by Noah Greenberg, the CEO of Stacker. It’s a great and dead-simple way to demonstrate the value your work has gotten you in an AI-first world.

Read his post (and give him a follow/like, he definitely earned it!), but here’s the gist:

1️⃣ Before your story comes out, write a handful of ChatGPT prompts that you hope will, in the future, cite the stories you’re placing. Get some baseline data: run those prompts and collect the results somewhere.

⭐ Tip: Make sure you lead with prompts that your target audience would likely use — focus on the people who drive your business outcomes!

2️⃣ After the story hits — maybe a few days after — run the same prompts. Pay particular attention to any results that specifically cite your stories.

3️⃣ Compare the before and after results, highlighting the changes that were driven by your team’s efforts.

Congratulations — you’ve just proved the impact of your work. As Noah says, “put [the] results in a nice deck and present [it] to the exec.” They’ll see the value right away, because as I said at the top, it’s obvious. 🎉

“When I first opened Attio, I instantly got the feeling this was the next generation of CRM.”
— Margaret Shen, Head of GTM at Modal

Attio is the AI-native CRM for modern teams. With automatic enrichment, call intelligence, AI agents, flexible workflows and more, Attio works for any business and only takes minutes to set up.

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉 Start for free today

This is mind-blowing (and really, really, ridiculous).

Let’s start with a quick quiz. Which prompt is better?

1) Here’s a list of 50 names. Find the 25th.

2) Here’s a list of 50 names. Find the 25th. Here’s a list of 50 names. Find the 25th.

That second choice isn’t a typo. It’s actually the right answer. A new paper from Google Research found that AI chatbots, when not in “reasoning” mode (basically, where you’re just asking them to look up something), do better when you repeat yourself. Per the paper, regular prompts returned the right result about 21% of the time, but the repeated version had a 97% accuracy rate. 🤯

Wait, really?: Yes, really! VentureBeat explains: “Most modern LLMs are trained as ‘causal’ language models. This means they process text strictly from left to right. When the model is processing the 5th token in your sentence, it can ‘attend’ (pay attention) to tokens 1 through 4, but it has zero knowledge of token 6, because it hasn't happened yet.” By repeating the query, you’re giving the chatbots the context that you hadn’t yet provided the first time around.

Why it matters: This could be very useful for any task that involves a lot of data processing but not conjecture. For example, if you’re using AI to process analytics, repeating yourself could avoid a lot of mistakes. And all you need to do is copy/paste your prompt so it appears twice!

And also, it serves as a reminder that AI chatbots aren’t human — sometimes, we need to talk to them in their own robotified way to get the best out of them.

The prompt below is for comms pros who haven’t built relationships with reporters yet. It turns any topic into a reporter-targeting engine, mapping journalists to angles they actually care about, backed by real coverage data.

It asks smart, clarifying questions upfront, searches credible sources across the last 90 days, and then spits out differentiated pitch ideas tied to what each reporter has written—plus what they haven’t (yet!). Steal it, tweak it, and go win some placements. 🚀

⚠ Caution: See the words “tweak it” above? That’s a must-do, not a recommendation. Take it from Gab Ferree of Off the Record, who earlier this month told Muck Rack that “your reporter is getting many, many, many perfectly customized pitches all the time that start to look like AI, whether or not they are actually AI.” Perfect can be your enemy — personalize the result instead, to make it uniquely yours (and human!).

⭐ Tip: I built this for Perplexity, but it works nicely in ChatGPT, Copilot, and Claude, too. They sometimes give different results, so try it across the board!

Download the prompt ⬇

This week’s lead was about generative engine optimization (GEO). I’m deep in it right now — learning everything I can about how GEO actually works in practice. I’m working on something really cool — a guide on GEO strategy and tactical implementation. More on that later.

As part of that work, I’ve been trying to get AI to turn a long, text-heavy GEO guide into something polished and presentation-ready. Short answer: It can’t yet.

Design tools with AI (like Piktochart and Visme) came closest, but still required heavy manual cleanup, paywalls, or generated new copy I didn’t want. Traditional AI chatbots struggled even more — Claude produced editable HTML, but changing layouts was painful. Gemini, despite being a Google product, couldn’t create a Google Slides template because it “cannot directly create a file in your Google Drive,” which was … unexpected. (And not in a good way.)

The takeaway: AI is excellent at writing, mediocre at structure, and still pretty bad at creating clean, editable design.

My next steps: I’ll probably just use an off-the-shelf template and manually put the content in. 🤷‍♂️ But I’m hopeful these tools mature — because it would unlock a lot of productivity.

Have something else I should try? If you’ve found a better workflow — or a tool that actually nails this — hit reply. I’d love to steal it.

AI has changed how consumers shop, but people still drive decisions. Levanta’s research shows affiliate and creator content continues to influence conversions, plus it now shapes the product recommendations AI delivers. Affiliate marketing isn’t being replaced by AI, it’s being amplified.

Download the full report

I’ve been playing with Sora, the ChatGPT/OpenAI video maker on and off for a while now. I don’t have a great professional use case for it, although I’m sure there are lots out there. But I have come up with a lot of less-than-professional ones. My favorite video so far came from this prompt:

“A Great British Bake Off contestant turns out to be made of cake when he accidentally cuts his hand.”

You can watch my baker discover his true origins here. 👨‍🍳🍰

Also:

Hit reply to let me know

Until next Wednesday,

Dan