I Read the AI Summary. I Felt Guilty. I Did It Anyway.

Happy Wednesday! This week:

Let's get started!

I like to read. I'm the person who follows the footnote, clicks the third link, reads the whole thread to the end. Curiosity is more or less my defining trait, and I'm a little proud of it.

So it bothered me when I caught what I'd started doing. A long, clearly labored-over email lands in my inbox with a tidy AI-generated summary pinned to the top — and I read the summary. Just the summary. A well-reported article opens with a "here's the gist" box, and some days the gist is all I take. I feel a little guilty every time, like I've cut the line on something someone worked hard to build. Most of the time, that guilt — and my curiosity plus my desire (in the case of the email) to fully understand what the other person is saying — takes over, and yes, I do read the whole thing.

But not all of the time. Here and there, I admit it — I rely on the AI-generated summary, and that’s good enough. And again, I’m a reader. If I'm doing it, everyone is doing it.

And as communicators — even if it hurts us to the core — we need to account for that.

Yes, something is lost when a message gets flattened to its bullet points — a story told well does something a summary can't, and we all know it. But the guilt I felt is useless as a plan, because the choice as communicators stopped being ours to make. Whether or not you write a summary, your reader now has one. We have to design for it anyway. Which means you're now writing for two audiences: the reader, and the summarizer.

Pre-AI, we could just decide to not toss a TL;DR into our comms. But now, every long thing you send is already being compressed — by the "summarize this" button your company bolted onto the intranet, by the automated bullets at the top of the all-hands note, or, if none of that exists yet, by the employee who pastes your message into ChatGPT and asks what it says. The summary is happening. The only real question is whether it's yours or a stranger's.

And a stranger's summary is a real risk. An LLM handed your 600-word change announcement will decide, on its own, what the headline is and what gets cut. Maybe it leads with the reassuring paragraph and quietly drops the one action you actually needed people to take. You wrote every word and vetted none of the version most people will read.

I want to be careful here, because the easy move is to call this progress and tell everyone to get over their nostalgia. I won't. When nobody reads the full thing, something real does go — the context, the nuance, the judgment you built into paragraph four. That isn't sentimental; it's the part of the work that made it good.

But naming the loss and refusing the reality are different things. You can believe a well-told story matters and accept that some of your audience just needs the four-line version — today, on their phone, between two meetings. Both are true at once. Pretending that second group doesn't exist doesn't protect the craft; it just hands their summary to a robot.

I know I’m not doing it here — maybe I should — but I think we need to start putting that TL;DR at the top of our own messages, in our own words, and make it carry the one thing an automated summary almost always fumbles: what this means for the person reading it. Is action required, yes or no? By when? What actually changes for you? A machine compresses for what the text says. You can write for what the reader needs to do — and that's judgment, which is exactly the thing we're supposed to be adding. (If you’re familiar with Axios’ Smart Brevity, that may be a good way to go.)

Do that, and the full piece still earns its keep for the people who want it, while the summary stops being a leak and becomes a choice you made on purpose.

If this feels familiar, it should — it's the comms twin of GEO. We've spent a year learning that AI decides how your brand shows up in search whether you participate or not, so you'd better shape it. This is the same lesson pointed inward: AI now decides how your message shows up in your own employees' inboxes whether you write the summary or not. Same two paths — the AI you use, and the AI acting on your words while you're not looking. Comms has to walk both.

Directly above, I say “write the summary yourself,” but I think there’s an intermediary step to consider. Run your comm through multiple LLMs — free versions are fine, but I’d do Claude, ChatGPT, Perplexity, Copilot and Gemini — and ask them to give you that bulleted summary. Check their results against your intended takeaways, and if something is missing, add it to the longer communication. (You don’t need to remove anything accidentally emphasized, because your hand-written TL;DR is going to not emphasize that point anyway.) Adjust the language that the LLMs give you to make the bullets appropriate for your audience and you should be good to go. You’re effectively optimizing the AI summary to meet your goals, getting ahead of what the reader would have done anyway — and ensuring that your message, not some third-party AI one, is winning the day.

Your next steps: Give this a try on an upcoming communication. If you can measure the click-through into the full text — doable if you’re publishing a blog post or an internal comm — even better. If most people stop at your summary, that's not a failure — it's the most useful thing you'll learn all quarter about how your audience actually reads you. Build the next message around what it tells you.

AI made PMs faster. Multiplayer mode is still broken.

A PM can summarize research, draft a PRD, and mock up a prototype before lunch. The hard part starts when the team has to decide what actually gets built.

Jira Product Discovery gives product teams one place to capture insights, prioritize ideas with consistent frameworks, and build living roadmaps stakeholders can rally around.

And because it’s connected to Jira, the context behind every decision stays with the work—so developers and their agents know not just what to build, but why.

AI helps PMs move faster. Jira Product Discovery helps the whole team build with confidence.

Get Started for Free.

A couple of weeks ago I wrote about Reddit falling off a cliff in ChatGPT's citations — from 3.83% of citations down to 0.52% in about a week. Striking number, but it left the more interesting question hanging: if ChatGPT stopped leaning on Reddit, what did it start leaning on instead?

Sean Stevens went and looked. Sean is the founder of brand·lore, an AI perception advisory for PR and comms teams, and he ran a small, admirably transparent experiment — 50 logged-out ChatGPT sessions (ten skincare questions, five fresh runs each, all in one day, with the raw data published so anyone can check his work). He shared it with me before it went public, and it went live yesterday. You can read the whole thing here.

The finding that matters: ChatGPT doesn't trust one set of sources — it trusts a completely different set depending on what you ask. Here's how Sean puts it:

In his runs that played out concretely: acne questions cited the American Academy of Dermatology in all five runs, retinol went to PubMed and clinical trials, "why does my moisturizer pill" pulled Allure and Glamour, and pregnancy-safety questions returned zero product recommendations — where Google Shopping happily served up forty. Reddit, the source everyone was optimizing for a month ago? Zero.

What jumped out at me was that the storefront barely moved while all of this happened underneath it — an 81% product overlap in Google Shopping's top 40, even as the sources ChatGPT actually cites got quietly rebuilt. One sunscreen brand, EltaMD, got recommended in every single run despite sitting at #15 on the Google Shopping ranking. Merchant position didn't earn that:

That's the whole ballgame for us. You can't buy your way onto the shelf that matters. What wins the recommendation is whether an authority ChatGPT trusts for that exact question is talking about you — and that you don't purchase, you earn. For acne it's a dermatology association; for retinol it's the journals; for "does this feel nice" it's Allure. It's earned authority and third-party credibility — the work comms has done for a century, just pointed at a new reader who happens to be a language model. (Or, in other words, we’re already doing GEO — we just didn’t do it intentionally. That’s good news for comms!)

And notice who's quietly setting the rules. The model decided a professional body owns medical questions, that peer-reviewed research owns efficacy, and that nobody gets a product recommendation for pregnancy at all. Those are editorial and gatekeeping calls, made at scale, that nobody voted on — and they're now shaping what your customers hear about you.

Your next step: Write down the five questions a real customer would type before buying from you. Run each in ChatGPT a few times and record the sources it cites, not just the answer. That list — five questions, and the outlet that wins each one — is an easy, starter GEO plan for the quarter.

Everything I write about here I also do hands-on: helping comms teams bring AI into the digital channels they run every day — the intranet, email and newsletters, social, the web — and building the workflows that make it stick. It's the intersection of digital communications and real AI adoption, and it's the work I love most.

A handful of clients at a time; three slots open this fall. Book a 30-minute intro call →, or just reply to this email.

The only platform that designs, licenses, and captures a Branded AI Voice from real, consenting actors—not from scraped data. Trusted by BMW, Superbloom, and Cresta.

Get a Demo

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