Sure, it’s easy to claim you’re the author of your work. But would an AI auditor say the same?

I received a sanity check about my AI usage thanks to Pangram, billed as "an AI detector that actually works" and the one I constantly see cited by journalists. It claims accuracy of 99.5-99.9% on detecting samples created by various AI models.

After seeing it come up often enough, I had to try it for myself. I recommend anyone who’s curious enough to try it too. It’s $20 monthly for a basic plan. Before you pay, you can get a week of access with enough credits to detect AI content from several books or other long-form content. 

Maybe you don't want to know your own scores, but I recommend checking. If your work becomes public enough to arouse suspicions, or it's a high-profile client project where they might have a subscription like this, you should know what Pangram reports. 

The first work I checked was a long-form PDF authored by someone else (we won’t resort to AI-shaming here) that I assumed was written with some AI assistance. I didn't expect Pangram to claim it was 57% AI-generated. They might as well say Claude is the primary author.

Then it was time to check my own book about AI and marketing published last year. When anyone asks me if AI wrote it, I say that the book was written by me, but I used AI for research, brainstorming, outlining, and other tasks. Still, I avow, for better or worse, the writing is mine.

Could I have misremembered it though? Could I have let some AI writing slip through? I liked to think my book was all my writing and voice. But maybe I cut a few corners or needed some help? A few lapses with some lazier copy-and-paste jobs over the course of writing a whole book wouldn’t be the biggest sin.

The results were clear: Pangram said that 99% of the book was written by a human. I'm pretty sure that human is me (with the assistance of Ideapress's human editorial team). 

As for the other 1%, it showed me the passages it flagged as AI-generated. They were quotes by others. Based on Pangram's analysis, I can stand by my claim. AI helped me plan the book. I wrote it. And I'm proud of that.

What I hope is that more people are proud enough of their own work to not let Claude and its peers become primary authors. AI could “write” a more detailed and perhaps technically accurate book than I did. It could have tripled the word count and produced a complete work within hours, and that’s even truer with today’s AI models. I like to think that the book wouldn't be as compelling a read though.

I still have hope that people will care about work with a human provenance. I'm reading the novel Under Story now by Chloe Benjamin, and I care that a human took the time to write it. I can't dismiss that AI one day could write such a beautiful, surprising story. But I'd like to prioritize work written by humans for humans. 

For those of us who do care, Pangram may become the new Better Homes & Gardens seal. Perhaps we'll see similar verification from Amazon and Barnes & Noble, and from media outlets. Perhaps more AI-authored works will be rejected.

Until Marketecture adds its own such rating, you'll have to check for yourself if this was really human-written. Claude might be able to write a better column; it didn't write this one, though. 

Maybe it’s a quaint notion. Maybe this idea of humans exclusively reading human-written books will become some quirky subculture. People will say on dating apps or during corporate icebreakers that their fun fact is they only read human-written books. And then others will ask them tons of questions about what it’s like being the Amish equivalent of a reader. They’ll make fun of us who read only human-written work. They’ll tease, “Do you only read writing on papyrus?” What a sick burn.

As a writer, and a reader, I am hopeful that such a scenario is a plot of a dystopian novel or a play. Maybe I’ll even write it. I won’t even let Claude write the foreword.

— David Berkowitz, Chief Community Officer, Marketecture Media

Who: Agency Executives, Media Operations Leads, CFOs, Marketing Technologists

What: Digiday reports that media agencies deploying AI agents for audience targeting, campaign setup, and buying are now building tracking and auditing tools to catch hallucinations and runaway token costs. The concern is backed by data: An April Gartner survey of 1,300 senior marketers found 56% of companies are implementing AI without clear usage policies, and Gartner estimates 60% of organizations using AI will face cost overruns tied to untracked usage.

Why it matters: The agentic conversation has moved from capability to controls. Agencies that can show clients an audit trail for their agents will win the trust argument; those that can't are one billing surprise from a hard conversation.

(Digiday)

Who: Brand Marketers, Commerce Media Leads, Publishers, Agency Strategists

What: Companies are moving past answer engine optimization toward advertising aimed directly at AI agents. Wayfair CTO Fiona Tan says the retailer is tracking agent behavior as an emerging channel. Time's COO Mark Howard says bot traffic now tops human traffic on a growing number of days, cites Ally Bank data showing AI-referred visitors convert better and arrive further down-funnel, and has Time partnering with Mobian on "Agent Ads," a format designed to be readable by agents.

Why it matters: First brands optimized their sites for agents; now they're buying media aimed at them. "Ads to agents" went from a thought experiment to a named ad format with a publisher attached.

(Marketing Brew)

Who: CMOs, Agency Talent Leads, Marketing Ops, L&D Leaders

What: The Drum asked trained marketers what skills they need most, and despite the industry's AI obsession, very few said more AI training; they went all in on strategy instead. The finding suggests working marketers see AI proficiency as table stakes they're already picking up, while strategic judgment is the gap they feel.

Why it matters: A contrarian data point for anyone building an AI upskilling program: The people doing the work want strategy training, not more prompt courses. Worth reading before next year's L&D budget gets allocated.

(The Drum)

Liking AI Brief? The Refresh is a weekly recap of all the most important advertising news, written by AdTechGod.

Thanks for reading this week’s AI BRIEF by Marketecture. 

Marketecture Media is committed to bringing you the latest news, opinions, and community updates. 

If you are interested in advertising on this newsletter, our podcasts, The Advertising Forum, or in our Slack community, you can contact us here.