I walked the floor at eTail Boston with a recorder and one question. Should companies use AI to create content, and why? When I got home I kept asking it in LinkedIn messages to people I already knew. The full report publishes soon, with every answer sorted by how the person was reached, and the numbers stay in the report until then. One pattern was there in the first conversation on the floor, before I had anything to count, and it has kept showing up ever since.
The answer people give
Almost nobody answered yes or no, and most of them pushed back on the choice before I finished asking. The answer that kept coming back, from a chief commercial officer, from a CTO, from working copywriters and from people who build AI systems for a living, was yes, with a human owning the output. Kelsey Knight, a commercial executive in consumer and B2B SaaS, put the line where most of them put it.
“Yes for research, fact-checking, and speed but never for origination. AI is derivative by construction, meaning it can only recombine what already exists, so anything that starts in a model starts as someone else’s work with the attribution stripped out. We should use it to sharpen our thinking, but never to have the thought.”
Kelsey’s answer is the shape of the majority answer, and I expected some version of it. People also volunteered a second answer that the question never asked for. Readers know.
Readers know
William Jowett of Bidnamic said it without a hedge, the way most people on the floor answered, in the few seconds they had between booths.
“I can really tell when it’s been procedurally generated by a computer. There’s no taking away from human creativity. People are latching on to the fact that they don’t want to just be fed this slop, and there’s no replacement for what a human brain and human hands can do.”
My first thought, standing there, was that humans are smarter than humans give humans recognition for. We found the patterns, and we can recognize AI way before anyone thought we could.
He was not the only one to reach for that word. Slop came up on the show floor and in the messages without me ever using it, from a growth lead, from a COO, from marketers, and from people who were otherwise enthusiastic about the tools. One person who answered by message said the mistake companies are making is assuming their audience cannot tell the difference between machine writing and a human’s, and that the audience can, because a model has reasoning and no intent. Julian V Taylor carried that thought one step further out.
“The best AI-created content still needs human experience, judgment and personality; otherwise, we’ll simply end up with a world full of perfectly written content that nobody remembers.”
Reba Hatcher, chief commercial officer at ButcherBox, answered in favor of using the tools and then described the work that keeps the result from reading like one.
“Companies should definitely use AI when creating content, but there is a process for figuring out how you fit yourself into it so it doesn’t feel AI generated. Certain things just need to be written by a human. Remember what makes you uniquely you.”
The tells readers say give a page away
People were specific about what gives a page away. A marketer who writes with AI told me that when every first character in a heading is capitalized, the page screams AI to him and he bounces. Another described corporate collateral that reads robotic even when a person wrote it, and said the personality is the thing a reader is scanning for. Readers found the patterns on their own, and they recognize the shape before they finish the first screen, and the AI search engines, which are readers with a bigger memory, pass over those pages when they pick what to cite. Rishabh Chatterjee, co-founder and CTO at Passionfruit, has watched it happen from the data side.
“Depending solely on AI outputs does not drive good results over time. We’ve seen companies push out thousands of AI-generated pages, it helps for two weeks and then everything starts tanking. It should be human in the middle.”
We see that pattern in a client’s Search Console too. The pages an AI engine cites are the ones that say a specific thing plainly, in a voice the engine has not read a thousand times, and I wrote about why the engines pass over generic answers in how to write user-oriented content that wins the click. The rule holds in what brands miss about product data for AI channels, where a machine reads the catalog and skips what it cannot trust.
Where that leaves the people who write for a living
The no answers came from people I did not expect to hear them from, because several of them build AI. Mathew Adams directs advanced R&D at Adeia, a patent licensing company, and his answer was a plain no, followed by the admission that his team uses the tools all the time and that the job now is figuring out how to do it in a way that protects the original artists. Kseniya Autukhovich, who answered by message, described her own process, dumping raw thoughts, letting a model pull them into a draft, then reworking it into her own voice, and drew the line at the byline. “Yes, they should, but as a co-pilot, not the author,” she wrote.
My own position has not moved much through this. A model can only build from what already exists, so it rearranges what has already been written, and that is sideways. People write forward. Research is looking back, and a model is good at looking back, so at Content Cucumber we use it for that and then hand the page to a person, and I wrote down how that works in the AI writing policy worth copying. The longer read on how the tools changed the work is in how AI impacts content marketing.
One post, your price
A lot of companies moved their writing to a prompt, and I understand why, it was fast and it was close to free. The people in this study are the readers those pages were written for, and they are telling me they can see it. Before you commit to anything, we will write one post, picked from your own Search Console data and written in your voice by a person, for whatever price you name. Name your price for one blog post, and you will know, warmingly, that a human made it for you.
The full breakdown, by how each person was reached and never blended, publishes with the report. The study page is where it lands, and the question stays open, so if you have an answer, I am still collecting them.
Frequently Asked Questions
In the answers collected so far, yes. People named the tells themselves, headings that all sound alike, a paragraph that opens by restating the question, and confident sentences that say nothing a competitor's page does not also say, and several said they bounce when they see them.
The answer that came back most often in the study was yes, with a human owning the output. AI for research, fact-checking and speed, and a person for the origination and the voice.
One CTO in the study described companies publishing thousands of AI-generated pages that helped for two weeks and then tanked. The pages AI engines cite say a specific thing plainly, in a voice the engine has not read a thousand times.
For research, which is looking back at what already exists. A person writes the page, and the writing policy is published on the blog.
The study page carries it when it lands, with every answer sorted by how the person was reached and never blended. The question stays open and answers are still being collected.
