I participate in several public and private AI communities, and one fun exercise is to take another creator’s image generation prompt, run it yourself (using the same AI tool or a different tool), and see what happens. But certain tools can yield similar results, for explicable reasons.
On Saturday morning in a private community Zayne Harbison shared his Nano Banana prompt (which I cannot share here) and the resulting output. So I ran his prompt in Nano Banana and other tools, including Microsoft Copilot and OpenAI ChatGPT.
The outputs from those two generative AI engines were remarkably similar.
But Harbison’s prompt was relatively simple. What if I provided a much more detailed prompt to both engines?
Create a realistic photograph of a coworking space in San Francisco in which coffee and hash brownies are available to the guests. A wildebeest, who is only partaking in a green bottle of sparkling water, is sitting at a laptop. A book next to the wildebeest is entitled “AI Image Generation Platforms.” There is a Grateful Dead poster on the brick wall behind the wildebeest, next to the hash brownies.
So here’s what I got from the Copilot and ChatGPT platforms.
Copilot.
ChatGPT.
For comparison, here is Google Gemini’s output for the same prompt.
Gemini.
So while there are more differences when using the more detailed prompt (see ChatGPT’s brownie placement), the Copilot and ChatGPT results still show similarities, most notably in the Grateful Dead logo and the color used in the book.
So what have we learned, Johnny? Not much, since Copilot and ChatGPT can perform many tasks other than image generation. There may be more differentiation when they perform SWOT analyses or other operations. As any good researcher would say, more funding is needed for further research.
But I will hazard two lessons learned:
More detailed prompts are better.
If the answer is critically important, submit your prompts to multiple generative AI tools.
I know it’s late on a Friday night (or perhaps Saturday morning where you are), but we need to speak about reality.
As you may know, I’ve grown tired of the word “trust” because of its overuse in the identity verification industry. When everyone repeatedly uses “trust” as a supposed differentiator, no one is differentiated.
But what happens if the overused word “trust” escapes the tired vision statements and starts to be taken seriously?
Why should my prospects trust Bredemarket?
Let’s bring this VERY close to home, why should Bredemarket’s prospects and clients trust me?
After all, there are many reasons why they shouldn’t trust me at all.
I claim to have worked for about two dozen clients (give or take) since 2000, but the majority of readers of this post cannot name one single Bredemarket client. A few of you can name one of my clients. (Especially if you’re the client in question.) Maybe someone can name two or three. This is by design, since I usually function as a de facto ghostwriter, where my work-for-hire words literally become the property of the client.
But at least in the 20th century you knew that a person was behind any claims. The person may have been lying through their teeth, but there was a person behind the lies. Today there may be no such person. What if Bredebot is NOT my only synthetic identity creation? What if I do not exist, and have never existed?
Ah…I can see the uncertainty entering your consciousness.
And now you’re thinking…that maybe you can’t trust anything I said over the last five years.
But now…
Why should your prospects trust you?
…think about how outsiders look at YOUR company.
And if outsiders have any reason to…um…trust you.
And what strategies and content you need to regain the trust of these outsiders.
Now I am not asking you to immediately trust my claim that Bredemarket can equip you with the content you need.
There are two problems with these “AI-powered” product marketing messages, and you probably don’t even realize the first one.
The first problem
Because you and your friends are so used to seeing the letters “AI” that you know to pronounce each letter separately, as in A I.
But most people don’t know this. Really, they don’t. So when they see those two capital letters next to each other, they think they’re supposed to emit a high-pitched scream.
Try it yourself. Read the sentence below, but instead of speaking the letters A and I in a normal tone of voice, yell them as a single interjection.
“State-of-the-art, frontier AI.”
Google Gemini.
Is that how you want your customers to talk about your product?
The second problem is more obvious…I hope.
The second problem
Despite its undeniable impact on all of us, artificial intelligence is just a feature. Like the Pentium, or Corinthian leather.
And it’s a feature that everyone has. Not a differentiator at all.
To say your software is AI-powered is like an automotive company saying their cars have tires.
Google Gemini.
How many times do you see Ford or Toyota saying their cars have tires?
They don’t waste their time talking about something that everyone has.
And you shouldn’t waste your time talking about your AI feature.
“Generative AI promised to relieve humans of the tedious, mechanical work — freeing them to be more strategic, more creative, more human.
“The reality? We’ve wrapped our rationalizations around this new concept called “humans in the loop.”
“This often means marketers are demoted to glorified spellcheckers and fact-checkers for machine output. Not creators. Not strategists. Just custodians of content they never had a hand in shaping.
Perhaps Rose’s thoughts are wishful thinking on the part of carbon-based marketers.
But if the “humans in the loop” thought persists…isn’t everyone using the same undifferentiated loop? When everyone yells “we use AI,” no one is differentiated. And no, it makes no difference with AI flavor of the week you’re using, since they all train on data. Human data.
And if the humans at all the companies are imprisoned by their identical loops…who has the competitive advantage? No one.
Except for those that use humans…especially humans who have been around for a while and remember this. If you don’t have a full five minutes, skip right to the three-minute mark.
Take a look at your most recent content. If you extracted this content from your channels, changed the names, and injected it into the channels of one of your competitors, would anyone know the difference?
This post looks at content created by human SEO experts, and my generative AI colleague Bredebot. And how to differentiate your content from that of your competitors. (Inserting a wildebeest isn’t enough.)
Several years ago
Several years ago (I won’t get more specific) I was a writer for a company’s blog, but I didn’t own the blog. Frankly, I don’t think anyone did. There were multiple writers, and we just wrote stuff.
One writer had the (apparent) goal of creating informational content. The writer would publish multiple articles, sometimes with the same publication date.
The posts were well-researched, well-written, and covered topics of interest to the company’s prospects.
They were clearly written with a focus on SEO—several years ago, AEO didn’t exist—and were optimized for keywords that interested the prospects.
The goal was simple: draw the prospects to the company website with resonating content.
What could be wrong with that?
This week
Now it’s 2025, I’m writing for the Bredemarket blog, and I own the blog and control what is in it.
Bredebot. (In the middle.)
But I’m not the only writer. I brought a new writer on staff—Bredebot. And like a managing editor, I’ve been giving Bredebot assignments to write about.
As of Sunday August 31 (when I’m drafting this post), the next three Bredebot posts to be published are as follows (subject to change):
Move Over, Authentic AI: Why You Shouldn’t Overlook AI’s Role in Modern Marketing
Power Up Your Sales: A CMO’s Guide to Sales Enablement (with a Wink and a Nudge)
What Is Liveness Detection? Let’s Re-Examine a Sentence
Bredebot just finished writing the sales enablement and liveness detection posts Sunday afternoon, and they blew me away.
The posts were well-researched, well-written, and covered topics of interest to Bredemarket’s prospects.
And while I’m not as much of an SEO/AEO expert as my colleague from several years ago, the posts do feature critical keywords. For example, the references to Chief Marketing Officers are intentional.
The goal is simple: draw prospects to the Bredemarket website with resonating content.
What could be wrong with that?
Next week
I’ll tell you what’s wrong with that:
Any other company could publish identical content.
My colleague from several years ago could produce identical content for any firm in that particular industry. Or some other writer could produce identical content.
Moving to the present day, my esteemed competitor Laurel Jew of Tandem Technical Writing could (if she wanted to; she probably wouldn’t) log in to her favorite generative AI engine and churn out bot-written posts on sales enablement and liveness detection that read just like mine—I mean Bredebot’s. Especially if she reverse engineers my prompts and includes things like “Include no more than one reference to wildebeests as marketing consultants and wombats as customers of these marketing consultants.” Once Bredebot has been easily cloned, game over.
TTW Bot?
As I noted Sunday, a correlation in which two bots use the same source data ends up with the same results.
Perhaps I could mitigate the risk by using a private LLM with its own super secret data (see Writer) to generate Bredebot’s content, but as of now that ain’t happening.
Another way to mitigate the risk is by careful prompt tailoring. I experimented with this in the pre-Bredebot days, back when Google Gemini was still Google Bard, and I told it to assert that “Kokomo” is the best Beach Boys song ever.
But in the end, no matter what data you use and what prompt you use, a generative AI bot is not going to produce anything original.
Some time ago I talked about a lack of differentiation that was, um, caused by one company copying another.
And one of those records was so unmemorable that it was memorable.
The album, recorded in the early to mid 1960s, trumpeted the fact that the group that recorded the album was extremely versatile. You see, the record not only included surf songs, but also included car songs!
The only problem? The album was NOT by the Beach Boys.
And I can’t even remember the name of the band.
But this sameness is not only a result of causation.
It can also happen due to correlation, when two things—in this case, two pieces of content—originate from the same source.
But I guess that isn’t “relevant,” so the company unveiled a new logo.
Without.
As someone noted on social media, the new logo removes the barrel and the…well, I shouldn’t go there.
So how did this attempt at relevancy play? According to CBS News, not well.
“Shares of Cracker Barrel fell as much as $8.74, or almost 15%, in Thursday trading, shaving as much as $194.6 million from the company’s market value. The stock regained some ground in early afternoon trading, with shares down $8.19, or 13.9%, to $50.84.”
Bredemarket 