5 reasons why relying on AI content is dangerous

• Is relying on AI content really intelligent?
• Reason #1 why relying on AI content is dangerous—inaccurate AI summaries 
• Reason #2 why relying on AI content is dangerous—AI is often just plain wrong
• ChatGPT—Outsmarted by a game
• Using AI can be frustrating—especially when it gets “stuck”
• Reason #3 why relying on AI content is dangerous—AI and AI-assisted tools aren’t infallible
• Reason #4 why relying on AI content is dangerous—AI can not only be wrong, it can contradict itself
• Reason #5 why relying on AI content is dangerous—it could cost you your job
• AI is a tool, it’s not a replacement—it still takes a skilled operator
• AI can do lots of things, but there’s still lots it can’t do

I see posts daily claiming that AI is our savior. Supporters claim it will revolutionize content creation, write better copy, and answer all our prayers. It won’t. It doesn’t. Trust me, it can’t. Don’t get me wrong—AI has its place. I use it almost daily. But as I’ve said before, it’s a tool. And a tool is only as good as its operator. But relying on AI content is dangerous.

Is relying on AI content really intelligent?

Perhaps AI is inappropriately named. My biggest beef with AI is the “intelligence” part. I don’t find it very intelligent. A graphic designer I work with shocked me when he warned me to count fingers and eyes in AI-generated images. What? How would the average person answer these questions?

• How many fingers do people have?
• How many eyes do people have?

If AI doesn’t know the answers are ten and two, what else doesn’t it know? AI-generated images with three eyes or nine fingers don’t seem very “intelligent” to me.

I also have a pet peeve about AI-generated written content. Even when I include “for an Australian audience” in the prompt, the answer uses American English. When I work for different clients, I ask them about their audience. I adjust the spelling, vocabulary, and grammar for that audience. Why can’t AI handle this seemingly simple task?

These are two indications why relying on AI content is dangerous.

Reason #1 why relying on AI content is dangerous—inaccurate AI summaries 

Users often tout AI’s ability to summarize vast amounts of data. Now, I’ve never used the paid version—$20/month ain’t cheap. But the free version of ChatGPT has limits on the amount of text you can paste for summaries. When I use it to summarize content, I often get errors related to copy length. I have to break it up into manageable chunks to accomplish my tasks.

Furthermore, the BBC recently reported that AI chatbots can’t accurately summarize news articles. Their study found that “51% of all AI answers to questions about the news were judged to have significant issues of some form.” What’s even worse, “19% of AI answers which cited BBC content introduced factual errors….” These factual errors included basic mistakes such as dates, numbers, and statements.

You should take those AI-generated summaries at the top of your Google search results with a grain of salt. More on that later. 😉 But if you’re using AI summaries in your content, be careful. Even if you’re going to edit the answer it gives you, this is a classic example of why relying on AI content is dangerous. You may be spreading incorrect information, which could damage your brand’s reputation.

Reason #2 why relying on AI content is dangerous—AI is often just plain wrong

In addition to incorrectly summarizing things, AI is often wrong. I wanted to confirm if Yoon Seok Yeol’s recent declaration of martial law was the shortest ever. Initially, ChatGPT told me Yoon “has not declared martial law during his presidency as of my last knowledge update (December 2024)”.

One of ChatGPT’s shortcomings is that it isn’t great for current events. It needs time to digest and analyze data. I pushed back and informed ChatGPT that he had declared martial law on Dec. 3, 2024 (I was there 😉). Then ChatGPT gave me this answer. 

Why did ChatGPT need this extra prompt? How many people just accept the first answer generated as true or accurate? Ever read the small print at the bottom of the ChatGPT screen?

ChatGPT can make mistakes. Check important info.

That is sage advice.

While writing this blog, I tested ChatGPT again. When asked for the shortest declarations of martial law, it gave me three examples. They ranged in length from 6–13 days. I asked for examples where martial law lasted less than 24 hours. Again, I got several answers ranging from a few hours to a day. Why was this follow-up necessary? The superlative “shortest” should return the one example shorter than all the others. 

If someone asked me, “What’s the tallest mountain in the world?”, I wouldn’t list random tall mountains. My answer would be simple—Mt. Everest. Why does ChatGPT struggle with the concept of a superlative? I’d expect artificial “intelligence” to be able to determine this easily.

BTW, the last time I checked, 157 minutes is substantially shorter than 6 days, a day, or a few hours. The first 10 search results for “shortest declaration of martial law” in Google all relate to Yoon’s recent declaration of martial law. What’s up, ChatGPT?

ChatGPT—Outsmarted by a game

I recently purchased a trivia game called Outsmarted, which my family and I have been playing a lot recently. One general knowledge question asked how many definitions there are for the word “run.” I’ve written about the versatility of “run” before. The correct answer seemed rather high, so I asked ChatGPT after the game.

The in-game answer was significantly higher than “over 30”, so I asked for clarification.

This response is much closer to the in-game answer (I think it was 627, but I can’t remember). I called ChatCPT out on its first answer.

So why is AI admitting it “should have been more consistent in my response?” Doesn’t that seem a “tad” odd to you? 

Should you run a query in ChatGPT, I hope the response it runs by you doesn’t leave you running around in circles, forcing you to run to a more trusted source.

Using AI can be frustrating—especially when it gets “stuck”

Earlier, I mentioned that I often have to break up text to get ChatGPT to summarize things. One day, I used it for a task I regularly use. Because I use it so often, I store the prompt in a text expander, so it’s always the same—only the content changes.

On that day, I did what I always do. 

But instead of getting the expected output, I got this.

I pasted the transcript yet again. I got this response, so I prompted ChatGPT again.

This happened several times. I was getting more than a little frustrated. Can you feel the sarcasm in my response? It took great restraint not to tack “, dumbass” onto my first response. 😉

Despite repeated attempts to clarify what I wanted, we were back to square one. This is where my frustration boiled over, and I “went off” on ChatGPT a bit. Ironically, this worked, and I got what I needed.

AI is supposed to learn. This is a prompt I’ve used several times before and never had an issue. I repeatedly tried to clarify my instructions. But it wasn’t until I basically called ChatGPT out for being a dumbass that I got what I needed. Again, not what I consider terribly intelligent. AI certainly isn’t doing a great job of learning my usage patterns to determine what I need.

Reason #3 why relying on AI content is dangerous—AI and AI-assisted tools aren’t infallible

I often Google grammar rules to double-check things. Recently, I needed to confirm something. I needed to know whether to hyphenate the exercise where you raise yourself off the floor with only your arms. I got this very “enlightening” answer.

Make this make sense.

• No, “push up” is not hyphenated…
• It is written as “push-up”…
• Correct spelling: push-up

Tell me how that answer isn’t confusing. We write some words differently depending on whether they’re a verb or noun/adjective.

• You need to log in to the system. To do so, you’ll need to provide your login information.
• I recently set up my new workstation. My new setup is so sweet!

I wanted to determine whether the noun form was written with a hyphen. I know most nouns are written as one word (login, setup), but I wanted to check. After all, email has morphed into email, and that’s much newer than the exercise. But when the answer begins with “No, ‘push up’ is not hyphenated…” and later states “Correct spelling: push-up”, you can understand why I needed to do some more research. 

BTW, push-up is correct when referring to the exercise.

Reason #4 why relying on AI content is dangerous—AI can not only be wrong, it can contradict itself

I recently had to put together a calendar of national holidays and observed days. My company is based in Australia, so to confirm I had the correct dates, I double-checked on Google. While the AI-generated summaries were easier to read, several were problematic.

For instance, I needed to find out when Red Nose Day is celebrated in Australia. The AI summary stated:

It states no less than three times that Red Nose Day will be held on 9th August, 2025. It also states it is always celebrated on the second Friday in August. There’s just one “teeny, tiny” problem.

When is National Bandana Day? That’s easy. It was held on October 24, 2024, and is always the second-to-last Thursday in October. So when is it in 2025?

Aside from not hyphenating “second-to-last Thursday” in the answer—which Grammarly (affiliate link) also missed—let’s look at the other problem here. “National Bandana Day is next observed on Friday, October 31st, 2025. It is observed on the last Friday in October.” Uhm…what happened to Thursday?

If you can believe it, it gets worse. When is Australian Wildlife Week?

Of course! It’s in October! You know, April 6-10…in October. WTAF?

Reason #5 why relying on AI content is dangerous—it could cost you your job

Are you a freelance writer? You’ve likely heard about clients incorrectly flagging original content as AI generated. But if you’re copying and pasting AI-generated content, it’s easily detectable. I was recently looking at a competitor’s website. It seemed odd that all their blogs were posted on the same day. I skimmed one and it just “screamed” AI at me. I pasted it into Grammarly, and unsurprisingly, it came back as 100% AI-generated.

So how do we differentiate ourselves from AI-generated content? I recently saw a funny (but also sad) post on LinkedIn. A freelance writer had submitted work to their client. Unfortunately, despite their best efforts, the submitted work contained a typo. Oh no!

But wait! The client was actually glad to see the typo. Why? The typo meant the freelance writer had done the work and hadn’t just copied and pasted an AI response. 

Is that what it’s come down to? Sure, most AI output is grammatically correct. But as I’ve pointed out, depending on your audience, it may or may not use the correct vocabulary or spelling. But will AI start strategically adding typos to make it seem more “human”? If it does, my money is on incredibly awkward, unnatural typos.

AI is a tool, it’s not a replacement—it still takes a skilled operator

AI isn’t taking your job, especially if you’re good at it. I’ve said it before, and I’ll say it again—AI is a tool. It’s currently unable to replicate human content creation. AI stock images (even with the right number of fingers and eyes) look like AI-generated images. AI content reads like AI content. Anyone with a slightly critical eye will know your content is AI generated. 

Know who else isn’t worried about AI? Hank Azaria. Yeah, that Hank Azaria, of The Simpsons fame. Why isn’t he worried—well, at least not really worried? Well, for starters, he’s incredible at his job. Also, he brings something to his voices that AI can’t—humanness. Finally, he’s much better at improv than AI could ever be, which allows him to create in the moment.

Will AI be able to recreate the voices he does? Sure. Will it be the same, or perhaps even better? Not any time soon.

AI can do lots of things, but there’s still lots it can’t do

Can AI generate grammatically accurate, error-free content? For the most part, yes. Especially if you’re writing for an American audience. Can it replicate your wit, nuance, and creativity? Not likely. If you’re looking for filler content to try and boost your site’s SEO—roll the dice. It’s still unclear if the Google algorithm penalizes AI-generated content. But I’ll err on the side of caution and avoid it—with one exception. 😉

Looking for engaging content that resonates with your audience? Then you’re better off hiring a real copywriter or content writer. If you can verify that they’ve submitted pure AI-generated content, don’t use them again. There are still lots of human writers who actually write what they submit out there looking for work. Many of them are very good. And affordable. Like me—even if we use AI to help us be more productive. 😉

Subscribe to this blog or follow me on Instagram, Facebook, LinkedIn, or Threads for more content and updates.