June 25, 2025 • 14 mins
A recent MIT study shows ChatGPT dulls our critical thinking skills over time. From maintaining high-quality campaigns to maintaining a creative mindset, Grace and Peter dive into the harms of AI overuse in business. Plus, a look at how NOT to use AI–and what to do instead. You’ll hear tips on how to avoid inaccurate or misleading information from LLMs, write better prompts, and achieve more accurate data analysis from tools like ChatGPT’s Deep Research.
Chapters
01:01 – The MIT Study: AI and Critical Thinking
03:14 – Why Generic AI Content Fails
05:22 – Smart AI Use Starts Narrow
08:59 – Prompting Mistakes to Avoid
13:09 – Call for Guest Experts
Sources
“ChatGPT May Be Eroding Critical Thinking Skills, According to a New MIT Study” by Andrew R. Chow for TimeMark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:16):
GraceWelcome to Accelerated Velocity, the podcast that helps you move faster, smarter, and more strategically in the rapidly evolving world of AI. We'll break down the latest AI news and innovations and explore how they impact your marketing, sales, and business growth. We'll dive in the practical use cases, unpack new tools, and cut through the noise so you and your team can adopt with confidence.
(00:40):
GraceI'm Grace Mathews, director of content at InboundAV, a business development agency and HubSpot solutions partner dedicated to driving sustainable business growth. Each week, I'll be here with Peter Malick, our CEO and founder. Join us as we make sense of what's changing and what to do about it.
(01:01):
GraceHey everyone, it's Grace. We had some technical difficulties at the start of our episode today, so we're just going to jump right into the meat of the conversation. To set the stage. Today's episode is all about what not to do with AI. We started by looking at a recent MIT study on ChatGPT and its effects on brain activity.
(01:21):
GraceThe study consisted of three different test groups, each of which were asked to write essays based on s.a.t prompts. The first group used ChatGPT to write their essays. The second group was allowed to use Google to get information and ideas, and the third group had no help or resources. They called this the quote brain only group, which I thought was hilarious.
(01:42):
GraceThe research found that those using ChatGPT unsurprisingly had the worst papers. They weren't interesting or original, and they all read very similarly to one another. On the other side, the brain only group had the most interesting and unique essays. Now, when it came to brain activity, the ChatGPT group showed the lowest levels of engagement, especially in brain regions associated with critical thinking.
(02:07):
GraceThe brain only groups EEG showed the highest levels of critical thinking activity. They had higher executive control, higher retention, and higher engagement. And though not quite as high, the group using Google similarly outperformed the ChatGPT group. So why does this matter on our podcast? How does this relate to business development? Let's pick up my original conversation with Peter from here.
(02:30):
PeterSo first of all, I'm going to preface this by saying, you know, we both belong to the same company. It's called InboundAV. And one of the prime things we do is help other businesses, solopreneurs, large companies with utilizing all elements, usually utilizing a genetic AI to streamline their processes, make things easier, and make their work go faster and at the same time.
(02:54):
PeterI love this subject because something we've been talking about for quite a while now is that we have all these amazing tools, and all of those amazing tools also have a back side, and the back side might be changing society in a positive way, and it might be changing society or even human beings in a negative way.
(03:14):
GraceAnd the actual quality of the writing itself being subpar from the ChatGPT group says a lot to anybody who's writing for content marketing, running email campaigns, running cold outreach campaigns, even, I think the, the, the very clear evidence that writing that is just 100% generated by ChatGPT is not resonating with people, no matter what industry you're in, is a very important thing to remember.
(03:43):
GraceAnd beyond that, there's it's not just the quality, but it's the necessity of originality and creativity.
PeterSo I have an overarching question. And that is, as I'm hearing you talk about the study, I'm thinking about the all of them themselves, that we've had these little pockets where this news comes out, LLMS or training on AI generated results. They're getting stupider. Most recently, it's been the alums are hallucinating big time. And so I wonder if the 11 brains are not entirely different from our own human brains in as much as, you know, it's it's kind of like real input in it's going to be a lot more valuable than generated input in and and in fact, if everybody's relying on homes for their entire day, is is is humankind going to get stupider?
(04:39):
PeterJust as the lambs have gotten stupider by reading their own stuff? And how how could they actually change this change this way?
GraceYeah, I think that's a really great way of framing it. I think if you're going to turn to ChatGPT or Perplexity or any other platform, I'm throwing those out because those are the ones I tend to tend to go towards. It's great to maybe use those as research hubs, as tools to identify patterns and trends in your own data or in your industry, but they're not great to develop a strategy from start to finish that you're not engaging with in some more thought provoking way.
(05:22):
PeterI really 100%, and I think I also want to tie it back. I think we mentioned this last week to what we are doing here in InboundAV, and we're building this HubSpot app that does very specific functions within the CRM, within HubSpot CRM, of enhancing data for potential prospects and researching their company, and maybe even have HubSpot respond in a different way based on the size of the company and automating all of that stuff.
(05:50):
PeterAnd so all stuff that's, you know, designed to save time, save research, etc. and the picture that it becomes clearer and clearer to us as we get into it is that we want these agents to do a very specific, very narrow thing whose results we can rely on. And so put them all together, and we have a package of results or results that we can rely on.
(06:14):
PeterAnd to me, that's kind of the conscientious and safe for business use of AI right now, where we stand in the middle of 2025, you know, the bottom line, it's just such an interesting time and there's so many things in play. And, you know, this has been something that's been discussed even before ChatGPT as AI capabilities were developing.
(06:34):
PeterIt's like, you know, is this a good thing? You know, is this something that can help humanity? Is it also something can destroy humanity? And, and most of the, the, you know, the AI specialists and professors and coders and etc. recognize that this is something that could really be a bad thing. And so I think it's really important to look at this, to look at studies like this one.
(06:56):
PeterAnd it just so we know what we're getting into and we can make informed decisions that not only for our business but for our own mental health.
GraceSo I guess we kind of sound like complete chat to be haters today. I think this is a good time to to pivot and, and look at, a few easy tips on how not to use and Lem.
(07:19):
PeterAll right. So yeah, I mean, one of my favorites is a concept of what kind of data you're getting to return. Is that data that has a historical value, or is this data that could be changing every week? And the perfect example is in email marketing, our email subject lines, an effective email subject line might have a shelf life of a week, and then everybody's on to it.
(07:44):
PeterThey're done with it, and it's worthless. So I think we touched on this last week that that ChatGPT was in our search, returning email subject lines that are like so far beyond, a valuable subject line and in fact will keep your email from being opened that, that you want to really be cognizant of. What's how old is the data that you are searching for?
(08:11):
PeterAnd if it's something that is changing as dynamically as email subject lines, as advertising copy, etc., it's not that LLMS can't be valuable, but you just have to be very, very careful that the data you're getting back is true today and not true necessarily two months ago, but it is up to date and true today.
(08:32):
GraceAll thing you can do to improve your results on that is to just give time frames to whatever you're using. Because if you say like, tell me the most recent subject line trends, you might still be getting articles that are recent to the LLM, but not really recent based on how fast everything moves. That leads us to another big mistake that a lot of people make, which is vague or underdeveloped.
(08:59):
GracePrompts. And there are millions of lists and resources out there on like prompt templates and what to do and what not to do when it comes to prompting. But I think we can just give a couple of of quick examples today that might be helpful.
PeterYeah. You know, I think that's the lesson for me and for many other people that I know is that it's very easy to go in as you're trying out an L on an M and ask it something like, what's a great marketing B2B strategy? And that is like, couldn't be more. And so consequently, I think that something that happens is people who are new to the LLMS, newer to using ChatGPT, Gemini or whatever it might be, we'll put in a prop like that, get that garbage and think like, well, this isn't worth my time.
(09:48):
PeterWhere in fact, if they had put in a prompt that was more specific, that would detail the type of output that you want, etc., that they would have had something they were pleased with. So I think that that's kind of an overall recommendation of what to do is to make sure that you learn a little bit about prompt and what works, how to structure it.
(10:10):
PeterAnd, you know, especially if you're on the fence about, you know, is this valuable? Is this just, you know, hype and it's not your site and, you know, as we're discussing this, you know, there's plenty of trapdoors, but at the same time, it can be unbelievably valuable to you, to your business, to to everything.
GraceAnd I'll throw out a little mini tip that I've had success with when it comes to prompting. If you're not sure what the best prompt approach might be for a specific task that you're working on or something you need help with from ChatGPT, let's say ask the LLM what what the best format for your prompt might be so, and what your intention is.
(10:51):
GraceSo you might say, my goal is to improve my marketing strategy at the decision phase with my audience. This is the information that I have. What is the best way for me to frame a prompt so that you can help me out with that? That actually can be very successful, because then you can get information back from whatever land that you're working on, kind of walking you through how it approaches those sorts of questions and how it can get you the best results possible.
(11:23):
PeterYou know, last week we touched on, the the data that LLMs deliver, and we talked about structured versus unstructured data.
GraceYeah, I would love to build on that a little bit because I think, Peter, what you and I found when we were playing around with ChatGPT deep Research feature last week was that especially for that feature, Prompt Engineering was really influential on the results that you got back. And one important factor in that when you're using deep research, or even just a regular prompt on any limb, is to make sure there's a clear distinction between structured data and unstructured data.
(12:03):
GraceFor the most part, a lot of these limbs that everybody has access to right now do not handle unstructured data analysis well. And that would be more having an album analyze more vague data sets, like let's say you upload your spreadsheet of all your email marketing information and you say, let me know what the most successful email marketing strategy was from this data set.
(12:27):
GraceThat's going to be way too vague. What you what you can do instead is you can direct the LLM to a specific data field. I could say factor in email open rates. And tell me what the most successful sending strategy was in terms of time and day on impacting our email open rate. So that's something that's a lot more targeted and specific.
GraceAnd as long as the LLM has all of the data fields available to it to analyze that, you're going to get much more insightful results back. And if you're if you have a CRM or a marketing automation platform or something like that that's integrated with an LLM of your choosing, the same goes for that as well as to think about how you're guiding the prompts towards structured data.
(13:07):
PeterIdeally. Awesome.
GraceYeah. So I know we've thrown out a lot of tips in terms of prompting and navigating AI and ultimately making sure that you don't stop thinking for yourself. But I know, Peter, you wanted to mention a couple other things before we close out.
PeterYeah, I just wanted to give a little call to action. We're going to make a pivot. Not every episode, but why don't you? Episodes a month, we're going to start to have experts on the show to talk about different aspects of AI. And so this is a call to action. If you know somebody who would be fantastic, who will blow our minds, who will not deliver unstructured data, then, you know.
(13:51):
GraceAwesome. All right. Well, sounds like a plan. And I personally I'm excited to see how this podcast keeps developing. Thanks for your time today, Peter. I will talk to you next week.
PeterThanks, Grace. Bye, everybody.
GraceThanks for listening to Accelerated Velocity, a link to our website, newsletter and all that other good stuff can be found down in the show notes. If you like what you heard, we'd appreciate a five star review and make sure to subscribe for weekly episodes on the latest in AI for business.
GraceWelcome to Accelerated Velocity, the podcast that helps you move faster, smarter, and more strategically in the rapidly evolving world of AI. We'll break down the latest AI news and innovations and explore how they impact your marketing, sales, and business growth. We'll dive in the practical use cases, unpack new tools, and cut through the noise so you and your team can adopt with confidence.
(00:40):
GraceI'm Grace Mathews, director of content at InboundAV, a business development agency and HubSpot solutions partner dedicated to driving sustainable business growth. Each week, I'll be here with Peter Malick, our CEO and founder. Join us as we make sense of what's changing and what to do about it.
(01:01):
GraceHey everyone, it's Grace. We had some technical difficulties at the start of our episode today, so we're just going to jump right into the meat of the conversation. To set the stage. Today's episode is all about what not to do with AI. We started by looking at a recent MIT study on ChatGPT and its effects on brain activity.
(01:21):
GraceThe study consisted of three different test groups, each of which were asked to write essays based on s.a.t prompts. The first group used ChatGPT to write their essays. The second group was allowed to use Google to get information and ideas, and the third group had no help or resources. They called this the quote brain only group, which I thought was hilarious.
(01:42):
GraceThe research found that those using ChatGPT unsurprisingly had the worst papers. They weren't interesting or original, and they all read very similarly to one another. On the other side, the brain only group had the most interesting and unique essays. Now, when it came to brain activity, the ChatGPT group showed the lowest levels of engagement, especially in brain regions associated with critical thinking.
(02:07):
GraceThe brain only groups EEG showed the highest levels of critical thinking activity. They had higher executive control, higher retention, and higher engagement. And though not quite as high, the group using Google similarly outperformed the ChatGPT group. So why does this matter on our podcast? How does this relate to business development? Let's pick up my original conversation with Peter from here.
(02:30):
PeterSo first of all, I'm going to preface this by saying, you know, we both belong to the same company. It's called InboundAV. And one of the prime things we do is help other businesses, solopreneurs, large companies with utilizing all elements, usually utilizing a genetic AI to streamline their processes, make things easier, and make their work go faster and at the same time.
(02:54):
PeterI love this subject because something we've been talking about for quite a while now is that we have all these amazing tools, and all of those amazing tools also have a back side, and the back side might be changing society in a positive way, and it might be changing society or even human beings in a negative way.
(03:14):
GraceAnd the actual quality of the writing itself being subpar from the ChatGPT group says a lot to anybody who's writing for content marketing, running email campaigns, running cold outreach campaigns, even, I think the, the, the very clear evidence that writing that is just 100% generated by ChatGPT is not resonating with people, no matter what industry you're in, is a very important thing to remember.
(03:43):
GraceAnd beyond that, there's it's not just the quality, but it's the necessity of originality and creativity.
PeterSo I have an overarching question. And that is, as I'm hearing you talk about the study, I'm thinking about the all of them themselves, that we've had these little pockets where this news comes out, LLMS or training on AI generated results. They're getting stupider. Most recently, it's been the alums are hallucinating big time. And so I wonder if the 11 brains are not entirely different from our own human brains in as much as, you know, it's it's kind of like real input in it's going to be a lot more valuable than generated input in and and in fact, if everybody's relying on homes for their entire day, is is is humankind going to get stupider?
(04:39):
PeterJust as the lambs have gotten stupider by reading their own stuff? And how how could they actually change this change this way?
GraceYeah, I think that's a really great way of framing it. I think if you're going to turn to ChatGPT or Perplexity or any other platform, I'm throwing those out because those are the ones I tend to tend to go towards. It's great to maybe use those as research hubs, as tools to identify patterns and trends in your own data or in your industry, but they're not great to develop a strategy from start to finish that you're not engaging with in some more thought provoking way.
(05:22):
PeterI really 100%, and I think I also want to tie it back. I think we mentioned this last week to what we are doing here in InboundAV, and we're building this HubSpot app that does very specific functions within the CRM, within HubSpot CRM, of enhancing data for potential prospects and researching their company, and maybe even have HubSpot respond in a different way based on the size of the company and automating all of that stuff.
(05:50):
PeterAnd so all stuff that's, you know, designed to save time, save research, etc. and the picture that it becomes clearer and clearer to us as we get into it is that we want these agents to do a very specific, very narrow thing whose results we can rely on. And so put them all together, and we have a package of results or results that we can rely on.
(06:14):
PeterAnd to me, that's kind of the conscientious and safe for business use of AI right now, where we stand in the middle of 2025, you know, the bottom line, it's just such an interesting time and there's so many things in play. And, you know, this has been something that's been discussed even before ChatGPT as AI capabilities were developing.
(06:34):
PeterIt's like, you know, is this a good thing? You know, is this something that can help humanity? Is it also something can destroy humanity? And, and most of the, the, you know, the AI specialists and professors and coders and etc. recognize that this is something that could really be a bad thing. And so I think it's really important to look at this, to look at studies like this one.
(06:56):
PeterAnd it just so we know what we're getting into and we can make informed decisions that not only for our business but for our own mental health.
GraceSo I guess we kind of sound like complete chat to be haters today. I think this is a good time to to pivot and, and look at, a few easy tips on how not to use and Lem.
(07:19):
PeterAll right. So yeah, I mean, one of my favorites is a concept of what kind of data you're getting to return. Is that data that has a historical value, or is this data that could be changing every week? And the perfect example is in email marketing, our email subject lines, an effective email subject line might have a shelf life of a week, and then everybody's on to it.
(07:44):
PeterThey're done with it, and it's worthless. So I think we touched on this last week that that ChatGPT was in our search, returning email subject lines that are like so far beyond, a valuable subject line and in fact will keep your email from being opened that, that you want to really be cognizant of. What's how old is the data that you are searching for?
(08:11):
PeterAnd if it's something that is changing as dynamically as email subject lines, as advertising copy, etc., it's not that LLMS can't be valuable, but you just have to be very, very careful that the data you're getting back is true today and not true necessarily two months ago, but it is up to date and true today.
(08:32):
GraceAll thing you can do to improve your results on that is to just give time frames to whatever you're using. Because if you say like, tell me the most recent subject line trends, you might still be getting articles that are recent to the LLM, but not really recent based on how fast everything moves. That leads us to another big mistake that a lot of people make, which is vague or underdeveloped.
(08:59):
GracePrompts. And there are millions of lists and resources out there on like prompt templates and what to do and what not to do when it comes to prompting. But I think we can just give a couple of of quick examples today that might be helpful.
PeterYeah. You know, I think that's the lesson for me and for many other people that I know is that it's very easy to go in as you're trying out an L on an M and ask it something like, what's a great marketing B2B strategy? And that is like, couldn't be more. And so consequently, I think that something that happens is people who are new to the LLMS, newer to using ChatGPT, Gemini or whatever it might be, we'll put in a prop like that, get that garbage and think like, well, this isn't worth my time.
(09:48):
PeterWhere in fact, if they had put in a prompt that was more specific, that would detail the type of output that you want, etc., that they would have had something they were pleased with. So I think that that's kind of an overall recommendation of what to do is to make sure that you learn a little bit about prompt and what works, how to structure it.
(10:10):
PeterAnd, you know, especially if you're on the fence about, you know, is this valuable? Is this just, you know, hype and it's not your site and, you know, as we're discussing this, you know, there's plenty of trapdoors, but at the same time, it can be unbelievably valuable to you, to your business, to to everything.
GraceAnd I'll throw out a little mini tip that I've had success with when it comes to prompting. If you're not sure what the best prompt approach might be for a specific task that you're working on or something you need help with from ChatGPT, let's say ask the LLM what what the best format for your prompt might be so, and what your intention is.
(10:51):
GraceSo you might say, my goal is to improve my marketing strategy at the decision phase with my audience. This is the information that I have. What is the best way for me to frame a prompt so that you can help me out with that? That actually can be very successful, because then you can get information back from whatever land that you're working on, kind of walking you through how it approaches those sorts of questions and how it can get you the best results possible.
(11:23):
PeterYou know, last week we touched on, the the data that LLMs deliver, and we talked about structured versus unstructured data.
GraceYeah, I would love to build on that a little bit because I think, Peter, what you and I found when we were playing around with ChatGPT deep Research feature last week was that especially for that feature, Prompt Engineering was really influential on the results that you got back. And one important factor in that when you're using deep research, or even just a regular prompt on any limb, is to make sure there's a clear distinction between structured data and unstructured data.
(12:03):
GraceFor the most part, a lot of these limbs that everybody has access to right now do not handle unstructured data analysis well. And that would be more having an album analyze more vague data sets, like let's say you upload your spreadsheet of all your email marketing information and you say, let me know what the most successful email marketing strategy was from this data set.
(12:27):
GraceThat's going to be way too vague. What you what you can do instead is you can direct the LLM to a specific data field. I could say factor in email open rates. And tell me what the most successful sending strategy was in terms of time and day on impacting our email open rate. So that's something that's a lot more targeted and specific.
GraceAnd as long as the LLM has all of the data fields available to it to analyze that, you're going to get much more insightful results back. And if you're if you have a CRM or a marketing automation platform or something like that that's integrated with an LLM of your choosing, the same goes for that as well as to think about how you're guiding the prompts towards structured data.
(13:07):
PeterIdeally. Awesome.
GraceYeah. So I know we've thrown out a lot of tips in terms of prompting and navigating AI and ultimately making sure that you don't stop thinking for yourself. But I know, Peter, you wanted to mention a couple other things before we close out.
PeterYeah, I just wanted to give a little call to action. We're going to make a pivot. Not every episode, but why don't you? Episodes a month, we're going to start to have experts on the show to talk about different aspects of AI. And so this is a call to action. If you know somebody who would be fantastic, who will blow our minds, who will not deliver unstructured data, then, you know.
(13:51):
GraceAwesome. All right. Well, sounds like a plan. And I personally I'm excited to see how this podcast keeps developing. Thanks for your time today, Peter. I will talk to you next week.
PeterThanks, Grace. Bye, everybody.
GraceThanks for listening to Accelerated Velocity, a link to our website, newsletter and all that other good stuff can be found down in the show notes. If you like what you heard, we'd appreciate a five star review and make sure to subscribe for weekly episodes on the latest in AI for business.
Popular Podcasts
New Heights with Jason & Travis Kelce
Football’s funniest family duo — Jason Kelce of the Philadelphia Eagles and Travis Kelce of the Kansas City Chiefs — team up to provide next-level access to life in the league as it unfolds. The two brothers and Super Bowl champions drop weekly insights about the weekly slate of games and share their INSIDE perspectives on trending NFL news and sports headlines. They also endlessly rag on each other as brothers do, chat the latest in pop culture and welcome some very popular and well-known friends to chat with them. Check out new episodes every Wednesday. Follow New Heights on the Wondery App, YouTube or wherever you get your podcasts. You can listen to new episodes early and ad-free, and get exclusive content on Wondery+. Join Wondery+ in the Wondery App, Apple Podcasts or Spotify. And join our new membership for a unique fan experience by going to the New Heights YouTube channel now!
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!