Do Not Let AI Write Checks Our Knowledge Can't Cash

A home inspection report is only as strong as your ability to stand behind every line of it and AI is testing that in real time. We pick up a thread from our conversation about legal issues with veteran inspector and legal expert Joe Ferry and dig into what happens when AI starts writing your narrative comments for you. The tool is powerful, but the moment your report sounds polished while your explanation sounds unsure, your credibility takes the hit.

We talk about the “credibility gap” that shows up when clients call with questions and you have to hunt through your own report to figure out what a paragraph means. AI can also inflate simple findings into long, authoritative language that feels disconnected from what you said on site. At a time when people are increasingly sensitive to AI written content, that mismatch can make you look less human and less trustworthy, even if your inspection work is solid.

Then we get into the liability side of AI in home inspections: impeachment in deposition, negligence arguments, and why a report you cannot defend is a report that can expose you. We also explore uncomfortable but real discovery concerns, like whether opposing counsel could push for drafts or AI records, and what E&O insurance carriers might think as AI becomes more common in report software.

We end with practical guardrails that keep AI useful without letting it drive the bus: using it for grammar, tightening your own draft, and reviewing every comment so your report stays clear, honest, and defensible. If this helps, subscribe, share it with another inspector, and leave a review so more people can find Inspector Toolbelt Talk.

Check out our home inspection app at www.inspectortoolbelt.com
Need a home inspection website? See samples of our website at www.inspectortoolbelt.com/home-inspection-websites

*The views and opinions expressed in this podcast, and the guests on it, do not necessarily reflect the views and opinions of Inspector Toolbelt and its associates.

Transcript

Ian Robertson  
Welcome back to the show, everybody. Today on Inspector Toolbelt Talk, I wanted to expand on something that Joe Ferry and I talked about last week on our episode with him. By the way, fantastic episode. Love that guy, just a wealth of 40 years of home inspection legal stuff, and he's a great guy. But we talked a bit about using AI in our reports. Now, listen, I'm not saying don't use AI, because I mean, it's everywhere, it's my favorite word, it's ubiquitous, right? It's all over the place. And using it for comments, I can't tell you how many times it's helped me personally wrap my head around how to say something. So it's a valuable tool, but it's also something that we want to be careful with. And the reason I bring that up is because we have seen something very interesting start to unfold in the industry.

Years past, writing narratives was a very specific skill, and actually one of probably the hardest skills for us as home inspectors to develop. I know for me it was. How do I write this up? We don't see that question very often anymore. How do I write this up? Because we go to AI and say, hey, here's all the facts, write it up for me. So it's great, you know, even our app, we have AI built into it, and we're expanding its uses and things like that, but there's a bit of a thing that's happened, and one of them was an instance where a home inspector, I asked him about a comment, and he goes, "Oh, I don't know, AI wrote that for me." And it immediately set off bells and whistles. And then I started reading through forums and talking with other inspectors, and I found a very similar theme, is that there's a divergence between us as inspectors and our comments. So, a comment is very personal, or at least it used to be. I used to spend years developing some of my comments, so when somebody said, "Hey, I saw this in your report." All day long, I could just sit there and talk about that comment. I could talk about why we wrote it that way, what the variables are, how we came to that conclusion. We could talk on the subject matter, even more importantly Talking about CMUs? Sure, all day long. But I've even had several inspectors I've personally talked to that had that same response as that inspector I just mentioned. I don't know, or I'd have to look that up. I'd have to see what that comment says. There's a danger in that, and here's a couple of reasons why. 

First of all, our market credibility—it really does hurt it. People still want that human interaction. We write in our report this long, wordy comment about, you know, crack variability, and frost lines, and all this other stuff. And then somebody calls us and says, "Hey, I have a few questions. You wrote this." We're scurrying, looking through the report, and we might even see a paragraph in our comment, which, by the way, that's—if we're reading through multiple paragraphs—probably too long of a comment. But we're reading through it, and we're like, I don't know what that means. Now we're quick looking it up, and they're like, "Hey, are you there?" Or even worse, on site. Like, we tell them one thing, and we say, oh, yeah, this crack's too big. And now they get a report with four paragraphs about this crack. And then people are like, hey, you gave us a very simple explanation on site, and then you wrote us a book. And that's what AI does. It just makes things big and fluffy, and adds more volume to it. That ruins our market credibility pretty quickly. In fact, just having this conversation with a home inspector the other day, who wants to change his messaging. So, hey out there, because he knows who he is, and he listens, but he wanted to change his messaging, and I said, don't change it to, hey, we utilize AI. Go just the opposite, saying yes, we utilize AI, but we use our expertise to get human eyes, real human contact with AI used as a tool. However we want to message that, people are looking for that more now than they were before. We're used to AI, but how many of us, when we have like a banking problem, and the bank will have that little pop-up window when we're talking with the AI, the most common response is human agent, the request for a human agent, not to talk to the AI. We as a society are becoming very sensitive to AI written content, so it hurts our market credibility. 

But talking about our experience with Joe Ferry, we dug into it a little bit more, and I am not a legal expert. Check with your attorney. Heck, check with Joe Ferry, but these are some of the reasons that we were able to find or possibilities that our reports can bite us in the butt later on if we use AI too heavily without fully understanding the issues that we're reporting on. Now, the old guys listening like me, listening to this podcast are saying, yeah, these new guys, they're using AI, and they don't understand this issue. Before they'd have to learn about it to write a comment, and there is truth to that. You'd have to fully learn and understand a subject before you could write a comment about it. Now, AI just does it. So there is a measure of not having to learn things as hard and as quick. But I see a lot of us older guys, guys like me, and anybody here listening that's been doing this for 20 plus years, we're using it heavily too. Here's a couple of the issues. 

The first one is the inspector can be impeached during deposition. I had never really known this. Being impeached during deposition. So, can you imagine your client suing the seller. That's probably the most common lawsuit that I've seen, usually not even suing the inspector, but now you get impeached during deposition. So, basically, like an attorney will ask, 'Hey, you wrote the differential settlement may indicate ongoing structural movement. What observations led you to that conclusion?" If the inspector says, "I'm not sure, AI generated that comment." It automatically damages your credibility. So they juries and courts tend to give less weight to somebody that does not seem to be an expert in their field. 

Second thing, it can make us look negligent, and that's the worst word. Negligent is the worst word. Remember every podcast that we've had about legal stuff, every attorney I've ever talked to has agreed, it's near impossible to sue an inspector successfully unless you can prove negligence, specifically gross negligence. So, here's a scenario: an attorney could argue a reasonable inspector would understand every statement in his report before delivering it to a client. AI is almost never liable. It's always the person who's using it. Remember, AI is a power tool. You don't sue the Sawzall company when you cut off your finger, you sue the person who was using the Sawzall. So the problem is basically writing a report where we can't stand behind every statement. Now we all say, well, I check every comment. I mean, most of us do, but come on, let's be honest. Have you ever put a comment in the report, you're three reports deep, and you're just exhausted? Did you really read through that comment completely and thoroughly, and then change a couple of sentences? Most of us will probably hop over a couple. Would we ever be surprised if somebody came back—because we're not as connected to our comments when AI writes it—would we be surprised when somebody says, 'Hey, you wrote this in the report," and you have to think, what did I mean by that? That's why canned comments actually had a little bit of an advantage. I knew what I meant, because I had that comment in hundreds of reports, when I found the same exact thing. There was a little bit of connection there, as we mentioned at the beginning.

The third point, the report becomes harder to defend. That's a problem if we're being sued and we can't defend the report because we don't understand it, that really opens us up to a lot of liability. 

The fourth one, so I had to do a little bit of extra digging, because this was pretty interesting. So, AI often writes language that sounds authoritative, and the more we use an LLM, any AI model, it's going to start writing like we do, or it's going to learn how we want things stated, and we as home inspectors, want things stated authoritatively, right? So, as it does that, it might write things that we don't fully understand or appreciate. So the argument would be, if we have this great comment about horizontal cracks in a CMU foundation, cool, but now if we get questioned on it, how much do you know about CMU foundations? What's the average expansion rate? What's the frost line in your area compared to an area that's four hours north. What about this? What about that? If we can't answer those questions, they're going to basically make the argument that we wrote something that exceeds our expertise, and that's incredibly dangerous, because now we're overselling ourselves. It's kind of like hiring a doctor to do heart surgery, and he was actually just a foot doctor, and the doctor's like, well, I'm a doctor, and I know a lot about hearts, and he botches the surgery. That doctor is probably going to lose that lawsuit, because they're like, you oversold your expertise. That one really got me. 

Five, the opposing counsel may request draft notes and AI records, and we already know from some of the court rulings that AI records for an individual can be called upon at any time by a secular authority. I think most of us would probably not want everybody to see everything we've asked AI. Hey, AI, what's that weird mole in my belly button, is it cancer? What if I pick at the mole? You know. I don't have a mole in my belly button, by the way, and that's a weird example, but there's a lot of weird stuff probably on our AI account. Can you imagine having that be read in a court of law? Everybody gets a copy of it. I thought that one was probably the weirder ones, But if they're like, hey, this inspector used AI to create this comment, we'd like to subpoena all of his records. That's got to be an interesting one. 

The sixth one, I have not been able to find evidence of, but it's still something I want everyone to keep in mind. E&O insurance carriers may not be happy about AI. There's not enough precedence out there, from what I can tell, and I'm not a legal expert, to know what insurance companies are going to do when we have a bunch of AI slop written into a report. 

So, are we able to defend our comments? Are we able to explain our comments? Do they oversell our expertise and make us sound smarter than we are? If any of those are even remotely true, step back and look at what we're using AI for. Here's what I personally do to make sure that I use AI in a way that is sustainable for my business, and to keep liability down, and to keep my market reputation down. 

Number one, silly simple grammar and language. I use it for that all the time. I do not know where to put a comma, and I am 44 years old, and I mess up on it all the time. Clearing up a thought, like writing our full thought out, writing our full comment out, and telling AI, make this clearer without changing the thought. That's much better. I have more connection to the comment, I know what I was trying to say, and then it usually does a pretty good job on it.

Number two, just reading through every comment, checking it, knowing what AI is saying. It's like having a really lame employee write things up for you. That employee can get us into a lot of trouble. So, again, I'm not saying don't use AI, but let's be careful about how we use it. Make sure that we use it in a way that protects our business, our future, our clients, and the whole kit and caboodle. Thank you very much for listening in, and we'll see you next time on Inspector Toolbelt Talk. 

Outro: On behalf of myself, Ian, and the entire ITB team, thank you for listening to this episode of Inspector Toolbelt Talk. We also love hearing your feedback, so please drop us a line at info@inspectortoolbelt.com.

If you’re enjoying the conversation, don’t forget to hit the subscribe button. Our podcast is available on all major podcast platforms. For more information on our services and our brand-new inspection app, please visit our website at Inspectortoolbelt.com.

*The views and opinions expressed in this podcast, and the guests on it, do not necessarily reflect the views and opinions of Inspector Toolbelt and its associates.