After careful prompting in AI and having AI ask me very specific questions to ensure accuracy, the results are so bad they're unusable. Or so I thought. After I finished laughing at the irony of it, I realized it's perfect for my newsletter and my blog post this week.
My plan was to write an article on the dangers of relying too heavily on quick internet searches and AI for students and coders. The pitfalls, inaccuracies and how they would negatively impact your audit findings and accuracy rates was the focus of the article.
To backtrack a bit…
I've been very sick for over a week with COVID. 10 days. 10 very long days spent mostly in bed sleeping. The fever is finally gone and I'm at least finally able to sit upright so I wanted to do something productive. My newsletter is a day late, so I thought, why not use AI to help me write this article for my newsletter and get it done quickly? It's usually very good at outlining so why not try having it write the actual article for me?
What follows is exactly what AI gave me for the content of my article. I've decided to break this up into 3 parts so it's a faster read. This week we'll cover Pitfall #1.
Keep in mind, I gave AI the instructions to write this article about the pitfalls of relying on the internet and AI as our sole source of coding reference. I gave AI the made up story about Joan, Willow Hospital and the reason for the patient admission that Joan coded. AI did the rest all on its own. And taken in the context of what I wanted to write about, it perfectly illustrates the reason we cannot rely solely on internet searches and AI in our study and everyday coding.
AI-Generated Article — Pitfall #1: Outdated, Inaccurate or Non-Authoritative Sources
As a newly certified inpatient coder, you've got the fundamentals down—but when it comes to complex chart scenarios, a quick internet search can lead you astray. Beyond the allure of instant answers, unofficial web posts often miss key sequencing rules, combination codes, and the latest guideline updates. In this deep-dive, we'll explore three real-world case studies featuring Joan, a recent hire at Willow Hospital, to illuminate the risks of "Google-only" coding. You'll walk away with mentor-tested Pro Tips, reflection prompts, and a roadmap of go-to resources to build confidence and defend your work during audits.
Now this starts off great. Exactly what I asked for! Then it goes into the practice coding for the made up scenarios I gave AI to use. Back to the AI content…
Many web articles aren't updated annually—and ICD-10-CM changes happen every October. Relying on a random blog post can mean using superseded codes or misreading chapter notes.
Spotlight Case Study: Joan and the Diabetic Foot Ulcer
Joan coded a patient admitted for Type 2 diabetes with ulcer of the left toe and gangrene, who underwent excisional debridement down to muscle tissue. A quick internet search led her to use E11.621 (Type 2 diabetes with foot ulcer) and E11.52 (Type 2 diabetes with diabetic peripheral angiopathy without gangrene) —However, the 2025 ICD-10-CM Official Guidelines require a combination code, E11.6211 (Type 2 diabetes mellitus with foot ulcer, gangrene), to capture both ulcer and gangrene in one entry.
This is incorrect. Here is why:
E11.52 is actually Type 2 diabetes with diabetic peripheral angiopathy with gangrene (not without).
Code E11.6211 does not exist. AI made that up. There isn't one code for DM with ulcer and gangrene in ICD-10-CM.
Joan would be wrong to assign E11.52 unless the provider specifically documents gangrene due to diabetes. They have stated Diabetic ulcer and gangrene, so they've linked the ulcer to diabetes, but not the gangrene.
ICD-10-CM does not link gangrene to DM as it does with other complications, and if the provider doesn't state the cause of gangrene we cannot assume it is due to peripheral angiopathy.
The Correct Coding Is:
- E11.621 (Type 2 DM with foot ulcer), with an additional code from L97.4x–L97.5x to specify ulcer site and depth.
- Per the L97 instructional notes, assign extra codes for associated gangrene (I96) and associated diabetic ulcers.
- L97.4–L97.5 further define ulcer depth and necrosis.
What a Google Search Returned Instead
I also did an internet search where I asked what are the ICD-10-CM codes for diabetic ulcer and gangrene. Google brought back these results:
Diabetes with ulcer and gangrene is E11.52, Type 2 DM with peripheral angiopathy with gangrene. Then goes on to explain that the underlying cause of gangrene is often Type 2 diabetes and that the gangrene is a complication of the diabetes.
But nowhere in the provider documentation did they state the gangrene was due to Diabetes, only the ulcer is documented as Diabetic. And ICD-10-CM does not link gangrene to DM as it does with other complications. This one would require query to the physician to determine if the gangrene was also due to Diabetes.
This is also missing the code for the diabetic ulcer, E11.621 and the L97.- code for the site and depth of the ulcer.
The next article that comes up in a Google search discusses pressure, or decubitus ulcers, which are not coded to Diabetic ulcers and code differently when gangrene is present but incorrectly states to code to E11.621 and E11.52 and other codes.
Next is an AAPC wiki from the boards discussing this but it's from 2016 and it still isn't leading to the correct answer — many have overlooked that Gangrene is not automatically linked to Diabetes the way other complications are.
If you keep going you'll find articles that tell you the provider must link gangrene to DM, but by that time you could have already done the appropriate search in your code book or encoder and come up with the correct codes on your own.
The key takeaway here is you need to rely on yourself, your codebooks, guidelines and other resources that the coding community as a whole considers official and defensible. And you need to understand how much time it takes to search for accurate codes and how quickly you can learn to look those up on your own.
Next week we'll get back to Part 2 of the AI generated story…