AI Slop Has Arrived in Solar Design

It didn't take long for AI slop to make its way into the solar industry.

We've already seen AI-generated articles, sales emails, product reviews, and marketing content (cough Linkedin cough). Now we're starting to see AI-generated single-line diagrams show up in service tickets, permit packages, and design reviews.

Recently, our service team received this diagram:

AI generated single line diagram

To be fair, it wasn't completely wrong. In fact, that's what makes it interesting. The danger wasn't that it looked terrible. It was that it looked legitimate (sort of).

For years, bad solar documentation was easy to spot. Missing equipment, hand-drawn corrections, random notes copied from previous projects, a coffee stain or two, and enough arrows to qualify as modern art. Most reviewers could immediately tell the drawing needed additional work. AI has changed that equation.

Today's AI tools are remarkably good at producing professional-looking documentation. They know what a single-line diagram should look like. They know common equipment names. They know solar terminology. They know where boxes and lines generally belong.

What they don't know is whether those boxes and lines actually represent a coherent system. The result is documentation that appears professional while quietly hiding ambiguity, missing details, and questionable assumptions.

The "It Looks Right" Problem

The diagram included Tigo EI Inverters, Tigo batteries, Acrel meters, a Tesla Backup Gateway, Tesla Powerwalls, and a PV array identified as using Tigo optimizers. Nothing immediately jumps off the page as completely ridiculous. In fact, many installers would probably glance at it and move on.

That's exactly why it's worth discussing.

A good single-line diagram should answer questions. This one created them.

• Why are there three separate battery groups?

• What role are the Acrel meters performing?

• How are the Powerwalls integrated into the architecture?

• What exactly is the Tesla Backup Gateway doing in this design?

• Where are the disconnects?

• Where is the rapid shutdown equipment?

• What conductors are actually being installed?
  

The more I looked at the drawing, the more questions I had, and that's not what a line diagram is supposed to do.

The Battery Numbers

Then there were the battery references. Under each inverter, the diagram includes labels such as:

• Batteries 68,79

• Batteries 77,35

• Battery 94

At this point, I found myself staring at the drawing and wondering: What are these, World of Warcraft map coordinates for a quest in Stranglethorn Vale? (yes, nerd alert...)

The diagram never explains.

This is one of the recurring issues I see with AI-generated technical content. The AI found information that appeared important and inserted it into the drawing, but it never provided the context needed to understand what the information actually means.

The labels look technical, intentional, and important. Yet they provide no useful information to the person trying to understand the system.

If a service technician, installer, designer, AHJ, or utility reviewer has to stop and ask what a label means, the drawing isn't simplifying the design. It's adding ambiguity- the last thing you want in electrical documentation. Particularly for those few who RTFM.

AI Loves Components

AI-generated technical content loves nouns. It knows equipment names, acronyms, product families, and enough industry terminology to sound convincing. What it often struggles with is understanding relationships.

Real system design isn't about collecting component names and arranging them on a page. It's about understanding how those components interact and how the system behaves as a whole. It's judged by whether someone can understand, install, commission, troubleshoot, and review the system using the information provided.

The Real Risk

The biggest risk isn't that AI gets things wrong. We've been doing that on our own for decades. The real risk is that AI-generated documentation creates a false sense of confidence.

An installer sees a polished diagram and assumes it was reviewed.

A homeowner sees a polished diagram and assumes it was engineered.

A salesperson sees a polished diagram and assumes it is ready to submit.

The cleaner the document looks, the less likely people are to question it. That's dangerous. Because, unlike a calculator, AI doesn't know when it's guessing.

AI Is a Tool, Not a Designer

I'm not anti-AI. I use AI regularly. It can help organize information, summarize documents, brainstorm ideas, and eliminate repetitive work. Used correctly, it's a powerful productivity tool.

But AI is not engineering judgment. It doesn't understand manufacturer requirements, utility requirements, the intent behind electrical codes. Most importantly, it doesn't understand consequences. It predicts what should appear next based on patterns it has seen before. Sometimes that's incredibly useful. Sometimes it creates a document that looks complete while quietly omitting the information that actually matters.

The challenge for our industry isn't deciding whether to use AI. That decision has already been made. The challenge is remembering that a professional-looking document is not evidence of a professional design.

Technical judgment still matters.

Fun fact!

Did you know AI doesn't use real fonts when it generates an image with text? When you ask AI to create a graphic, it is not loading a font library. Instead, it's doing something more like:

"I've seen millions of posters that contain big bold text. I'll draw something that resembles those letters."

The AI is literally generating pixels. It is drawing the appearance of letters, not typesetting text.

That's why AI-generated graphics have errors like this:

• "SOLAR" becomes "S0LAR"

• "POWER" becomes "POWFR"

• "TIGO" becomes "TI6O"

The AI isn't reading words. It's drawing what it thinks words look like. But I digest...

If your single-line diagram contains mystery battery coordinates, multiple energy-storage architectures, and enough unanswered questions to keep engineering and service arguing for an hour over Teams, it might be worth having an actual (competent) human being review it before hitting Submit.

Or just go back to cocktail napkins. They contain about the same amount of engineering, but they're honest about it.