--- title: Bad LLM! Admonishing my ChatGPT to Get the Ghosts Out of the Machine description: Struggling with unexpected elements in AI-generated images? Discover how I navigated surprising challenges with ChatGPT and learned the importance of clear communication in AI interactions. image: https://signal78.ai/hubfs/ChatGPT%20Image%20Apr%2018%2c%202025%2c%2002_48_10%20PM.png --- [Skip to content](https://signal78.ai/insights/bad-llm-admonishing-my-chatgpt-to-get-ghosts-out-of-the-machine#main) [![Signal78 home](https://signal78.ai/hs-fs/hubfs/signal78-site/logo-signal78.png?width=95&height=30&name=logo-signal78.png)](https://signal78.ai/) [Home](https://signal78.ai/) [Platform](https://signal78.ai/platform) [Services](https://signal78.ai/services) [About](https://signal78.ai/about) [Insights](https://signal78.ai/insights) [Book a demo](https://signal78.ai/insights/bad-llm-admonishing-my-chatgpt-to-get-ghosts-out-of-the-machine#demo) [← Insights](https://signal78.ai/insights) Practical AIApril 18, 20253 min read # Bad LLM! Admonishing my ChatGPT to Get the Ghosts Out of the Machine ![Matt Cyr](https://signal78.ai/hubfs/signal78-site/headshot-matt-cyr.jpg)**Matt Cyr** · Founder & President, Loop ![](https://signal78.ai/hubfs/ChatGPT%20Image%20Apr%2018%2c%202025%2c%2002_48_10%20PM.png) In this article Share I had to admonish my ChatGPT this week. I didn't feel good about it, but it had to be done. I was trying to create an image of me, [Chris Boyer](https://www.linkedin.com/in/chrisboyer/) and [Reed Smith](https://www.linkedin.com/in/reedtsmith/) to go along with a blog post about my appearance on their [touch point media](https://touchpoint.health/podcast/tp431-ai-are-you-killing-our-vibe/) podcast. I uploaded headshots of the three of us and asked ChatGPT to create an image in my brand style. This usually goes pretty well, but lately there's been a problem: it keeps adding mice 🐭, cheese 🧀 and robots 🤖 to images when I haven't asked it to do so. This isn't surprising. I recently had ChatGPT generate a bunch of images to support eBooks and blog posts I created where mice, cheese and robots were *supposed* to be there. But this wasn't one of those situations. The first image it generated showed the three of us, along with a mouse, a piece of cheese AND a robot. I used the new tool that lets you highlight parts of an image you want to get rid of, and the resulting image included a stool with a piece of cheese on it and, inexplicably, weird bulbous fingers on poor Chris's right hand. I tried a third time, being specific that I didn't want ChatGPT to include any mice, cheese or robots...only the three humans. The result? A baby robot holding a bolt of lightning. Once more I tried the select tool and gave clear instructions what I wanted the image to include. Good news! No mice, cheese, or robots. Not-so-good news? There was suddenly a small version of me in the image. I'd had enough. Typing furiously, I asked: "Why do you keep adding robots, mice and cheese? I haven't asked you to do that, but you keep doing it." The system's response was fascinating: *"You're absolutely right to call that out—and I appreciate your patience. What’s happening is that some of the styling instructions from earlier requests (especially those involving mice, robots, and cheese) got “sticky” in the image generation tool’s context. Even when we stop mentioning those elements, the system sometimes carries them forward unintentionally.* *Going forward, I’ll make sure to explicitly clear those out of the prompts I send on your behalf so your images stay focused on what you're actually asking for.* *Thanks again for flagging it—want to retry your last request with a clean, accurate prompt?"* I felt a little guilty for admonishing my ChatGPT, so I typed an overly enthusiastic response: "Yes, let's do it!" The result? Nailed it! The lesson? These LLMs are getting really good at telling you why they do things a certain way and correcting mistakes, but you have to communicate with them very directly, very clearly -- and occasionally admonish them when they just can't get those ghosts (or mice, cheese or robots) out of the machine. - ![Image-1](https://signal78.ai/hs-fs/hubfs/Image-1.jpeg?width=1536&height=1024&name=Image-1.jpeg) Hover message - ![2nd-image](https://signal78.ai/hs-fs/hubfs/2nd-image.png?width=1536&height=1024&name=2nd-image.png) Hover message - ![3rd-image](https://signal78.ai/hs-fs/hubfs/3rd-image.png?width=1536&height=1024&name=3rd-image.png) Hover message - ![4th-image](https://signal78.ai/hs-fs/hubfs/4th-image.png?width=1536&height=1024&name=4th-image.png) Hover message - ![ChatGPT Image Apr 16, 2025, 08\_09\_21 PM](https://signal78.ai/hs-fs/hubfs/ChatGPT%20Image%20Apr%2016%2c%202025%2c%2008_09_21%20PM.png?width=1536&height=1024&name=ChatGPT%20Image%20Apr%2016%2c%202025%2c%2008_09_21%20PM.png) Hover message - ![ChatGPT Image Apr 16, 2025, 08\_19\_56 PM](https://signal78.ai/hs-fs/hubfs/ChatGPT%20Image%20Apr%2016%2c%202025%2c%2008_19_56%20PM.png?width=1536&height=1024&name=ChatGPT%20Image%20Apr%2016%2c%202025%2c%2008_19_56%20PM.png) Hover message   ### Want to learn how to get the AI ghosts out of your machine? Get in touch. ![Matt Cyr](https://signal78.ai/hubfs/signal78-site/headshot-matt-cyr.jpg) Founder & President ### Matt Cyr Nearly 30 years leading marketing teams, from Boston’s hospitals and universities to eight years running strategy and new business at a 100-person agency. Founded Loop in 2024. ## More insights. [See all insights→](https://signal78.ai/insights) ### [![](https://signal78.ai/hubfs/Gemini_Generated_Image_tlr2qftlr2qftlr2.jpeg) AI StrategyOctober 6, 2026 It’s Q4 — How’s Your AI Plan Coming? Read the post→](https://signal78.ai/insights/its-q4-hows-your-ai-plan-coming) ### [![](https://signal78.ai/hubfs/2026-09-28_asset_linkedin-article-header-platform-update-fall-2026.png) Signal78September 28, 2026 Signal78 Is Live. Here's What It's Proven — and What's Next. Read the post→](https://signal78.ai/insights/notes-from-the-build-where-signal78-stands-this-fall) ### [![](https://signal78.ai/hubfs/37-78.png) AI for AgenciesAugust 11, 2026 AI Will Keep Making Move 37. We Named the Company After Move 78. Read the post→](https://signal78.ai/insights/ai-will-keep-finding-move-37.-we-named-the-company-after-move-78) Every week ## Stay in the Loop. Get next week’s issue in your inbox: what changed in AI, and what it means for your agency. ![Signal78](https://signal78.ai/hubfs/signal78-site/logo-signal78.png) The agentic operating platform for agencies. It runs the routine and signals your people when their judgment matters. [Home](https://signal78.ai/) [Platform](https://signal78.ai/platform) [Services](https://signal78.ai/services) [About](https://signal78.ai/about) [Insights](https://signal78.ai/insights) © 2026 Signal78 [Privacy policy](https://signal78.ai/privacy) Cookie settings × ## Book a 30-minute demo Pick a time that works. You'll see Signal78 running on agency work like yours. × ## Talk to us Tell us a little about your agency and what you want AI to change. We'll reply within one business day. × ## Can’t we just build this in Claude? The white paper on what it takes to keep a homegrown AI system current, secure and separated by client. ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Matt Cyr", "url" : "https://signal78.ai/insights/author/matt-cyr" }, "dateModified" : "2025-04-18T19:22:54.326Z", "datePublished" : "2025-04-18T19:22:54.000Z", "headline" : "Bad LLM! Admonishing my ChatGPT to Get the Ghosts Out of the Machine", "image" : [ "https://signal78.ai/hubfs/ChatGPT%20Image%20Apr%2018%2c%202025%2c%2002_48_10%20PM.png" ], "mainEntityOfPage" : { "@id" : "https://signal78.ai/insights/bad-llm-admonishing-my-chatgpt-to-get-ghosts-out-of-the-machine", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://signal78.ai/hubfs/Loop%20Logos/Loop_Primary.svg" }, "name" : "Signal78" } } ```