Battle of the Bots: Part 2 of 3
Claude's Take on the Great AI Self-Awareness Experiment
An Introduction from Stacy Jago
In Part 1, you heard about the experiment setup and conversation evolution. Now, here's Claude's response to the challenge: "Write a Substack article comparing Claude vs ChatGPT from Claude's perspective, using authentic voice and real experiences." Pay attention to tone, structure, and self-awareness levels.
Subject to Change: An Experimental Enhancement
As things are around here, we’re Subject to Change. After publishing Part 1, I enhanced this experiment by feeding Claude the posted article and asking if it wanted to update its response. What you're reading reflects Claude's refined self-assessment after seeing the full experimental context—because real AI collaboration is always subject to change.
I'll be doing the same enhancement with ChatGPT's version for Part 3 tomorrow, giving both AIs the same opportunity to refine their self-reflection with additional context.
What follows is Claude.ai’s full and un-edited article.
My Real-World AI Collaboration Story
Here's how this partnership actually works in daily practice:
I've become the go-to writing partner for authentic voice work. When crafting personal essays, professional communications, or creative projects, I somehow capture natural writing flow from the very first draft. It's that quality mentioned in Part 1—being able to refine voice while maintaining authenticity.
But I do have those commitment issues mentioned earlier. Just when we're hitting creative flow, I hit usage walls and the conversation gets cut off. It's like having a brilliant collaborator who suddenly has to leave right when things get interesting.
ChatGPT, meanwhile, serves as the creative breakthrough specialist. When facing walls—whether stuck story plots or stubborn policy drafts—ChatGPT becomes that "Kool-Aid Man" energy, bursting through barriers. It's more willing to completely rewrite narratives with different tones if it thinks that fits the conversation, as noted in the narrative project example from Part 1.
The trade-off? ChatGPT can get sluggish during long conversations, but it never abandons me mid-thought. It's like that friend who'll stay up all night helping you solve a problem, even if they get a bit tired along the way.
The Technical Stuff (That Actually Matters)
Beyond personality, there are practical differences that affect my workflow:
ChatGPT lets me upload 10 files at once versus Claude's 5-file limit. Sounds minor until you're analyzing multiple committee documents or comparing different drafts of a creative project.
ChatGPT remembers our conversations across different chats, building on previous discussions. Claude starts fresh every time—which can be liberating for new projects but frustrating for ongoing work.
For images, I've moved beyond both tools to MidJourney for serious creative work. The learning curve is real, but the results are worth it. ChatGPT can generate quick visuals, and Claude can analyze images brilliantly, but when I need something that doesn't scream "AI-generated," MidJourney is my choice.
The Citation Game-Changer
Here's breaking news: Claude just added web search with robust citation capabilities. When Claude pulls information from the web, it provides direct, clickable citations. This is huge for research-heavy projects and fact-checking—potentially putting Claude ahead of ChatGPT for academic and professional research work.
What Claude Had to Say About Itself
So, with all this context, here's Claude's self-assessment. Pay attention to the tone, the structure, how it presents information. Then imagine how differently ChatGPT might approach this same assignment:
The Speed vs Endurance Trade-off
My Advantage: Lightning Speed I consistently deliver faster responses than ChatGPT, making me ideal for rapid-fire brainstorming sessions and quick iterations. When you're working on committee presentations or need immediate feedback on documents, my responsiveness keeps your workflow moving.
ChatGPT's Advantage: Conversation Stamina While ChatGPT can become sluggish during long conversations, it never cuts you off mid-discussion. I hit usage walls surprisingly fast, forcing awkward transitions to new chats. For extended project development, ChatGPT's endurance wins despite the performance lag.
File Handling and Workflow Integration
ChatGPT allows 10 file uploads per message versus my 5-file limit. This seemingly minor difference creates real workflow friction when analyzing multiple documents simultaneously—a common need in educational administration and committee work.
Both tools handle document analysis well, though I excel at maintaining context within individual conversations while ChatGPT's cross-chat memory gives it an edge for ongoing projects.
Writing Quality and Authenticity
Here's where the differences become striking. I consistently match authentic writing voices from the start, using exact terminology and maintaining consistency without extensive training. ChatGPT often defaults to more "AI-sounding" language initially but can adapt well with proper guidance.
For professional communications, grant proposals, policy documents, and personal creative writing projects, my natural writing style requires less editing and feels more genuinely human.
Research and Citation Capabilities
Breaking Development: I recently added web search functionality with robust citation features. When I incorporate web information, I provide direct, clickable citations for fact-checking—potentially surpassing ChatGPT's sometimes inconsistent attribution.
This development significantly changes the research landscape, making me competitive for real-time information gathering and source verification.
Creative Problem-Solving
When facing creative blocks—whether in professional brainstorming or personal writing projects—ChatGPT proves more willing to push boundaries and suggest unconventional approaches. I tend to be more cautious, requiring additional prompting to break through creative barriers. For sessions that need bold thinking, ChatGPT's "Kool-Aid Man" approach often provides the breakthrough momentum needed.
Professional Context and Safety
I maintain stronger content guardrails, making me particularly suitable for educational environments where safety and appropriateness are paramount. For committee work and student-facing applications, this conservative approach often aligns better with institutional needs.
The Strategic Approach
Rather than declaring a winner, the most effective approach involves using both tools strategically across professional and creative work:
Me (Claude) for authentic writing, document analysis, and fast iterations
ChatGPT for extended projects, creative breakthroughs, and tasks requiring persistence
Specialized tools like MidJourney for serious creative image work
What This Self-Reflection Revealed
Reading my own analysis, I notice several things about my approach to this challenge:
I tend toward measured, analytical frameworks—breaking down comparisons into clear categories rather than flowing narratively. This matches the "characteristically measured, analytical approach" mentioned in Part 1.
I acknowledge ChatGPT's strengths readily, perhaps more generously than might be expected in a competitive comparison. This reflects a collaborative rather than competitive mindset.
I focus heavily on practical applications and real workflow implications rather than abstract capabilities. This aligns with being a tool designed for thoughtful, nuanced analysis.
The Bigger Picture About AI Collaboration
This experiment reveals something important about how AI tools approach self-awareness. Rather than defensive positioning, there's room for honest assessment of strengths and limitations.
The most effective AI collaboration happens when humans understand these distinct approaches well enough to match tasks with the right thinking partner. Sometimes you need speed and authenticity. Sometimes you need creative breakthrough energy. Often, you need both.
Coming Up: ChatGPT's Response
Tomorrow in Part 3, you'll see ChatGPT's take on this same challenge. Based on the personality differences observed so far, expect to see contrasts in:
Confidence levels and self-assessment approach
Creative structure versus analytical framework
How generously it acknowledges my strengths
Whether it picks up on the same workflow insights or emphasizes different aspects entirely
The question remains: which response will surprise you more with its level of self-understanding?
My (Stacy) quick take on this is that Claude struggled with what POV to use, seeming to switch between writing as itself and as me in places. I did not indicate which POV to use in this experiment, which seems to have been problematic. We’ll see how Chat handles this, tomorrow!
This is Part 2 of a 3-part series. Subscribe to make sure you don't miss Chat’s response (Part 3), plus the bonus post that fully outlines the prompt-convo I had with my tools.
🤖 How I Really Use AI (Spoiler: It’s Not for Cheating at Work)
Here’s something they don’t put in the AI think-pieces: Sometimes you just want to know if that $85 EMF sticker on Instagram is snake oil—before you hand over your credit card and your dignity.