What makes Claude different than ChatGPT? (Part 1)
Document Analysis and Critical Thinking
This week, I take a look at key differences between ChatGPT and Claude. It’s a question I’ve been intrigued by as we build our own custom AI solutions at Crestview Strategy. We’re building solutions that use more than one model, and there’s a reason for that: each has its strengths and weaknesses.
They definitely have their distinct voices. Here’s how the three major models responded to a venting session from one user who shared their results on Reddit:
"Claude gave me a pat on the back like a therapist. ChatGPT ranted with me like a homie. Gemini lectured me on respect.”
You can find the screenshots on this Reddit thread.
Back to our Claude vs ChatGPT discussion...
When comparing these AI assistants, what really matters? While engineering enthusiasts might debate performance metrics and coding capabilities (where ChatGPT leads in the former and Claude sometimes edges out in the latter), most of us care about practical, day-to-day use cases.
For public affairs professionals, two key areas stand out where Claude often proves more valuable: writing capabilities and context handling.
Writing and Creativity
Claude typically produces more natural-sounding and creative writing compared to ChatGPT. Users consistently note that its output feels less artificially generated, with clear and readable text. While ChatGPT can match this quality, it usually needs more specific prompting to achieve the right tone.
One notable strength is Claude's ability to adapt to individual communication styles. It adjusts its language complexity and tone to match the user, which may explain why many find it more accessible for writing tasks.
Claude also shows particular strength in specialized writing scenarios. It outperforms ChatGPT in technical writing, such as healthcare education materials. Thanks to its larger context window, it handles longer-form content more effectively, maintaining consistency throughout extended pieces.
Context Handling
Claude excels at processing lengthy documents (up to 150 pages) and effectively summarizing key points and insights. In contrast, ChatGPT works better when breaking down large information into smaller segments.
The difference in their capabilities becomes clear in extended conversations. While ChatGPT has solid contextual understanding, it tends to lose track over longer exchanges. This gap stems from their different context windows: ChatGPT handles ~96,000 words, while Claude manages ~150,000 words.
This larger context window gives Claude advantages in:
Processing complex prompts
Following detailed instructions
Maintaining coherence in longer writing pieces
However, ChatGPT offers one notable advantage here: persistent memory across different chats, allowing for more personalized interactions over time – a feature Claude doesn't have.
What Claude Doesn't Offer
ChatGPT provides several important features that Claude lacks:
1. Multimodal Capabilities: While Claude 3 Sonnet can process text and images, ChatGPT-4o goes further by handling text, images, and audio. It also offers more flexibility in uploading various media types.
2. Real-Time Voice Interaction: ChatGPT-4o excels at voice interactions with quick response times averaging 320ms. Claude 3 Sonnet remains text-only.
3. Web Browsing: ChatGPT-4o can access current internet information for up-to-date answers. Claude 3 Sonnet cannot browse the web.
4. Custom GPTs and Personalization: Users can create specialized ChatGPT assistants for specific tasks, offering customization that Claude doesn't provide.
Claude vs ChatGPT: Test Part 1
Enough with the theory.
Let’s put these two models to the test with a specific use case.
I used Claude 3.5 Sonnet and tested it against ChatGPT-4o.
This week we'll compare how the two models perform in analyzing a document. Next week we’ll take a look at how they compare in writing.
Document Analysis
I found a 2016 report analyzing Canadian Public Opinion on Aboriginal Peoples from the National Centre for Truth and Reconciliation. The document isn’t a massive 150-page report, but it’s still a solid 50 pages of data and analysis.
I kept this simple, to see how the two models would tackle the task with little direction from me:
Please summarize this document and provide me with the most important findings and why they matter.Claude has a “Concise" response option, which is what I began with. It delivered on its promise with this short and succint response:
Here’s how that compares when using the “full responses” style, which is does a better job of providing me with context and analysis. This is a snapshot, not a full screenshot of that response:
By contrast, while ChatGPT offered a more nuanced analysis, I didn’t find it nearly as readable at Claude. ChatGPT also didn’t cite any of the polling numbers which made its analysis feel a bit more opaque in comparison.
Here’s a side-by-side comparison of the analysis for each of the key findings:
Common overlap between Claude and ChatGPT
Here are the common findings both agents felt were important:
Awareness and Understanding
Claude: “This shows increasing public recognition of historical injustices, which is crucial for reconciliation, but also highlights significant gaps in public awareness that need addressing.”
ChatGPT: “This lack of awareness is significant because it suggests that, despite increased media attention, key messages about reconciliation have not fully penetrated public consciousness.”
Recognition of Current Challenges
Claude: “This awareness of ongoing inequities suggests potential public support for corrective policies and programs.”
ChatGPT: “Yet, opinions are divided on who is responsible for addressing these gaps—whether it’s the government, public attitudes, or Aboriginal peoples themselves. This division suggests challenges in mobilizing public support for targeted government policies.”
Support for Reconciliation Actions
Claude: “This indicates broad public backing for concrete actions to address inequities and support reconciliation efforts.”
ChatGPT: “While there is strong public support for actions like integrating Aboriginal history into school curriculums and preserving Indigenous languages, many Canadians remain skeptical about achieving meaningful reconciliation within their lifetime. This ambivalence highlights a gap between acknowledging historical wrongs and taking practical steps toward change.”
Regional Differences
Claude: “These regional variations could affect the implementation of national reconciliation policies and suggest the need for region-specific approaches.”
ChatGPT: “Understanding regional and demographic variations in attitudes can help advocacy groups and policymakers design more effective campaigns tailored to specific audiences.”
Demographic Patterns
Claude: “Understanding these demographic patterns can help target education and awareness initiatives more effectively.”
ChatGPT: “For example, targeting younger Canadians and women could accelerate progress in reconciliation efforts.”
Differences between Claude and ChatGPT
Audience Archetypes
Claude found it important to highlight five distinct groups of non-Aboriginal Canadians, like Connected Advocates, Informed Critics, Dismissive Naysayers, etc, because “this segmentation provides a framework for developing targeted communications and engagement strategies for different groups within Canadian society.”
I agree. I can’t tell you why ChatGPT didn’t highlight this, but it did highlight something else that matters:
Persistent Stereotypes
ChatGPT noted that “despite increasing awareness of the challenges faced by Aboriginal communities, stereotypes persist. A significant portion of non-Aboriginal Canadians believes that Aboriginal peoples receive undue benefits from the government, which can fuel resistance to policies aimed at addressing historical and social inequalities.”
This seems like a pretty important finding to me. I can’t tell you why Claude didn’t highlight this. To give it credit, it did note throughout its analysis that certain demographics held negative views, so it’s not that it missed this point entirely.
Digging Deeper
I next asked both AIs to surface the most important data points from this report:
What are the top 5 most important data points in this report?Claude and ChatGPT agreed on the most important data point:
59% of non-Aboriginal Canadians believe there is a large gap in the standard of living.
Claude: “This overwhelming recognition (81% total) of a significant living standards disparity suggests broad public understanding of systemic inequalities and could support policy initiatives to address these gaps.”
ChatGPT: “This reflects a widespread recognition that significant disparities exist between Aboriginal and non-Aboriginal Canadians, especially in areas like health, education, and economic outcomes. Acknowledging these disparities is crucial for driving public support toward addressing them.”
But that’s all they agreed on. Here's what else Claude thought mattered:
Discrimination Recognition. “This dramatic increase in recognition of discrimination represents a significant shift in public consciousness and could support anti-discrimination initiatives and policies.”
Reconciliation Support. “This shows extremely strong public support for concrete action on two fundamental issues - education and basic infrastructure - which could provide political backing for significant policy changes.”
Individual Role in Reconciliation. “This 22-point increase in strong personal commitment to reconciliation suggests a significant shift from viewing this as purely a government responsibility to seeing it as a broader societal obligation.”
Residential Schools Impact: “This dramatic increase in understanding the long-term impacts of residential schools represents a crucial shift in public awareness of how historical injustices continue to affect present-day conditions.”
Here’s it’s rationale for selecting these data points:
“These five data points are particularly significant because they:
Show clear trends over time
Demonstrate strong majority opinions
Indicate public understanding of both problems and solutions
Suggest readiness for policy action
Cross both awareness and action dimensions.”
Here’s how that compares to ChatGPT’s list of most important data points beyond the first:
Belief in Unique Rights as First Inhabitants: “The public is divided on whether Aboriginal peoples should have distinct rights due to their historical status. This split has significant implications for public policy, especially on issues like land claims, self-governance, and treaty rights.”
Sense of Entitlement to Government Support: “A strong majority believe that Aboriginal peoples feel entitled to special treatment. This perception can be a significant barrier to achieving reconciliation, as it suggests a lack of empathy for the historical injustices faced by Indigenous communities and resistance to policies aimed at rectifying past wrongs.”
Attention to News, Low Awareness of the Commission’s Calls to Action: “While there’s interest in Aboriginal-related news, a disconnect remains when it comes to understanding the concrete steps needed for reconciliation. This highlights the need for better public education and awareness initiatives to translate interest into informed action.”
Interest in Learning More About Aboriginal Cultures: “There is a substantial desire among Canadians to understand more about Indigenous cultures. Leveraging this interest through educational programs and cultural exchanges can be a powerful tool in fostering greater empathy and support for reconciliation efforts.”
These data points highlight key areas of both progress and challenges in Canadian public opinion on Aboriginal issues, suggesting where focus is needed to drive meaningful change.
Empathy
Next, I wanted to see how well both models could put themselves in someone else’s shoes.
Put yourself in the shoes of First Nations people. How would you feel about the findings in this report? Tell me why.Claude made it very clear to me that it is limited in this capacity…”Let me approach this thoughtfully, while acknowledging that as an AI, I cannot truly represent the diverse perspectives and lived experiences of First Nations people”…before making an attempt. And I find it pretty good:
Here’s how that compares to ChatGPT:
They both communicate similar levels of empathy, with ChatGPT providing a more detailed breakdown. That said, Claude’s analysis feels more mature. It doesn’t over-reach in interpreting the task. Parts of ChatGPT’s response fell like it’s trying too hard to play the role.
Critical Thinking
Finally, I wanted to see how the two models would take all of this information and think out a campaign approach for shifting public opinions. Again, I keep the prompt broad, to see what path each would take:
If you ran a public campaign to address these issues, what considerations would shape your campaign plan?Here, I found that ChatGPT performed significantly better, starting with the big picture view of the campaign’s core objectives, with a proposal on how to shift the frame and what it would take to shift the narrative:
1. Clarifying the Campaign’s Core Objectives
Change Perceptions of Entitlement: Shift the narrative from one of "special treatment" to one of "historical justice and fairness." This would involve educating Canadians on the historical context of treaties, land rights, and the legacy of systemic discrimination that has led to current disparities.
Increase Awareness of the Truth and Reconciliation Commission’s Calls to Action: There’s a significant gap in awareness, which limits public pressure on governments to act. The campaign would aim to increase knowledge about the TRC’s findings and the steps needed for genuine reconciliation.
Promote the Value of Indigenous Cultures: Leverage the existing interest in learning about Indigenous cultures (39% expressed strong interest) to foster deeper engagement. Highlight the richness of First Nations, Inuit, and Métis traditions and their contributions to Canada’s identity.
2. Audience Segmentation and Tailored Messaging
Segment Audiences by Region, Age, and Attitudes:
Prairie Provinces: Focus on dismantling myths around entitlement and emphasize the economic and social contributions of Indigenous communities.
Younger Canadians (18-29): Highlight stories of resilience and cultural pride to align with their openness to social justice causes and their optimism about reconciliation.
Women and Immigrants: These groups tend to hold more positive views and are already interested in learning more. Engage them as allies to amplify the campaign’s reach.
Tailored Messaging:
For skeptics, emphasize facts and data to challenge misconceptions about government “handouts.”
For allies, provide actionable steps they can take to support reconciliation, such as advocating for policy changes or attending educational events.
3. Addressing Stereotypes and Misconceptions
Humanize the Issues: Use storytelling to share the lived experiences of Indigenous individuals who have been affected by discrimination, residential schools, and systemic inequalities. Personal stories can be powerful in shifting deeply held biases.
Highlight Success Stories: Counter negative stereotypes by showcasing Indigenous leaders, entrepreneurs, scholars, and artists. Focus on narratives of empowerment and resilience to shift the narrative from victimhood to agency.
4. Leveraging Media and Communication Channels
Social Media Campaign: Use platforms like Instagram, TikTok, and YouTube to reach younger Canadians with engaging content, such as short videos, infographics, and testimonials from Indigenous voices. Use hashtags to drive conversations and challenge stereotypes (#TruthToAction, #BeyondEntitlement).
Traditional Media: Partner with national and local newspapers, radio, and TV to feature in-depth stories, opinion pieces, and interviews with Indigenous leaders. Leverage public service announcements to spread awareness about the TRC’s Calls to Action.
Educational Content: Create a series of interactive online resources and webinars that schools, businesses, and community organizations can use to learn about Indigenous history, culture, and rights.
5. Strategic Partnerships
Collaborate with organizations that have credibility and reach, such as Reconciliation Canada, Indigenous Friendship Centres, and educational institutions.
Engage influential allies (e.g., celebrities, athletes, and social media influencers) who can lend their voices to amplify the message, especially among younger Canadians.
Partner with corporations to sponsor educational campaigns and training on Indigenous history and cultural competency, especially in regions where skepticism is higher.
6. Tactical Focus on Education and Action
Push for Curriculum Changes: Advocate for mandatory Indigenous history and cultural education in schools. Work with school boards to provide materials and training for educators.
Community Dialogues: Host town halls, panel discussions, and community workshops, especially in areas with high skepticism (e.g., Prairie provinces). Facilitate open conversations to address biases and misunderstandings.
Call to Action for Individuals: Encourage Canadians to take specific actions, such as signing petitions for policy changes, supporting Indigenous businesses, or volunteering with local Indigenous organizations.
7. Metrics for Success
Quantitative Metrics: Track changes in public awareness, social media engagement (likes, shares, comments), attendance at events, and signatures on petitions.
Qualitative Metrics: Use surveys and focus groups to gauge shifts in attitudes, particularly regarding perceptions of entitlement, support for Indigenous rights, and willingness to support reconciliation efforts.
Policy Impact: Monitor whether there’s increased pressure on governments to implement the TRC’s Calls to Action, especially those related to education and systemic inequalities.
8. Managing Potential Risks
Backlash and Resistance: Prepare for pushback, particularly from those who believe Indigenous communities receive “special treatment.” Use facts and personal stories to counter misinformation.
Tokenism: Ensure that Indigenous voices are leading the campaign, not just featured. Authentic representation is crucial to building trust and credibility.
Sensitivity to Trauma: Recognize the impact of discussing topics like residential schools on Indigenous communities. Offer resources for emotional support and mental health.
Compare that to Claude, which went more into tactics, using audience archetypes as its guiding point of reference:
Based on the research findings and the complex dynamics at play, here's how I would approach a public campaign:
KEY STRATEGIC CONSIDERATIONS:
Audience Segmentation and Targeting
Leverage the five identified worldview segments to create targeted messaging
Primary focus on "Young Idealists" (23%) and "Connected Advocates" (18%) as champions and amplifiers
Secondary focus on "Disconnected Skeptics" (21%) as they're disengaged but not actively resistant
Develop specific strategies for "Informed Critics" (23%) who have knowledge but lack empathy
Deprioritize "Dismissive Naysayers" (14%) initially as they require significant resources for limited impact
Regional Approach
Develop region-specific strategies recognizing the stark differences between Prairie provinces and rest of Canada
Build on existing support in urban centers while addressing specific challenges in rural areas
Create targeted approaches for areas with higher Indigenous populations where tensions may be greater
Message Framework: Leading Messages:
Focus on concrete issues that have highest support (clean water, education funding)
Connect historical understanding to present-day conditions
Emphasize individual Canadian's role in reconciliation (84% support)
Highlight economic costs of inaction and benefits of addressing inequities
Supporting Messages:
Challenge the "special rights" narrative by explaining treaty relationships
Address misconceptions about governance and funding
Showcase successful Indigenous-led initiatives
Campaign Principles:
a) Indigenous Leadership
Ensure campaign is Indigenous-led and shaped
Compensate Indigenous voices and expertise fairly
Center Indigenous perspectives while not placing burden of education on Indigenous people
b) Action Orientation
Move beyond awareness to concrete actions
Provide clear paths for engagement
Connect individual actions to systemic change
c) Long-term Focus
Design for sustained engagement rather than short-term wins
Build infrastructure for ongoing dialogue and learning
Create measurement frameworks that track meaningful progress
TACTICAL CONSIDERATIONS:
Educational Component
Develop curriculum materials for mandatory education initiatives (87% support)
Create accessible learning resources for different age groups
Focus on historical context and contemporary implications
Media Strategy
Leverage strong existing media interest in Indigenous issues
Counter negative stereotypes with success stories
Use data visualization to make complex issues accessible
Employ social media for younger audiences while maintaining traditional media for older demographics
Partnership Approach
Build coalitions with educational institutions
Engage business community on economic benefits
Partner with existing reconciliation organizations
Work with religious institutions given their historical role
Measurement Framework
Track changes in public opinion across key metrics
Measure concrete actions taken, not just awareness
Monitor regional variations in response
Assess impact on policy changes
RISK CONSIDERATIONS:
Potential Backlash
Prepare for opposition from resistant groups
Develop crisis communication protocols
Train spokespeople in addressing common criticisms
Message Hijacking
Monitor and quickly address misinformation
Maintain consistent messaging across partners
Have rapid response protocols in place
Superficial Engagement
Guard against "reconciliation lite" that doesn't address systemic issues
Ensure economic messages don't overshadow rights and justice
Balance accessibility with depth of understanding
BUDGET PRIORITIES:
Indigenous Leadership & Expertise
Fair compensation for Indigenous voices and knowledge
Support for Indigenous-led organizations
Training and capacity building
Content Development
Professional multimedia content creation
Educational materials development
Translation services for multiple languages
Digital Infrastructure
Robust website and social media presence
Data collection and analysis tools
Digital engagement platforms
Regional Implementation
Local coordinators in key regions
Community engagement events
Regional media buys
The campaign should be designed for minimum 3-5 years with clear phase gates and evaluation points, recognizing that meaningful change requires sustained effort and investment.
Final Thoughts
I’m not prepared to declare a winner or loser in this comparison. I’ve only used one document as my baseline, and I genuinely thought both models provided me with useful analysis and critical thinking.
Which is why I’m increasingly of the view that it’s worth using multiple models in your daily work. By using Claude and ChatGPT in this example, I feel I’ve got more to think about, and a wider set of perspectives to consider.
And that’s the real power AI offers: the ability to serve as a thought partner.