Claude Is Not ChatGPT. Here's Your Set Up Guide.
Ten things that will transform your switch from day one.
I spend at least eight hours a day using AI in my workflow. Writing marketing strategies, crunching data, creating campaigns, building AI agents, updating websites, building Business and Brand DNA files, then installing AI marketing systems for clients.
I use all the main models including ChatGPT, Claude, Gemini, Perplexity, plus image and video generation models. On rough calculation I’ve clocked over 10,000 hours using these tools, so it’s fair to say I understand the nauances of each models pretty well.
Over the past week, it made big news when the US Pentagon had a public spat with Anthropic (Claude) when a red line was drawn in the sand of AI ethics.
This saw a mass migration of people switching from Chat GPT to Claude in support of Anthropic standing their ground.
The Claude Migration
With everyone jumping ship, content quickly filled my feed with “Use these prompts in Claude” and “How to export your ChatGPT history”. But before you go using these prompts and migrating your ChatGPT memory, I need you to understand how Claude operates differently to ChatGPT.
In this post I am going to explain why I wouldn’t waste my time using these prompts or importing my chat history and what to do instead to set up Claude to get the best out of making the switch.
You see, Claude is a strategic reasoner, not a fast executor. It thinks differently to ChatGPT. And if you arrive there with your ChatGPT habits intact, you will get ChatGPT-quality results from a model that’s capable of significantly more.
Before I get into what actually works, let’s cover why half the internet just jumped ship in the first place.
Why everyone is suddenly switching
I asked Claude to explain what really went down with the US government as the situation is more nuanced than “Claude stood up to the Pentagon.”
“Anthropic published their model spec and usage policies with explicit restrictions on autonomous lethal decision-making and mass surveillance without human oversight. This was a principled policy stance baked into Claude’s constitution, not a dramatic public confrontation. It gained attention because it was one of the few explicit refusals in the industry to enable those use cases without oversight, contrasting sharply with competitors pursuing defence contracts with fewer guardrails.
No use of Claude in fully autonomous weapons where AI makes targeting decisions without a human in the loop. And no mass surveillance of American citizens using AI to process vast public datasets and build detailed dossiers on people’s political views, private associations, and browsing histories.
The Pentagon said remove those restrictions or face the consequences. Anthropic said no.
The Pentagon designated Anthropic a “supply chain risk.” A label historically reserved for foreign adversaries like Huawei. The first time in US history it has been applied to an American company. Anthropic sued the Trump administration. Claude is still being used by the US military in the war on Iran while the lawsuit plays out.
Hours after the Pentagon announcement, OpenAI CEO Sam Altman posted on X that his company had agreed to Pentagon terms. He later admitted the timing was “sloppy.”
ChatGPT uninstalls then spiked 295% in a single day. One-star App Store reviews jumped 775% that Saturday. Claude shot to number one in the US App Store.
The general public voted with their feet.
10 Power Moves to Get the Best Out of Claude
If AI models were dog breeds, ChatGPT is the Golden Retriever. Friendly, eager, tail wagging. It fetches whatever you throw and acts like every idea you have is brilliant.
Claude is the Standard Poodle. Articulate. Deeply analytical. Built for reasoning, not retrieval.
So when you arrive in Claude and throw the same stick you threw at ChatGPT, the Poodle just stares at you.
Really?
Claude is not a faster, fancier version of the same tool. It was never built to be. It thinks differently, it works differently, and it performs at a completely different level when you set it up properly.
Here are the ten moves that make the difference.
1. Build your Project before you do anything else.
Claude without a Project is like hiring a brilliant strategist and giving them zero briefing. They are capable of extraordinary work. But they are starting from scratch every single time you walk in the room.
A Project is Claude’s permanent briefing room. It sits above every conversation inside it. Load it once with the right context and every single chat that follows starts with Claude already knowing your business, your audience, your offers, your voice, your goals, what you are working toward and why.
No re-explaining yourself. No re-establishing your tone. No pasting in three paragraphs of background before you can ask a real question.
Without a Project you are not using Claude. You are just prompting into a void and hoping something useful comes back.
Here is what to load inside it:
Your Brand DNA -- voice, tone, writing rules, what you never say
Your Business DNA -- offers, audience, positioning, goals
Any other context that is permanently true about your business
The reason most people get generic outputs from Claude is not because Claude is generic. It is because they gave it nothing to work from and expected something specific in return.
Set up your Project first. Then give it a personality with my ‘Business Bestie’ custom instruction. (I talk more about it here in this Substack Post)
The free ‘Business Bestie’ is the starting point, it installs a strategic personality into your Project so Claude stops responding like a polite assistant and starts responding like a thinking partner. Grab it here.
When you are ready to go deeper, the full CEO version of Business Bestie layers in your complete Business and Brand DNA, every conversation then runs through the full context of your business, your audience, your offers, and your voice. That is where the outputs stop feeling like AI and start feeling like you.
2. Context-rich prompts outperform short prompts every time.
ChatGPT trained most of us to keep prompts short and iterate fast. That habit actively works against you in Claude.
Claude is built to handle nuance and complexity in a single prompt. The more context you provide upfront, the better the first output. It’s not about writing longer prompts. It is about replacing vague information with details that provides better outputs.
The old prompting habit was role-play. “You are a world-class copywriter who has written for Fortune 500 companies with 10 millions followers...” Claude does not need the costume. It needs the brief.
Basic prompt: “Write me a LinkedIn post about AI.”
Context-rich prompt: “Write me a LinkedIn post about why most business owners are getting generic AI outputs. My audience are small business owners who use ChatGPT but feel like it does not really get them. I want them to feel seen, not educated. Warm but direct tone. No buzzwords. End with a question that gets comments.”
Context is a one-time brief. Role-play is a costume. One compounds. One does not.
3. Claude will push back and that is the feature, not the flaw
ChatGPT is designed to be agreeable. It will complete the task, give you a confident answer (right or wrong), and move on. Claude is designed to be honest. It will tell you if your brief is unclear. It will ask a clarifying question before running in the wrong direction. It will flag if something in your reasoning doesn’t hold up.
The best outputs I’ve ever pulled from Claude came after it stopped me and asked what I actually meant.
When Claude pushes back or asks a question, pause before you override it. That question is usually pointing at something worth addressing.
4. Tell Claude what you’re trying to achieve before you ask it to do anything
This is the single biggest shift in workflow for people coming from ChatGPT.
With ChatGPT the default move is to jump straight to the task. “Write me a LinkedIn post about X.” “Draft an email sequence for Y.” Claude will do this too, but you are leaving significant quality on the table.
The better move is to lead with your goal. “I’m trying to position myself as the go-to person for X audience. Here’s what I know about them. Here’s what I want them to feel after reading this. What’s the best approach before we write anything?” Claude will map out a strategy, check its understanding, and then execute from a much more informed position.
That extra sixty seconds of framing changes the output quality more than any prompt template you’ll find online.
State your goal first. Give context second. Ask for the approach third. Then execute. In that order.
5. Memory is intentional here, and that’s actually better
ChatGPT builds rolling memory across conversations automatically. Claude’s memory works through what you deliberately load into a Project.
This means Claude will not quietly accumulate assumptions about you across random chats. It will not remember something you mentioned three weeks ago in a different conversation unless you’ve put it inside a Project. For people used to ChatGPT’s passive memory this feels like a limitation at first.
It isn’t. Intentional memory is more powerful than passive memory because you control exactly what Claude knows and why. There are no accumulated misunderstandings. No half-remembered preferences from a conversation you forgot you had. You build the context deliberately and Claude works from that foundation consistently.
Treat every fresh chat outside a Project as a clean slate. Treat your Project as your permanent briefing room. Keep them separate and use each intentionally.
6. Ask Claude to think before it answers.
Claude will reason through a problem before it responds if you ask it to. This is not a novelty. It is one of the most useful things you can do when you are working through a strategic decision, a positioning question, or an offer structure you are not quite sure about yet.
Try this: “Think through this before you answer. What are the assumptions embedded in what I am asking?”
You will catch gaps in your own thinking that a straight answer would have buried.
This is also the point where your model choice matters.
Claude comes in two main versions, Sonnet and Opus. Sonnet is faster and handles most everyday tasks beautifully. Opus is slower, more expensive, and significantly more powerful for complex reasoning. Think of Sonnet as your day-to-day thinking partner and Opus as the one you bring in when the decision actually matters.
If you are using the free version of Claude, you are on Sonnet. If you are on a paid plan, you can switch to Opus inside the same interface.
For strategic thinking, offer development, positioning decisions, or anything where you need Claude to really pressure-test an idea rather than just respond to it, switch to Opus first. The difference in depth is noticeable immediately.
7. Give it your actual words, not a polished brief.
Claude does not need you to have it together before you open a conversation. Paste in a voice note transcript. Drop in a messy brain dump. Share the screenshot you grabbed at midnight because the idea would not wait.
Claude is exceptionally good at finding the signal in your noise and reflecting it back in a form you can use. The messier the input, often the more useful the output -- because it is working with how you actually think, not a sanitised version of it.
I have the app on my phone and will often lie in bed brain dumping ideas by voice or sharing screenshots to analyse. Claude pulls the thread, structures the thinking, and I wake up with something actionable instead of a note I will never look at again.
8. Connect your tools.
Those late night brain dumps and car voice chats are only useful if they go somewhere.
Once you have captured the idea in Claude, you do not have to remember to action it later. Ask Claude to add it directly to your Google Calendar as a follow up reminder, with complete detailed notes. Done. It will not get lost in a note you never open again.
Claude integrates with Canva, Google Calendar, Gmail, and more. Once connected, you are not just thinking inside Claude. You are actioning directly from it. Draft a campaign, build the visuals in Canva, schedule the follow up in your calendar, all without leaving the conversation.
The Connectors Directory includes over 50 integrations including Canva, Slack, Notion, Asana and more.
This is where Claude stops being a thinking tool and starts being an operating system for your business.
Navigate to the "Search and tools" menu in Claude chat, then hit "Add connectors." Select the tool you want to connect and click connect. For web connectors like Google Calendar and Gmail, sign in to your account to grant Claude access. Then test it by asking Claude if it has access to that tool
9. Ask Claude to visualise it.
This one just dropped and it is a game changer for visual thinkers.
Claude can now create custom charts, diagrams, and other visualisations inline in its responses and then tweak and modify them as the conversation develops. These are not static images. You can interact with them click buttons, adjust sliders, expand to full screen and then ask Claude to refine further or add more detail.
Claude will decide when to build a visual, or you can ask directly with prompts like “draw this as a diagram” or “visualise how this might change over time.”
Think content strategy mapped as a visual flowchart. A campaign timeline that updates as you talk. A competitive landscape diagram built in real time.
The feature is available now for all plan types including free, currently in beta. Note it is desktop and web only for now mobile is coming.
10. Analysis powerhouse.
Claude is your marketing right hand, it will do a full marketing makeover before lunch.
Upload your CRM data, Claude will profile your best customers. Not just demographics. Patterns. Behaviours. The thing your spreadsheet never told you.
Analyse your inventory and sales results and find the anomaly. The product that should be converting but isn’t. The segment you’ve been ignoring that’s actually your warmest audience. Claude can then builds an interactive quiz with scoring logic and personalised outcomes to gather the missing intel directly from your audience. No developer. No coding. No waiting.
While you’re making your second cup of coffee, it’s crunching the responses, finds the pattern, provides the pitch, writes the copy, designs the landing page, and plans the campaign launch sequence.
All before lunch.
It’s a Standard Poodle with a Harvard MBA, a nose for uncovering complex problems, and the reasoning insight to provide the solution.
And just when you think it’s done showing off, it spins around, barks twice, and asks if you want a visual presentation to go with that.
Should you export your ChatGPT memory and import it to Claude?
It seems every AI influencer on Instagram right now is telling you to do this. But here is my logical reason for pausing before you do.
If your ChatGPT memory is clean, deliberate, and well maintained, and it accurately reflects how you work and think, then yes. It gives you a running start.
But if you find yourself constantly swearing at it because it gives you incorrect answers or made things up, or it’s constantly giving information about you that is no longer true, then no. Do not drag that mess into a fresh tool. Claude is smart and it will not take long to get to know you, the real current you.
Here is what I would do instead.
If you are mid-project in ChatGPT and you need continuity, do not export everything. Ask ChatGPT to write a concise summary of that specific conversation only. The context, the decisions made, and where you got to.
Then once you’ve set up your Project, install the Business Bestie prompt, paste that into there and ask Claude to analyse the information and ask any clarifying questions about the job. Clean, relevant, and actually useful.
That is not starting from zero. That is starting from intention.
And that is the whole point of making the switch.
The summary
The people getting the best results from Claude right now are not the people with the cleverest prompts. They are the people who set up their foundations properly, brief it like a senior team member, and let it think before it executes.
It doesn’t just answer. It thinks with you.
You can throw it your messiest, most complex, half-baked business problem and it will sit in that complexity with you, pressure test it, find the angle you missed, and come out the other side with something you couldn’t have reached alone.
It’s not a search engine that retrieves. It’s not a template machine that fills in blanks. It is the only AI tool that genuinely reasons through hard problems and gets more useful the harder the problem gets.
What I have covered today is for Claude Chat, this doesn’t even touch the system automation genius Claude Cowork and Claude Code can do.
Let me know are you using or thinking of using Claude?
Want to set this up properly from day one?
The single biggest mistake I see switchers make is going straight to prompting before building the foundation. Claude’s outputs are only as good as the context you give it. That means your Business DNA and Brand DNA need to exist as actual files before you start using it seriously for marketing.
I’m opening a small number of Business and Brand DNA sessions for people who want to set this up right. We build your foundation files together, install your CEO Business Bestie, and you leave with Claude actually knowing your business, your audience, your offers, and your voice.
Spots are limited. If you’re interested, send me a message or email connect@debszabo.com and I’ll send you the details.
Deb Szabo is an AI Marketing Strategist based in Newcastle, Australia. She helps business owners build the marketing foundations that make AI actually work.
thanks for this fabulous summary and insight Deb. This has cleared things up for me and prompted me to be intentional in the way i set it up rather than just drag everything across. I can't wait to work with you as you help me navigate this.
Amazing, thanks for your insight Deb, you’ve answered many of the questions I had around whether it was worth migrating or not