Free Training: Build Custom GPTs for Marketing | theCLICK

What you’ll be able to do

Create custom GPTs that write, plan, organize, etc—so you can extend your strategic thinking and chew through your daily to-do list
Fix the two big AI problems: cold start / “new guy” (re-explaining context every session) and boiled chicken (reasonable-but-generic output)
Use the core idea of grounding (trusted sources of truth) so outputs are verifiable and usable
Build two high-ROI GPT patterns: a planning GPT (Content Calendar) and a writing/editor GPT (Draper-style punchy persuasion)
Use Markdown to preserve structure (headings, tables) so the AI reliably follows long instructions
Work inside Projects and pull in GPT tools via @ so context compounds (offer + persona + plan) and execution becomes consistent

The 4 ChatGPT tools to master (in order)

Custom GPTs: reusable tools that run repeatable workflows
ChatGPT Projects: the workspace where work stays organized and grounded
Deep Research: slower, source-backed research you can verify
Canvas: living documents you build, edit, and ship from

Fastest path: build GPT tools, run them inside Projects, and use Deep Research + Canvas as your source-of-truth engine.

Key terms (quick definitions)

Growth OS: a proven system that helps founders and marketers plan → build → ship business growth campaigns with astonishing speed and quality
GPT (Custom GPT): a purpose-built AI tool with instructions (and optional files) designed to do one job repeatedly and well
Instructions: the operating rules your GPT follows every time you run it (your “built-in prompt”)
Knowledge base: files you attach to a GPT (or Project) so it can reference your source-of-truth materials (including Markdown files)
Grounding: anchoring output to trusted sources (instructions + files + vetted inputs + URLs) to reduce drift and hallucinations
Cold start / “new guy”: the AI is capable but needs onboarding; without structure you re-explain everything every session
Boiled chicken: output is reasonable but generic because it isn’t grounded in your business reality
Markdown: a formatting language (e.g., # headings, **bold**, tables) that helps AI understand document structure
@ menu: the in-chat picker where you can call GPT tools (and other connected tools) into a conversation
ChatGPT Projects: an organized workspace that groups chats plus shared instructions and shared files so your work stays grounded and compounds over time
Thinking vs fast model: use a thinking model for plans/calendars/strategy; use a fast model for quick drafts and rewrites

The 3 traps that keep GPTs from helping

Tinkering overload: tool-hopping creates half-finished work and no compounding advantage
Cold start / “new guy”: the AI never gets onboarded, so you start over every session
Boiled chicken: output is “reasonable” but generic because it isn’t grounded

Rule of thumb: build a GPT with clear instructions, show it what “good” looks like, and ground it with the minimum files/URLs needed for the job.

Grounding sources (pick a “source of truth”)

Instructions: the built-in workflow your GPT follows
Knowledge files: persona, offer, plan, examples, swipe files (including Markdown)
Web pages: ground to a URL when you’re working from a landing page or offer page
Deep Research: source-backed research outputs (quotes + URLs) you can verify
Canvas docs: living documents (drafts, pages, decks) that keep work coherent

Shortcut: when you get a strong output, highlight the exact part you want and use “Ask ChatGPT” to keep the next step anchored to that slice.

GPTs vs Projects (how to structure your workflow)

GPTs (Custom GPTs)	ChatGPT Projects
Tool you pick up to do a specific job (plan, write, summarize, generate hooks)	Workspace / container for a big initiative (e.g., Marketing, Podcast, Product Launch)
Has instructions + optional knowledge files	Has shared instructions + shared files across all threads in the Project
Best for repeatable, narrow tasks (your “toolbelt”)	Best for compounding context and organizing work over time (your “intelligent folder”)
Use via the @ menu inside any chat	Do the work inside the Project so it stays grounded in offer + persona + plan

Best practice: plan inside a Project, then call GPT tools (Content Calendar, Draper) at the moment you need them.

Build your first Custom GPT (setup checklist)

Start with a meta prompt to generate draft GPT instructions (purpose → required context → process → “what good looks like”).
Paste the draft into Google Docs so you can refine structure, headings, and tables.
Enable Markdown (Tools → Preferences → enable Markdown), then Copy as Markdown.
Open ChatGPT → Explore GPTs → Create and use the Configure screen (not the “chat-to-build” screen).
Name + describe the GPT, then paste the Markdown instructions into the Instructions box.
Attach files to the GPT’s knowledge base if needed (persona, offer, plan, examples, formats). Markdown files are allowed.
Choose a model: use a thinking model for planning/strategy; use a fast model for quick drafts.
Run it in workflow: in any chat, hit @, select the GPT, run it, then dismiss it when done.
Optional (high leverage): work inside a Project so offer + persona + plan context compounds.

Minimum viable GPT instructions: Purpose + Required Context + Process + One example of “what good looks like.”

FAQ

What is a Custom GPT?

A Custom GPT is a purpose-built tool with instructions (and optional files) that runs a repeatable workflow—so you don’t have to re-prompt or reinvent the process every time.

Do I need a paid ChatGPT plan to create GPTs?

Typically yes—creating Custom GPTs is a paid feature. You can still use Projects on some free plans, but GPT creation generally requires a subscription.

What should I include in GPT instructions?

Start with Purpose, then Required Context, then the Process. Finish by showing an example of exactly what “good output” looks like.

Why does my GPT still feel generic?

Usually it’s a grounding problem: the workflow isn’t specific, “what good looks like” is missing, or you didn’t attach the minimum truth files/URLs. Tighten instructions, add a concrete example output, and ground the GPT in a persona/offer/plan when needed.

What files should I add to a GPT’s knowledge base?

Add files that define the truth: your offer, your persona, your plan, and any examples you want it to mimic (tables, formats, best-performing copy). Markdown versions of these files work extremely well.

When should I use a thinking model vs a fast model?

Use a thinking model for planning, strategy, and calendars. Use a fast model for quick drafts, rewrites, or lightweight tasks where speed matters.

How do I use a GPT inside my normal workflow?

In any chat (or inside a Project), hit @ and select the GPT. Run it like a tool, then dismiss it when you’re done. For best results, highlight the specific slice you want to transform and use “Ask ChatGPT” so the next step stays anchored.

How do Projects make Custom GPTs work better?

Projects can hold shared files and instructions (offer + persona + plan). When you call a GPT inside a Project, it can operate with richer context and produce better results faster.

Can I use Custom GPTs to create deliverables like decks or email promos?

Yes—ground the GPT in a source of truth (a URL, a Canvas doc, or your offer/persona files) and then ask for a specific deliverable format (e.g., slide-by-slide deck outline, 3–5 email storyboard, landing page skeleton).

Do Custom GPTs replace Deep Research or Canvas?

No—think of GPTs as repeatable workflows. Deep Research and Canvas are where your truth and drafts live. Best flow: research in Deep Research → draft/build in Canvas → use GPT tools via @ to plan, write, and polish.

Build Custom GPTs for Marketing