Lesson 07 · ChatGPT Mastery Pro ~11 min read 3 workflows

Multimodal ChatGPT: the Swiss army knife.

ChatGPT isn't just a text AI anymore. It searches the web, analyzes images you upload, and generates images via DALL-E. Most users still treat it like text-only. This lesson covers the three multimodal workflows where the integration genuinely changes the work.

The mental model

Multimodal isn't a feature — it's a capability you forget exists.

ChatGPT can now see what you see, read what you point at, and generate images for what you describe. Most users default to typing as if it's still 2023. The workflows below are the ones where breaking out of text saves real time.

Workflow 01 Image input: photo, screenshot, document

1

Upload an image, ask anything

ChatGPT can read text in images, describe what's in them, extract structured data, and answer questions about them.

The prompt that works

Image use cases(Upload screenshot of error message) What's causing this and how do I fix it? (Upload photo of a whiteboard) Transcribe everything written here. (Upload an Excel chart) What's the headline finding and what's misleading?

Best use cases

  • Error message debugging
  • Whiteboard photo transcription
  • Receipt and invoice processing
  • Reading documents you don't want to retype
  • Analyzing charts or data visualizations
Image processing is generally accurate but can misread handwriting, low-quality photos, and dense charts. Verify the transcription against the original.
Time savings: Manual transcription/data entry: gone for many use cases.

Workflow 02 Web search: stop guessing, start finding

2

Trigger search when freshness matters

ChatGPT searches the web automatically for time-sensitive questions, or you can force it.

The prompt that works

Search-forcing promptsSearch the web for the current [stat/fact/news] on [topic]. With current information from the web, answer: [question that needs fresh data].

Best use cases

  • Current events and breaking news
  • Recent product updates and prices
  • Verifying time-sensitive claims
  • Research that needs sources
Web search adds latency and cost. For evergreen questions (concepts, definitions, fundamentals), don't force search — it's slower with no benefit.
Time savings: Out-of-date answers: avoided when it matters.

Workflow 03 DALL-E for work, not just art

3

Generate visual assets for real work

DALL-E can produce diagrams, mockups, illustrations, social images. Not perfect, but often good enough for internal use or starting points.

The prompt that works

Work promptsGenerate a simple flowchart showing 4 stages: input, processing, review, output. Clean, minimal, blue accent. Create a social media graphic: 'AI Mastery for Knowledge Workers' as the headline, modern dark theme, abstract geometric background. Generate a hero image for a blog post about productivity — calm, focused, person at a desk.

Best use cases

  • Internal slide deck visuals
  • Blog post hero images
  • Social media graphics
  • Concept diagrams and flowcharts
  • Mock UI elements
DALL-E is bad at text in images (tends to misspell), specific facial features (don't use for real people), and exact brand colors (give it hex codes but expect approximation).
Time savings: Hours hunting stock images: minutes generating fit-for-purpose visuals.

Final challenge: one multimodal day

For one workday, deliberately use a non-text capability every time the opportunity arises. Photo of a whiteboard → upload it. Need a current stat → force web search. Need a visual for a deck → generate it. Count how often these capabilities helped vs. didn't.

Back to ChatGPT Mastery → See pricing →

What you can do now

  • Upload images and ask ChatGPT to read, transcribe, or analyze them
  • Force web search when freshness matters
  • Generate work-appropriate visuals with DALL-E
  • Recognize the limits of each multimodal feature
  • Stop defaulting to text-only when other inputs would be faster
Pro+
Up next in ChatGPT Mastery

Lesson 8+ · Plus: API basics, projects, advanced prompting, and more

When to use the API instead of the app, project organization for power users, evaluation patterns, advanced reasoning prompts. The advanced curriculum. See pricing →