Review Diagrams and Photos in ChatGPT

The same image can mean very different things depending on the domain. A messy sketch on a whiteboard could be a system diagram, a floor plan, or a shopping list drawn by a tired parent. ChatGPT does not know which one until you say. That one line of context can change the entire review.

When you tell ChatGPT what kind of image it is looking at, you unlock domain thinking. A system diagram gets checked for missing arrows. A receipt gets checked for math. A UI mock gets checked for accessibility. Skip the context and you get a shallow reply that could apply to anything.

What a good context line looks like

A context line names three things. The subject, the source, and the purpose. For example, this is a whiteboard photo from a planning meeting for a payments backend. That single sentence tells ChatGPT the domain, the format, and the reason you care.

When domain context matters most

• Whiteboard photos of architecture or process flows
• Photos of a receipt, invoice, or bank statement
• UI mockups from a design tool
• Photos of a physical space such as a room or a shelf
• Diagrams from a textbook, paper, or slide deck

Prompt for a whiteboard system diagram

Photo: a whiteboard from a backend planning meeting.
Domain: payments processing at an online store.
Task: review the diagram for missing or unclear parts.
Return a numbered list of gaps. Each item names the missing piece and why it matters.
Do not redesign the system. Only note what is missing from what is drawn.

Why this prompt works

The domain line changes the review from generic to specific. ChatGPT will now think about retries, refunds, and idempotency instead of shapes and arrows in the abstract. The rule against redesign keeps the reply focused on the picture you actually sent.

Prompt for a receipt or invoice photo

Photo: a restaurant receipt.
Domain: business meal for expense reporting.
Task: extract each line item and check the math.
Return a table: item, amount, notes.
Add a subtotal, tax, and total row. Flag any charge you cannot read clearly.

Prompt for a UI mockup

Screenshot: a mobile app checkout screen mockup.
Domain: fashion ecommerce for first time buyers.
Task: review for clarity and trust signals.
Return a numbered list of issues. For each, note the visible element and one concrete change.
Skip visual polish, focus on comprehension.

Prompt for a real world photo

Photo: a shelf inside a small grocery store.
Domain: store layout review by a category manager.
Task: describe the layout and call out anything unusual such as gaps, misplaced items, or unclear pricing.
Return two short paragraphs: layout, then issues.
Skip lighting and camera quality.

What to include in your message

• One line naming the subject, source, and purpose
• The domain so ChatGPT can apply domain thinking
• The task and the format for the reply
• A rule about what to skip so the review stays useful
• A clear way to flag uncertainty

How to refine when the review misses the point

If the reply reads like generic advice, the domain line probably was not specific enough. Rename the domain more tightly. For example, replace ecommerce with checkout for first time mobile buyers. A narrower domain almost always produces sharper feedback.

Common mistakes

• Uploading the image with no domain context at all
• Using a domain word so broad it does not narrow the review
• Asking for polish when you actually need a gap analysis
• Trusting a receipt total without checking the math yourself
• Letting ChatGPT redesign the image when you only asked for a review

How to check the result

Pick the two most important claims from the reply and confirm them against the image. For a receipt, confirm one number. For a diagram, confirm one missing arrow. For a UI, confirm one element the reply named. If those hold, the rest is worth reading.

Takeaway

Context turns a shallow reaction into a real review. Name the subject, name the domain, and ChatGPT will meet you where the work actually lives.