Home · Business · Product & Strategy · UX & prototyping

YES

As of 13 August 2026, AI can convert a sketch into a digital prototype.

This still needs a person who signs their name to it.

Can you do it?

15 minutesto a draft.

1 hourto something you’d act on.

Cost, all in£0

Skill neededchat-fluent

Who has to check ita colleague

What the alternative costsRelume creates AI-assisted sitemaps and wireframes that can be exported to Figma and Webflow.

If this goes wrong: the prototype follows an incorrect interpretation of the sketch, which you can usually fix before it is used for a product decision.

What to actually do

  1. Hand it to a person

    The route this page recommends

    A person who owns the outcome does this end to end, worth it when the failure is dear.

  2. Do it yourself

    Second choice

    A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

    How to actually do it

    1. Open the sketch as a clear image and add short annotations for the intended user, the starting screen and the outcome the user should reach.
    2. Gather any existing screen copy, brand colours, content rules and known constraints, and remove confidential information that the prototype does not need.
    3. Paste the image and the copyable prompt into an image-capable chatbot, then answer its clarification questions using only decisions you have actually made.
    4. Use the resulting screen list and interaction map to create or import the prototype in a prototyping tool such as Relume, or run the generated HTML if the tool produced it.
    5. Click through the complete flow from the starting point, checking every button, link, form field, empty state, error state and back action against the sketch and the stated assumptions.
    6. Send the prototype to a colleague who understands the product and ask them to complete the main task without your explanation, then correct any confusing screen or transition and repeat the flow.

    Prompt

    I have attached a sketch for a digital product flow. Convert it into a low-fidelity clickable prototype specification, preserving the intent of the sketch rather than inventing extra features.

First, identify each screen, its purpose, the visible components and the links or actions between screens. Then list any unclear parts of the sketch as explicit assumptions and ask no more than five questions that would materially change the prototype. Do not invent business rules, content, user research findings or accessibility claims.

Create the prototype in the format your environment supports. If you can generate a clickable prototype or runnable HTML, do so. Otherwise provide screen-by-screen wireframe instructions with exact labels, states and interaction links that I can recreate in a prototyping tool. Keep the visual treatment simple and use placeholder copy where the sketch does not provide wording.

Include these flows if they are present in the sketch: the starting point, the main task, empty states, error states, back navigation and the completion state. Mark every part that is inferred rather than visible. Finish with a short test checklist covering each link, button, form field and back-navigation path.

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

  3. Use a tool built for this

    The distant third

What it gets wrong

  • AI cannot recover the product intent that the sketch leaves unstated; it fills gaps with plausible assumptions.
  • AI cannot decide whether the proposed flow is useful, commercially sensible or consistent with your wider product strategy.
  • AI cannot replace observation of people using the prototype, so it can miss confusion that is obvious in a real test.
  • AI can produce screens that look connected while omitting important states, permissions, content rules or technical constraints.

Even on a YES, the friction has a name: judgement under ambiguity, verification cost and context depth.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output2
Inputs2
Verification1
Liability2
Effort delta2
Total9 / 10

The methodology and its thresholds are published in full.

FAQ

Can ChatGPT turn my hand-drawn wireframe into a prototype?
Yes. An image-capable chatbot can interpret the sketch and produce a screen map, interaction specification or runnable front-end, while a dedicated tool such as Relume can turn the structure into editable wireframes. Check every inferred interaction because a sketch rarely records all the states.
Can AI make a clickable prototype from a photo?
Yes, if the photo is clear and you explain the intended flow. AI may misread labels, controls or connections, so compare the clickable result with the photograph and test the full path yourself.
Will AI preserve my original sketch when making a prototype?
It can preserve the visible layout and main flow, but it will not reliably preserve unstated intent. Tell it not to add features, mark its assumptions and approve each change before showing the prototype to your team.
Is an AI prototype good enough to test with users?
It can be good enough for an early usability test if the main flow, labels and states are accurate. It is not evidence that the product works: watch people use it and change the prototype when they misunderstand it.

Nearby answers

All UX & prototyping rulings

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.