AutoDraw (autodraw.com) — Google’s sketch-to-clip-art drawing tool

Sketch a rough shape and AutoDraw guesses what you meant, swapping your doodle for clip art from an artist-drawn library. Free, no sign-in.

  • Sketch Recognition
  • Clip Art
  • Drawing Tools
  • Google AI Experiments
Publisher
Google Creative Lab
Type
Sketch to Clip Art
Pricing
Free
Reviewed
19 September 2026
Official site

Quick verdict

Use when

  • You need a clean icon and cannot draw
  • The subject is a common everyday object
  • You want something usable in under a minute
  • A rough sketch is all you have to work from

Skip when

  • You need illustration beyond the library
  • You want layered or vector output
  • The drawing has to be filed and revisited
  • You need the tool guaranteed to outlive you

Try instead

Same job, different trade-offs — getting a drawing out of a browser.

AutoDraw vs Excalidraw vs Canva

AutoDraw answers one narrow question — what was this scribble meant to be — and then hands you a finished piece of art from a fixed set. Excalidraw draws whatever you draw, in a house style built for diagrams rather than illustrations, and Canva skips the sketch entirely by letting you search a huge library of ready elements, so the choice is really about whether your rough shape is the input or the thing you want to escape.

Tap a dimension to focus

Pricing

AutoDraw leads
  • AutoDrawThis page
    • Free with no account, no paid tier, and no upgrade path
    • There is no store, no credits, and no subscription to cancel
    • Google has never charged for it or gated a feature
    • Free and open source; the hosted app costs nothing to use
    • Paid plans add team sharing, libraries, and longer history
    • You can self-host it because the source is public
    • Free tier with a large set of free elements and templates
    • Paid tiers unlock premium assets, brand kits, and storage
    • Team plans are billed per seat
  1. AutoDraw — official tool
  2. Experiments with Google — AutoDraw by Google Creative Lab
  3. Google — the 2017 launch post naming the contributing artists
  4. AutoDraw — the Privacy & Terms link it points to (Google policies)
  5. Excalidraw — plans for the paid hosted tier
  6. Canva — plan tiers

Preview of AutoDraw - not the live app. Confirm details on the official site.

Does this overview help you decide?

(—)

Learn more

Details below the decision summary—features, workflow, and scope notes.

What is AutoDraw?

Sketch a rough shape and AutoDraw guesses what you meant, swapping your doodle for clip art from an artist-drawn library. Free, no sign-in.

What it costs

Nothing — and there is no pricing page to check, which is itself the finding.
Free tier
Yes
Pricing summary
AutoDraw is free in the plain sense of the word: no account, no trial, no credit system, no watermark, and no paid tier sitting behind a feature you eventually need. Google has never charged for it and the project page states plainly that there is nothing to download and nothing to pay for. That is not a freemium funnel with a thin free plan — there is no upgrade path at all, so there is nothing to evaluate on price and nothing to cancel later. The practical cost is different and worth naming: the tool has been left as it was in 2017, it still offers a share button for a Google product that shut down years ago, and Google retires its experiments from time to time. Free here also means unsupported — no roadmap, no support channel, and no commitment that it will be running next year, so anything you make should leave the canvas as a download rather than live only in the tool.

Reviewed on 19 September 2026 · Experiments with Google — AutoDraw project page

What AutoDraw does

Capabilities taken from the running tool and Google’s own project pages.
  • The suggestion row does the work

    Draw a rough shape with the Draw tool and a strip of finished artwork appears along the top of the canvas without you asking for it. Click one and your sketch is replaced by that piece of art, scaled to fit roughly where you drew. The suggestions refresh as you keep drawing, so the row tracks what the tool currently thinks you are making rather than only your first stroke.

  • It matches, it does not generate

    The suggestion tool uses the same recognition technology as Quick, Draw! — another of Google’s AI Experiments — to guess what a sketch represents. Nothing is invented on the spot: every option it offers is a drawing a human made earlier, and the model’s only job is choosing which existing piece fits your ink. That is why the options are always good and occasionally completely unrelated.

  • A library counted in hundreds, drawn by named artists

    The recognition set covers several hundred objects, from a star, a heart and a house to a colosseum, a taj mahal and a venus fly trap. The artwork behind it is a little over 1,600 individual drawings contributed by eight credited makers — chiefly Selman Design of New York, alongside HAWRAF, Simone Noronha, Tori Hinn, Pei Liew, Erin Butner, Julia Melograna and Melia Tandiono.

  • A small, complete drawing tool underneath

    Select, Draw, Fill, Type, Shape, Undo, Delete, and Zoom sit in the toolbar, with a swatch palette for colour and a stroke-width control. It is not a stripped-back canvas with only the AI trick — you can build a whole simple composition with fills, text labels, and geometric shapes, then move, resize, and rotate the pieces.

  • Download as PNG, share straight out

    Finished work downloads as a PNG, and there is a share menu for posting the canvas to Facebook or X. There is no vector export, no project file, and no cloud save — the drawing is kept in your own browser storage, so clearing site data clears the artwork with it.

  • Nothing to install and nothing to sign up for

    It runs in the browser on a phone, tablet, laptop, or desktop, and it works with a mouse, a trackpad, or a finger. There is no account, no email, no onboarding wall, and no app to download — which is a large part of why a 2017 experiment is still the fastest answer to "I need a small clean icon".

How people use AutoDraw

The loop the tool is built around, from blank canvas to a finished image.
  1. Open it and start drawing

    There is no sign-up step to get past. Press Start Drawing on the intro screen and you land on a blank canvas with the Draw tool already selected — the fastest possible route from arriving to making a mark.

  2. Sketch the shape badly, on purpose

    Draw the object you want as roughly as you like. Precision works against you here: the recogniser is reading the overall silhouette, so a loose approximation of a cat is more useful than a careful outline that spends a minute on the ears.

  3. Pick from the row that appears

    When the suggestions show up, click the one that matches. Your sketch is replaced by the finished drawing. If nothing in the row fits, draw the shape again slightly differently and a fresh set of options is offered — the tool is re-guessing rather than accumulating.

  4. Build the rest of the picture by hand

    Use Shape for rectangles, circles, and lines, Type to add labels, and Fill to colour areas. Pieces can be selected, moved, resized, and rotated, and Undo steps back through recent changes, so a simple diagram or card comes together without leaving the canvas.

  5. Download before you close the tab

    Export the finished image as a PNG. Worth doing deliberately: work in progress is held in browser storage rather than an account, so it does not follow you to another device and it does not survive clearing site data.

Who AutoDraw is for

The people whose actual task it removes.
  • Anyone who needs an icon and cannot draw one

    The core audience the tool was built for. If your ceiling is a stick figure and your requirement is a presentable lightbulb, rocket, or birthday cake, the suggestion row closes that gap in one click and saves you searching a stock library.

  • Teachers building worksheets and slides

    Recognisable everyday objects map neatly onto classroom material — animals, food, weather, transport, shapes — and the whole thing runs on a school browser with no login, no install, and no licence to administer.

  • Product managers and founders sketching a visual

    For a quick illustration in a deck, a doc, or a landing-page placeholder, drawing the shape yourself and letting the tool clean it up is faster than briefing a designer and cheaper than buying an asset for something temporary.

  • Designers blocking out an idea before the real work

    A rough icon at the right scale, made in under a minute, is useful for testing a layout before anyone invests in the finished illustration. Treat the PNG as a placeholder and the library’s limits stop mattering.

When AutoDraw is the right pick

AutoDraw earns its place when the gap between what you can draw and what you need to show is small but real — a birthday card, a worksheet, a slide that needs an icon, a quick visual where a hand-drawn arrow will not do. In those moments it is close to magic: you scribble something approximate, the suggestion row appears without being asked, and one click replaces your effort with art a professional made. Nothing else does that specific trick, and it does it without an account, an install, or a payment. It stops being the right pick the moment your subject is not an everyday object. The library is fixed and finite, so a niche idea will produce a row of confidently wrong suggestions, and you will end up fighting the tool rather than using it. It is also the wrong home for anything you intend to keep: there is no vector export, no layer file, and no account to save into, so if a project will be reopened and revised, draw it somewhere that gives you an editable file. Treat AutoDraw as a fast first move that ends in a PNG, and it will rarely disappoint.

Platform and licensing notes

What the tool technically is, and what it is not.
Recognition is a service call, not local processing
The suggestion tool works by sending your ink to a Google handwriting-recognition endpoint and reading back a ranked list of matches against the artwork library. The canvas is in your browser; the guessing is not. No account is created, but the strokes are not staying on your machine either.
The library is closed and fixed
Around five hundred recognisable objects, backed by a little over 1,600 individual drawings. Nothing is generated and nothing is learned from your session, so the tool will never improve at the specific thing you keep trying to draw — a subject outside the set stays outside it.
This is a 2017 experiment that was never modernised
AutoDraw launched in May 2017 from Google Creative Lab, built by Dan Motzenbecker and Kyle Phillips. It is still live and the recognition still works, but the interface has not moved on: the share menu still lists Google+ under a Google+ icon, years after that product shut down. Read that as a signal about support, not about whether the tool functions.
Browser storage is the only place your work lives
Artwork is kept in the browser under a local storage key rather than on a server account. There is no login to recover from, no cross-device sync, no version history, and no server-side copy to retrieve. A PNG download is the real save button.
Artwork licensing is the artists’, not a standard stock licence
The drawings were contributed by named designers and studios explicitly for public use within the tool, and the project page states that AutoDraw can guess hundreds of drawings with more to be added over time. Before putting a stencil into commercial material, check the terms yourself — there is no per-asset licence page to point at.
What it is not
Not an image generator, not a vector editor, not a diagramming tool, and not a design suite. There are no layers, no SVG export, no collaboration, no API, and no account system. It is one narrow capability — a sketch interpreter wired to a curated art library — plus a minimal canvas around it.

Frequently Asked Questions

Quick answers about this tool—open a question to read more.