Say What You See

Say What You See is a clever little Google game that teaches you one of the most valuable skills of the AI era: how to write a good image prompt. You are shown an AI-generated image, and your job is to describe it well enough, in a short prompt, that the AI generates a close match. It sounds simple, but it quickly reveals how much precision goes into turning words into the exact picture you have in mind. It is part puzzle, part training ground, and a genuinely fun way to get better at prompting.

Created by an artist in residence at the Google Arts & Culture Lab, the game flips the usual text-to-image flow on its head. Instead of imagining a picture and describing it, you reverse-engineer an existing image into the prompt that could have made it, sharpening both your prompt-writing and your ability to truly see and articulate visual detail. As AI image tools become everyday creative instruments, that skill is increasingly worth having.

This guide covers everything that matters about Say What You See in 2026: what it is, how the gameplay works, what it actually teaches, who it is for, why prompt skills matter, and the limitations of a focused experiment. By the end you will know whether it can level up your prompting.

What Is Say What You See?

Say What You See is an experimental Google Arts & Culture game that teaches the art of the prompt by having you describe AI-generated images. You are presented with a target image and a text box, and you write a prompt of up to 120 characters, aiming to make the AI generate an image that closely matches the one you are looking at. If your generated image clears the level's visual similarity threshold, you pass; if not, you refine and try again.

The premise turns image generation into a reverse-engineering challenge. Rather than dreaming up a picture and prompting for it, you study an existing image and figure out the words that would recreate it: its medium, subject, composition, mood, and details. That exercise builds two skills at once: writing effective prompts and reading images carefully enough to describe them precisely.

It is a free experiment created by artist Jack Wild as an artist in residence at the Google Arts & Culture Lab. Like Google's other creative experiments, it exists to make learning about AI playful and hands-on rather than abstract.

How the Game Works

The game is structured in levels of increasing difficulty, easing you in before raising the stakes.

1. Study the target image. An AI-generated picture you need to recreate through words.
2. Write a prompt. Describe it in up to 120 characters, covering medium, subject, and context.
3. Generate and compare. The AI produces an image from your prompt, measured against the target.
4. Pass or retry. Clear the level's visual threshold to advance; you get three attempts per image.
5. Level up. Each level grows more complex, demanding more precise and complete descriptions.

Early levels invite you to describe basic, relevant information (the medium, subject, and context) while later ones require richer, more exact prompts to hit the mark. With three attempts per image and built-in tips and hints, the game guides you toward better descriptions rather than leaving you to guess, so you learn as you play.

What It Teaches

Beneath the game lies real, transferable skill-building.

1. Prompt-Writing Craft

The core lesson is how to construct an effective image prompt: what details matter, how to prioritize them within a tight character limit, and how word choices steer the result. These are exactly the skills that make any AI image tool work better for you.

2. Visual Literacy

To describe an image well, you first have to see it well. The game trains you to notice and articulate medium, style, composition, lighting, and mood, sharpening the way you read visuals, not just generate them.

3. Iterative Thinking

With three attempts and a clear target, you learn to refine a prompt based on what the AI produced, the same iterative loop that effective prompting relies on in real creative work.

Who Say What You See Is For

It suits anyone who wants to get better at working with AI image tools, or just enjoys a clever puzzle.

Aspiring Prompt Writers

Anyone learning to use AI image generators can build real, practical prompting skills here in a fun, low-stakes way, skills that pay off immediately in tools you actually use.

Designers and Creatives

Creatives who work with generative imagery can sharpen their descriptive precision and visual analysis, getting closer to the results they intend with fewer tries.

Educators and the Curious

Teachers exploring AI literacy and curious players alike will find it an engaging, hands-on introduction to how text-to-image AI interprets language.

Why Prompt Skills Matter

The ability to write a good prompt has quietly become a core creative skill. Every AI image tool, and increasingly every AI tool of any kind, depends on how well you can translate intent into language the model understands. Say What You See makes that abstract skill concrete and measurable: you instantly see how close your words got you, and learn what to change.

Google has noted the game also helps improve its own AI by surfacing how people describe images, but for users the payoff is direct: practice here and you will get better, faster results from generators like those built into the Gemini App and across the wider creative-AI landscape. It is a small investment that compounds across every AI tool you touch.

Pricing and Availability

Say What You See is a free Google Arts & Culture experiment, with no subscription. You play it in your browser through the Google Arts & Culture platform. As an experimental project, its availability may change over time, as with any Google Labs experiment. Check the official Arts & Culture page for the current version.

Limitations to Keep in Mind

Final Verdict

Say What You See is a small, smart gem: a game that turns the genuinely useful skill of prompt-writing into an addictive puzzle. By challenging you to recreate AI images through precise, concise descriptions, it trains both your prompting craft and your visual literacy at once, with the immediate feedback that makes learning stick. For anyone working with AI image tools, it is fun and quietly valuable.

It is a focused experiment about image prompting rather than a broad tool, and its threshold scoring favors accuracy over creative interpretation, but as a playful way to level up a skill that pays off everywhere, Say What You See is well worth your time. Put the skills to use in the Gemini App, and browse more free AI tools to round out your stack.