CASE #WP0-A

Spot the Pipeline

OPEN INVESTIGATION

Test your ability to identify which AI image generator created each image.

The Assignment

You'll examine 25 images generated using five different text-to-image models and pipelines. For each image, identify which of these pipelines was used to create it, based on the stylistic qualities of the image alone. The prompts used to generate the images have been deliberately stylistically underspecified to capture the model's default aesthetic character.

Data Handling

We record your choices, response times, and experience survey answers. No personal information is required. Data is stored securely and used for research on AI image aesthetics. You can optionally submit a handle (e.g., Discord or Reddit) to join the leaderboard, and provide an email address at the end for result notifications.

Time Commitment

8–12 minutes

EVIDENCE 01/26

00:00

EXHIBIT A-001

Which generator produced this image?

Investigator Profile

Before we reveal your results, tell us about your experience with text-to-image AI tools.

How long have you been using text-to-image AI tools? *

I haven't used text-to-image AI tools Less than 1 month 1–6 months 6 months – 1 year 1–2 years 4–6 years (OG prompter with DALL-E) 6 years + (I was there for DeepDream and StyleGAN)

Select an option

On average, how many hours per week do you spend generating images with AI tools? *

0 (I don't actively generate images) Up to 1 hour 1–5 hours 5–10 hours 10–20 hours 20–40 hours 40hrs/FTE + (I do this for a living)

Select an option

How familiar are you with each of these specific tools? *

CASE #WP0-A

Investigation Complete

CASE CLOSED

--/25

(--%)>

Chance performance: 5/25 (20%)

Thank you for participating! Your responses help us understand how recognizable different AI generators are to experienced users.

Optional feedback questions

Share as much or as little as you like. All fields are optional.

1. Any tells that help you identify particular models/pipelines? (Optional)

2. Has spotting default styles become easier or harder as models have developed? (Optional)

3. Which models have the greatest stylistic range? (Optional)

4. What properties influence your selection beyond style? (Optional)

5. Any other observations or feedback? (Optional)

Join the leaderboard with a handle (optional)

Send me results/updates about this research (optional)

CASE #WP0-A

Leaderboard

CASE CLOSED

Case filed. Thank you for your contribution.

Top investigators

Rank

Handle

Score

Loading leaderboard...