Games & Virtual Worlds Series

Google Images (& Pinterest) vs Generative AI for Board Game Visual Design Research

Notes on the Inspiration Tool Belt

Michael Filimowicz, PhD

Published in

Narrative and New Media

11 min readMay 7, 2023

For years I’ve been promoting the virtues of Google Images in my courses for doing image-based research, especially when seeking ‘inspirations’ for one’s design. Generative AI has added a new method for doing image research based on prompt writing which differs significantly from search writing.

When using Google Images, you have to conduct a web search, which means, regrettably, that you can only search for things that already exist. With generative AI, you can ‘search’ for things that have never existed, in effect asking a program to dream up images to match your text.

Google Image Search

The purpose of a Google Image search is to find and retrieve existing images based on specific keywords or criteria. It relies on an index of images available on the web. Users input text-based queries, and the search engine returns relevant images from its database. The results are typically real photographs or illustrations sourced from various websites.

Google Image search provides access to a vast collection of images created by photographers, artists, and other content creators. The images are pre-existing and have been captured or designed by real people, though one might anticipate that in the future, more AI-generated content will show up in searches. The search results typically represent the creative work of others and are subject to copyright or licensing restrictions.

Users have limited control over the specific details or characteristics of the images they find through a Google Image search. They can refine their search terms or apply filters, but the results depend on the available images in the search engine’s index. Interpreting images is left to the viewer, as the search engine does not have context about the user’s intent.

Pinterest searches are similar in this way to Google Image searches, in fact often the results you find via Google Images are essentially others’ Pinterest boards. Pinterest, like Google Images, relies on already-existing visuals but Pinterest boards have the particular advantage of expressing the curatorial intent of others who have deliberately created collections of images that are organized under some theme or heading.

Thus, you might use Pinterest board searches as an alternative or complement to Google Image searches, but note that even Pinterest shows a marked bias towards snaky paths just as Google Images does, when doing a search on ‘board game design.’

a Pinterest search result on ‘board game design’ showing plenty of snaky paths

Generative AI Image Creation

In contrast, generative AI image creation involves using artificial intelligence algorithms to generate new, unique images based on textual prompts. Instead of retrieving existing images, the AI model generates novel visuals from scratch. It utilizes deep learning techniques, such as generative adversarial networks (GANs) or variational autoencoders (VAEs), to synthesize images that match the given text description.

Generative AI models can produce entirely new and original images based on textual prompts. The AI algorithms learn from a dataset of existing images during the training phase and then generate novel images that resemble those in the training set. The generated images can be unique and imaginative, as the AI model extrapolates from learned patterns and generates visuals that it has not seen before.

With generative AI image creation, users can exercise more control over the visual outcome. By providing specific text prompts, users can guide the AI model to generate images according to their desired attributes, styles, or compositions. The textual input can include detailed instructions, such as specifying object colors, shapes, or even providing a scene description. Users can influence the interpretation of the image creation process.

Seeking Novel Board Game Visual Concepts

A typical thing one might do with a collection of images found via Google Image search is to create a mood board, similar to how one creates image boards in Pinterest. A mood board is a visual tool or collage that captures and communicates the overall mood, style, and aesthetic of a concept or project.

It is typically composed of a collection of images, textures, colors, typography, and other visual elements assembled on a physical board or in a digital format. Mood boards are used in various creative fields, such as design, fashion, interior decorating, and marketing, to convey ideas, inspire creativity, and establish a visual direction. They serve as a visual reference and help stakeholders and collaborators understand and align with the desired mood or atmosphere for a project.

Below, for example, is a mood board created by a Reddit user for refining the look of their game:

A Google Image search for “board game design” returns some very strange results, strange in that there some kind of evident bias towards snaky paths:

The results aren’t much better if you search “board game design ideas” which shows the same infatuation with snaky paths:

You can see that getting a decent mood/Pinterest board with Google Images might present quite a challenge when the results are so similar (and kind of boring, too)! The point of an image collection for inspiration purposes is to obtain a certain balance of uniqueness, similarity and diversity across a broad range of examples to feed your creativity.

Now, let’s compare the images above to some generative AI image results, based on the prompt a monochromatic cyber punk board game with neon metal pieces and something very exciting starts to happen — to understand just how exciting, enter the same text in a Google Image search and be prepared for major disappointment. The generative AI tool doesn’t disappoint, however:

Google Images gave us umpteen variations of snaky paths, but here just two images present entirely different board game concepts which is much more useful for thinking outside the box and getting the visual design inspiration juices flowing.

Now you might ask, “Hey, where’d you come up with that prompt?” but the beauty of generative AI is that you’re asking an algorithm to imagine something, not search for something that already exists, so you can prompt anything that you can think of. Comparatively, Google Image search seems rather finite while generative AI feels quite infinite in the possibilities.

If you want actual historical references in your game, so that players are reminded of canonical classics with deep cultural nostalgic resonances, either tool will have its uses, but you might find that by asking generative AI to imagine things that already exist, its results might be boring. For instance, here’s a comparison of a Google Image search and a generative AI prompt for innovative board game with retro chess board feel.

Google Image Search Result (source) AI art

Generative AI result, trying to be innovative with Chess. AI art

In the above example, I would say that Google Images is the clear winner, because it found something much more interesting than what generative AI could invent. On the downside, the best image it found is really just a traditional chess game with very artsy pieces and game board, whereas the generative AI made chess into more of a maze game which is a different kind of novelty.

Google Images will usually return a lot of useless images, forcing you to hunt for good finds, whereas generative AI has a strong likelihood of nailing it on the first go. Here’s a comparison of the results for outrageously complex tangible board game element for creative affordances and interesting mechanics:

Let’s try something very simple, directly in the graphic design vein: monochromatic game board. Generative AI emerges as a clear winner, with its ability to imagine radical new possibilities, rather than regurgitate familiar stuff that already exists!

Based on the next experiment, it seems that Google Images can get very confused, focusing on the wrong part of a phrase, in this case getting hung up on color theory with the prompt board game design in triadic color scheme whereas generative AI keeps intact the understanding that what you’re searching for is a board game design:

Hung Up on Color Theory in a Google Image Search (source)

Next prompt comparison: plastic pieces of a board game

Google Images returns pretty much the exact same plastic piece type, just as it did with the snaky paths above, whereas generative AI imagines a wider range of alternative possibilities quite easily:

Finally, let’s compare Google Images and generative AI results for board game design for a spooky dungeon game world:

This result is intriguing, because AI tends to generate 3D isometric game boards with this prompt (at least with the tool that I am using), which adds an intriguing immersive dimension to board games that usually have flat (2D) graphic design.

Google Images — which can be thought of as a web-based mood board that is pre-curated, needing an extra curatorial pass from human users— produces a strangely more cheerful (i.e. colorful) result, and includes game packaging and marketing graphics.

If you find an AI-generated design that you’d like to use outright, find an AI-based upscaling tool to enlarge the image, which will generate the missing pixels and often yields a high quality and printable (i.e. 300 ppi) graphic. Photoshop, for example, uses AI to enlarge images at up to 400% without artifacts, which might be a useful tool in your arsenal for prototyping board games. The same idea of course applies to smaller game elements, such as player pieces, cards etc.

Counteracting Algorithmic Bias with Daft Punk (or Mandalorians, even)

An annoying thing about using generative AI is that it tends to produce mostly white guys if you want to include human players in your design research inspiration images. It can also do some very odd things with faces and hands, even if you make an effort to add “multiracial” or “multiethnic” as words in your prompt writing, as shown below.

One way I sometimes solve these kinds of issues —the strange faces and hands, or too many images of white men — is to write the prompt so that Daft Punk is somehow involved, which puts helmets on their heads and sometimes gloves on their hands, so that I don’t feel like I’m always cranking out images of white dudes. Technically I prompt-craft ‘androgynous’ Daft Punks but generally one can’t discern a sex difference because of how suited up they are.

This first image below is especially great because it has little Daft Punk action figures on the game table:

Daft Punk to the Algorithmic Bias and AI Distorted Hand/Face Rescue

Or if you prefer Mandalorians to Daft Punk for anonymous helmeted people, that works too:

So, Just Experiment

Both Google Images and generative AI can be great resources when doing visual design research as part of your board game prototyping process, though Google Image results are increasingly seeming inferior and old-fashioned compared to the possibilities of generative AI. The only conclusion to a lecture like this is to encourage you to experiment as creatively and freely as you can, which the images below (all from generative AI of course) are meant to inspire in your inspiration seeking.

The prompts used to produce the images below use terms like affordances, game elements and game mechanics, which should get you going down some interesting paths in your visual design research.