I’ve been thinking about this a lot lately.

As an editor, writer, and translator, I’ve been experimenting with AI—not because I think it’s a replacement for artists or writers, but because I want to understand what it can and can’t do. I’ve even used it with students to explore visual storytelling and occasionally for conceptual artwork to accompany a news story. I used to make stuff with Illustrator and Photoshop but I just can’t pay the monthly subscription for Adobe Creative Suites. I liked the idea that maybe we could create our own glossy cartoons with kids that looks Marshallese and not just vaguely Hawaiian.

After some experimenting, I created a series of postcards with students that made cartoon illustrations of them saying hi from their home islands in their island languages. AI can be useful, but it isn’t magic. The best results we’ve had came only after providing very specific prompts, reference photos, and source material. You can’t just say “add Marshallese flower” expect something culturally accurate. Careful that it doesn’t add Chuukese motifs to a postcard of the Marshall Islands, or add Marshallese flowers to a postcard of a waterfall in Haiti.

I’ve learned that AI is only as good as the cultural knowledge you put into it—and even then, it still gets things wrong.

One of the best examples came from the recent discussion about the Miss Marshall Islands pageant artwork. Ronnie Reimers pointed out that the illustration distorted the 24-point star on the Marshall Islands flag, altered the National Seal, and even created a canoe that, in his words, “as soon as wind blows on that sail, Moana is going straight down to meet Lobaibaat.” https://www.facebook.com/20531316728/posts/10154009990506729/

It was funny—but it also made an important point.

Those aren’t just artistic details. They’re cultural knowledge.

At Chikin Melele, we’ve spent far more time telling AI what not to do than asking it what to create. Don’t alter the photographs. Don’t invent people. Don’t change the flag. Don’t add U.S. seals where they don’t belong. Don’t change the shape of a fan or change the details on a jaki.

One time it replaced Kathy Jetñil-Kijiner with another Marshallese woman just when I was splicing three photos together and adding a text. Another time it invented a crew for the MV Meram. I’ve had to tell it repeatedly not to “improve” photographs because it starts hallucinating details that were never there.

Ironically, after we published a video illustrating the fire aboard the MV Meram, some people accused us of using AI to fabricate the footage. We hadn’t. We sourced verified photos from government sources and used conventional video editing tools to create a visualization and clearly labeled it as such because, obviously, no one escaping a burning ship out at sea stopped to take a selfie. Good journalism sometimes requires visual explanation—but it also requires transparency about what is real and what has been recreated.

Working in translation has probably made me more sensitive to these issues than most. Every day I work with translators trying to preserve the meaning of their languages and represent cultures faithfully. Not just Marshallese and Chuukese translation but also Tuvaluan, Rotuman, Solomon Pijin, and Fiji Hindi. A single word can change meaning. A single mark can matter. A language name, a place name, a chiefly title, or the direction of a canoe sail all carry knowledge that outsiders often don’t recognize.

Images deserve that same care.

When AI draws the Marshall Islands flag incorrectly or invents an official seal, it’s the visual equivalent of misspelling someone’s name or publishing the wrong translation. Most people may never notice. The people who belong to that community always will.

That doesn’t mean AI has no place. I’ve actually found it incredibly useful for brainstorming layouts, testing concepts, and creating visual placeholders, even thinking through possible translation approaches. I’ve followed (and commissioned) a Māori artist who intentionally trains AI on his own work so he remains the authors of what the technology produces. I think that’s a fascinating direction.

But for me, AI is still just a tool like Photoshop and Grammarly.

I’m still saving up to commission original artwork from artists like Debby, Chewy, and Ronnie whose lived experience, cultural knowledge, and artistic voices simply can’t be replicated by a prompt.

Until then, I’ll keep experimenting, keep learning, and keep checking every flag, every seal, every canoe, every caption, every translation, and every image.

Just see what it did when I took screenshots for the headline image (above) and asked it to just splice the screenshots together. Who the heck is that guy?! It’s definitely not Ronnie Reimers. They spelled his name wrong too. And, is that canoe supposed to be Chikin Melele’s logo? The logo that Ronnie designed for us?! Not even.

Because representation isn’t just about looking right.

It’s about getting the story right.