I’m a makeup artist who asked ChatGPT to recommend products for me. The results seemed promising at first, but the cons kept adding up.

To get the best shade matches, I needed to identify the exact colors of my skin, hair, and eyes. To do this, I went to a website called ImageColorPicker.com and uploaded a photo of myself I’d taken in natural lighting.

Then, I clicked on each feature I wanted to capture — my hair, eyes, and skin — to get their HEX codes, a unique combination of letters and numbers specific to a certain hue.

I copied my HEX codes and pasted them into a prompt I’d written out to send to ChatGPT.

I added my HEX codes to my ChatGPT prompt and sent the following message:

Pick makeup that will best flatter me based on the following HEX codes for eye color, skin color, and hair color. Choose five of each: foundation, lip product, and eye-shadow palette. Limit options to products available at drugstores like Walmart and Target.Skin color: [d5916f]Eye color: [494956]Hair color: [353942]

When I sent my prompt to ChatGPT, the software recommended five drugstore makeup products for each category, all from an impressive variety of brands.

ChatGPT even explained why each product might work for my skin tone, noting the specific undertones and coverage levels where applicable.

ChatGPT’s response seemed promising, but I was surprised to find that many of the recommended shades either didn’t seem to exist or weren’t stocked at my Canada-based stores.

For example, recommended products like the Revlon Super Lustrous lipstick in Cocoa Bronze only seemed to be available from online resellers, and I couldn’t find the L’Oréal True Match super-blendable foundation in N7 Classic Tan — though the foundation is available in a shade called N7 Medium Deep on L’Oréal’s website.

Finally, after some searching, I was able to find a lip cream and eye-shadow palette in the shades ChatGPT recommended at Shoppers Drug Mart. However, I couldn’t find any of the suggested foundation shades in-store, so I ordered one on Amazon.

I had the Maybelline FitMe matte and poreless foundation in my kit when I first started doing professional makeup, so I was most excited to try this product again.

Though I was skeptical when I saw shade 338 in the bottle, after applying it to my face with a brush, it ended up being a perfect match. The foundation’s natural, matte coverage was also exactly what I wanted for an everyday makeup look.

I’ll definitely be reaching for this product again. 

Out of all the recommendations I tried, the CoverGirl TruNaked nudes eye-shadow palette was probably the most disappointing.

ChatGPT described this palette as having “versatile neutrals and browns to create everyday or smoky looks,” but when I examined the colors, I thought the shade range looked too cool-toned, or gray, for my skin tone.

The palette didn’t have a diverse range of shades, which seemed limiting, and I had a hunch that the color payoff — how much pigment would be left on my skin — wouldn’t be great.

I applied a brown-toned eye shadow onto my lid with a brush. As I suspected, the shades looked too cool-toned for my complexion. Even the deepest-brown shade ended up looking too gray, instantly washing me out and aging me.

Though I’ve used the Nyx soft matte lip creams before, I’ve never tried the shade Abu Dhabi. According to ChatGPT, this color was supposed to be “a soft nude that complements warm skin tones without washing you out.”

In the bottle, the lip product looked similar to the nude shades I usually apply. As a Black woman, I knew I’d have to pair the light color with a darker brown lip liner to create more dimension, so I followed my usual process.

Unfortunately, even pairing the product with a liner couldn’t save this color. When I applied the product, it looked like I had concealer on my lips, which washed me out. I wouldn’t go out of my way to find this shade again.

Overall, the biggest limitation I noticed with ChatGPT was its inability to correctly identify items in existing shades or recommend products that are still in stock at drugstores (and not just resale sites).

I had to look up the recommendations online to confirm they existed, so using the software didn’t save me much time, either.

Though giving ChatGPT the HEX codes for my skin, hair, and eye colors seemed like the easiest way for the chatbot to determine appropriate shades, I don’t think this method can accurately assess warmth or capture the depth and dimension of one’s complexion.

The foundation I tried looked good and matched my skin tone, but the eye shadow and lip cream weren’t flattering on me at all.

In the future, I’ll probably use ChatGPT to pinpoint prices or get generic product recommendations for oily or dry skin. However, I’m going to leave the shade-matching to the humans.

The beauty of professional makeup artistry lies in understanding that every face is a unique canvas. If you ask me, ChatGPT offers a paint-by-numbers approach in a world that demands custom artwork.