The Best Alt Text Artist in the World Describes Her Craft

Caroline Desrosiers smiling brightly

Descriptions are essential for those who are blind or visually impaired.

Whether it’s meaningful images on the web, illustrations in a comic book, displays at a museum, or another important visual, descriptions must be available.

But not just good descriptions, great ones.

Scribely CEO Caroline Desrosiers specializes in great descriptions — in fact, she’s built an entire business around delivering visual content that everyone can experience and enjoy.

“The web has formed deep roots in our daily lives — at any moment we can dive in and answer a burning question, connect with friends, study a beautiful image. Missing and inadequate descriptions block people from information and businesses from people. I started Scribely to help change the web’s visual bias and set a new standard for digital images that includes everyone.”

Caroline’s primary service involves providing high quality alt text for digital images but the scope of Caroline’s work extends much further than the web. Scribely supports all types of organizations, from museums and galleries to audio and voiceover production studios to create rich visual content experiences that optimize the experience for everyone.

“Technology increases the pace of our daily lives. We tend to prioritize speed and efficiency to increase productivity and maximize time. This has helped us advance as a society in so many ways, but sometimes it means we forget to stop and smell the poppies. Scribely’s team of expert writers are humans writing for humans. We believe every image is an opportunity to convey purpose, beauty, or detail to another person.”

In this article, I’ll focus in on alt text specific to the digital world but keep in mind that the formula behind great alt text extends beyond the digital realm to anywhere great descriptions are needed.

The Alt Text Creative Process

Alternative text or “alt text” is the text alternative programmatically added to non-text content (typically images) on the web.

Alt text helps those with visual disabilities who use screen readers know what the images are.

While simplistic on the surface, assigning the proper alt text value to an image is a difficult task. In fact, WebAIM, the best source of online accessibility information says, it’s “one of the most difficult to properly implement.”

When organizations drill down into the specifics and really start trying to optimize the web experience for their audience, they quickly realize why Caroline’s expertise is so important.

I asked Caroline to explain what her team takes into account for every client project and there’s a lot to consider.

Let’s run through the key considerations in Scribely’s formula.


Above all else, alt text must be accurate. It requires taking the extra time to research the descriptive terminology to properly identify people, objects and actions. Writers must be precise in communicating the content, context, and purpose of the image so as not to mislead or misrepresent the visual message. Accuracy is always important and especially when it comes to educational content like an electronic textbook or journal article.


One of the most difficult steps involved in writing great alt text is selecting the details that fit into a brief description, optimally 125 characters maximum. This takes a discerning approach to detail, succinct writing practice, and strong language skills. Mastering this skill takes practice and is critical to effectively describing digital images.


In situations where there are multiple images of the same subject, accounting for the distinctive details of each image is crucial. Repetitive alt text is not only dull for the listener, it misses the opportunity to describe what makes each image unique. This is especially important for images that appear in a sequence, for example product photos. When you look closely, each photo describes a different aspect of the product’s function and purpose. Image specificity helps users collect visual information, learn about products, and decide what they want to buy.


Writing excellent alt text is not just about identifying details; it’s about describing the purpose of those details. The same image may be described in different ways depending on its application. For example, an artistic photo might be used to achieve a brand’s specific look and feel. In this case, the writer selects adjectives that match the brand’s message. Another application for the same photo might be a virtual art gallery. In this case, the writer captures the aesthetic details of the photo (camera angle, blurred effect, contrast).


It’s important to read through and incorporate relevant text that is in close proximity to an image. For example, an article might include a written caption below each image. The caption will be read out loud to people using Text to Speech (TTS) or screen reader technology so it is not necessary to repeat it. The surrounding text and captions in an article are often written to accompany rather than describe what is in an image. In this case, the focus of the alt text is on the visual details the caption leaves out.


Some media types require a completely different approach altogether. For example, images in audiobooks must be described in a way that flows with the author’s writing style and voice. Many audiobook listeners are not accustomed to the typical short form approach of writing alt text. For this reason, writers should pay attention to the narrative flow to help create a listening experience that everyone can enjoy together.


Alt text requires empathy; the ability to describe visuals in a way that recreates the visual experience. As is the case with memes and GIFs — sometimes writers must switch up the order of information as it is presented in the image to recreate the experience. This involves an understanding of how to “break down” a joke or clever statement and place the punch line at the right moment.

Describe, Don’t Interpret

Adding your personal or company’s brand voice is a great way to create engaging alt text. With that said, it’s important to avoid using “positive” or “negative” words that would sway the user’s opinion of the image. For example, beautiful, hideous, adorable, bad, good, bland, exciting, funny, boring, perfect. Great alt text involves painting a picture that helps inspire listeners to interpret the meaning for themselves.

Is AI Good Enough?

One of the most common questions Caroline receives is, what about AI — can’t artificial intelligence solve my alt text problems in a jiffy?

With AI, machines use image or object recognition to insert words as alt text. We’ve seen AI image recognition attempted by Facebook and others with no success other than a checked box, so to speak.

What happens is an alt text value will technically be populated, but even the best results have only yielded related keywords separated by commas. For example, for an image of a woman at a coffee shop, AI might generate the alt text of: woman, coffee, table.

This is not the user experience any reputable company associates with.

Having studied automatically generated alt text at length, Caroline has found AI alt text to be highly problematic on multiple fronts.

Object Recognition

While AI object recognition is improving, it is far from perfect and will continue to make errors for the foreseeable future. A few examples of object recognition failures include, missing relevant or distinctive details that are unrecognizable to the technology, and misidentifying textured patterns that look like something else (ripples in the sand that look like water). The risks of incorrect or misleading descriptions are unacceptable and may lead to offensive descriptions, improper search tagging, and inaccuracies in educational or business content.

Logical Flow

For graphs, charts and other complex images, automated technologies fail to capture the logical flow of information. In these cases, it’s helpful to use a human-generated long description to break down the complexity of an image. Humans writing for humans. One approach to this is to start with the most general information and then work towards describing more specific details. This helps the user visualize a complex graphic gradually coming into view.

Making Comparisons

AI struggles to describe objects using descriptive comparisons. However, this is often the best way to quickly and efficiently describe something to another person. For example, when describing the size or shape of an object, a writer could say that a globe is about the size of a basketball.

Decorative vs. Meaningful

Another important decision Caroline’s team makes is when alt text is not necessary.

All images should have an alt attribute but alt attributes for decorative or non-meaningful images should be left blank. This falls in line with the requirements for Web Content Accessibility Guidelines (WCAG) and it also makes good sense.

Screen reader users don’t want every image on a page described to them in the same way sighted website users don’t pay attention to every single image.

Deciding which images are meaningful vs. decorative requires thoughtfulness and a keen eye.

Not only that, but sometimes an alt attribute should be left empty because adding a descriptive alt text value to the image would be redundant with the surrounding text.

Looking Forward

As Caroline’s clients continue to grow, she continues to refine her process as she strives to become even better than the best.

“2020 was packed with lessons for all of us. We learned it is critical is to be able to operate virtually. We learned diversity, equity, and inclusion are an essential part of business. We learned we need to work together to overcome adversity. Accessibility has never been more important than it is right now. Alt text represents an opportunity to break down barriers, reach more human beings, and help people visualize beauty.”

Caroline Desrosiers is changing the world, one alt text value at a time.