Long Live Lorem Ipsum
Think before you 'GPT it.'
When designing, ask yourself one key question: Who are you talking to at what stage of the process? Managing expectations around Design is crucial at every point for every audience, and copy plays a critical role. Here's how that plays out, from low-fidelity wires to ready-to-ship experiences.
Context
Chat GPT and other Foundational Models are good at creating copy that is 'good enough.' Writers rightfully critique the stock, out-of-box quality. I'm not a stellar writer, but I can spot a GPT-generated listicle when I see one.
When properly trained, a Foundational Model can write something with a voice and tone that is passable for the person who created the original content. The source content can be written, transcribed voice, or video and often only needs a little pre-tagging or cleaning or tagging for vector DB conversion. Forgive me for getting all nerdy there. These machines are hungry and want to use all the computing we give them. They are no longer task-based, requiring manual training like they used to.
Reminder
There is still a place for all things ipsum. If you are out of time and need to place something into a layout or design, there are many auto-generated tools on the 'web. As far as I can tell, Hipster, Pokemon, and Jeoff Goldblum ipsum are all free.
Coverage
I will only address the Middle and High Fidelity context here, as Low Fidelity is hand or speedy digital sketching for layout purposes. If you intend to do user testing with Low-Fidelity artifacts, then you should not use Chat GPT for placeholder copy.
You want the experience design and the content to have matching levels of fidelity. If the experience and content are unequal in refinement, the artifact will confuse more than it provides clarity.
Mid Fidelity
If you are creating wires, use Ipsum. Wireframes, even color-blocked wireframes, are a gesture towards a solution. Sure, wireframes are better than a hand sketch, but they lack resolution enough to be paired with anything other than an H1 or H2 to inform the type of content they present.
Exception: You may have final copy for a product experience redesign, which is great. That said, when placing that copy into a new, better vehicle for experience, the content could benefit from some iteration along with the design. I have been there and seen this opportunity.
For Product Redesign copy, your decision to use Chat GPT is driven by time and resources. If you use Chat GPT at this stage to generate copy, do so with transparency and care--you will make copy part of the conversation that may eclipse your designs.
High Fidelity
At this stage, you have at least the first copy draft to place into a design that has a Design Language applied. If you don't have draft copy, then it's ok to use Chat GPT to spin something up. If you use Chat GPT, do so transparently and ask someone for guidelines or high-level guidance. Otherwise, the machine will give you the wrong output because you need to specify what you want.
High-fidelity designs that have Ipsum are making a statement. These designs intentionally say, 'You can put just about anything in this space.' High-fidelity designs with Ipsum drive a conversation about copy and the need for something that pairs with the rigor of the UI and UX, enhancing the experience for users/consumers/buyers.
Provocation
If you are curious and want to play around with an LLM, consider developing an internal "lorem ipsum" tool. You likely have access to content to feed into the model based on technical documentation or training videos. Or riff off of something trending like "The Bear" or "Saltburn," or a passion of a key stakeholder, like golf or maybe the local sportsball team (go <insert team!>).
So What?
Only use Chat GPT to generate copy if you are:
Open and honest about using the tool.
Prepared to defend the use of the tool.
Directing prompts with purpose and care.
Editing the output, even slightly.
Using the copy for size, proportion, and styling--not as a replacement writer for final copy.
If you are a designer using Chat GPT to write final copy, be aware that if/when it is acceptable, you now have a new job. You solved a problem and might be a temporary hero of sorts--but you also increased your workload. Even with the assistance of AI, was this your goal?
