Why user testing matters more than ever

Editing for plain language is often a humbling exercise. The idea that communication should be so clear, well-organized, and well-designed that an intended audience can easily find what they need, understand what they find, and use the information seems straightforward enough. Yet no matter how much we think we’re streamlining the structure and clarifying the language, user testing consistently shows us there’s room for improvement.

When WCEA partner Lana Okerlund and I updated one of our plain language courses this summer, one table in particular showed us just how valuable real human feedback can be.

In our course, learners work through a scenario in which the final exercise has them thinking about how to present a paragraph full of numbers. Most learners suggest a table, and we wanted to show them a possible solution that illustrates some best practices for plain language design. Here, I’ve created a fictional scenario similar to the one in our course to show how human feedback helped us arrive at the table we use in our answer key.

A family of four is requesting a refund for cancelled flights purchased through a travel agency. They have written the following summary:

The amount owing for cancelled booking QX14369 breaks down as follows. On June 12, we paid you a deposit of $1,000 for the flights. We paid the balance of the flights of $2,936.85 on August 25. This amount includes the balance on the airfares of $2,500 plus the Air Travellers Security Charge of $132.00, Airport Improvement Fee of $105.00, NAV Canada surcharge of $60.00, and GST of $139.85. On August 25, we also paid your agency’s booking charge of $420.00. The total we paid you is therefore $4,356.85. We are also claiming compensation for inconvenience of $1,600 from the airline.

Many of our learners use artificial intelligence tools to draft text. So, although Lana and I did not ask an AI agent to propose a solution to our course scenario, I was curious what it might do with this paragraph. Here’s what Gemini did with the data when I asked for a table:

AI version

Item Description	Date Paid	Amount (CAD)
Flight Deposit	June 12	$1,000.00
Flight Balance (Airfares, Surcharges & GST)	August 25	$2,936.85
Agency Booking Charge	August 25	$420.00
Total Amount Paid		$4,356.85
Airline Inconvenience Claim	Pending	$1,600.00

When it comes to evaluating plain language, many researchers have noted the limitations of software tools. And this table shows some of those limitations. Admittedly, the AI agent didn’t have a lot of context to work with, but it was not able to infer from the paragraph what a reader might have been looking for. Here are a few of this table’s limitations:

• The table has no title to guide readers to its purpose.
• The heading label “Item Description” is unclear. What connects these items? And why are all of the item descriptions in bold type?
• The table highlights the amount that’s been paid, not the amount owing to the family.
• All of the columns, including the money column, are left-aligned. (Columns with decimals should line up along the decimal.)
• The table includes grid lines and repeated dollar signs that add extra clutter.

Readers are left not knowing why they’re looking at this table or what it’s showing them.

By contrast, our first attempt to create this table ourselves avoided some of these problems:

Human version #1

Breakdown of amount owing for cancelled booking QX14369

Payments owing	Canadian dollars ($)
Refund for airfares (including surcharges and GST)	3,936.85
Deposit paid ($1,000.00 on June 12)
Balance paid ($2,936.85 on August 25)
Compensation for inconvenience	1,600.00
Total owing from airline	5,536.85
Refund for booking charge paid to Agency X (August 25)	420.00
Total amount owing	5,956.85

In this table, we:

• Added a title to make the table’s purpose clear
• Added column headers and descriptors (refund, compensation) to clarify that money was owed
• Left-aligned the descriptive columns and right-aligned the numbers column to support ease of reading
• Separated the amounts owing from the airline from the amount owing from the travel agency
• Deleted grid lines and dollar signs to reduce visual clutter, and added more white space between lines (Gemini’s version was single spaced)
• Highlighted the total amount owing to guide readers to the most important information

The content was closer to what we were looking for, but the table still seemed unnecessarily complicated. We went back to the plain language standards published by the International Organization for Standardization and Accessibility Standards Canada and drew on advice from Sally Bigwood and Melissa Spore’s excellent 2013 book, The Designer’s Guide to Presenting Numbers, Tables, and Charts.

Human version #2

Breakdown of amount owing for cancelled booking QX14369

Payments owing	Date paid	Canadian dollars
+ Refund for flight deposit	June 12	1,000.00
+ Refund for flight balance	August 25	2,936.85
+ Compensation for inconvenience		1,600.00
= Total payments owing from airline		5,536.85
+ Refund for booking fee owing from Agency X	August 25	420.00
= Total amount owing		5,956.85

In this second version, we made the following revisions:

• Three clear columns without indents made comparing data easier. (Vertical columns are easier to compare than horizontal ones.)
• Adding plus and equals signs made clear the relationships between individual lines.
• Setting the dates in their own column made it possible to scan this information quickly.
• Grid lines clearly separated the header line, bottom line, and total payments due from the airline.

While this version looks cleaner and clarifies the relationships between individual lines, we weren’t finished. When we gave this table to some proxy learners, they helped us improve it further. They wondered if the dates were really important and if readers might misunderstand them as being dates on which the amounts owing from the airline and the travel agency were paid. They also suggested bringing back the dollar signs to make it clear at a glance that the numbers signified money.

Human version #3

Breakdown of amount owing for cancelled booking QX14369

Payments owing	Canadian dollars
+ Refund for flight deposit	$1,000.00
+ Refund for flight balance	$2,936.85
+ Compensation for inconvenience	$1,600.00
= Total payments owing from airline	$5,536.85
+ Refund for booking fee owing from Agency X	$420.00
= Total amount owing	$5,956.85

We used this version of the table in our answer key, and we’ll continue to revise it based on learner feedback and emerging best practices. But the point is that we could not have achieved this table either with AI alone or on our own. Testing our material with real users produced the best result—and will continue to drive future changes.

While AI agents recognize patterns, these increasingly common tools don’t understand context or emotion. In this case, AI could not identify the purpose of the table and focused on the wrong information. Editors excel at structure and meaning, but we anticipate how people should or might react, rather than how they actually react in the real world. Only users can tell us what they understand and what they find confusing. Together, we can come up with the clearest, most useful solutions.

Want to know more? The Center for Plain Language is hosting User Testing: Why It Matters, What Gets in the Way, and How to Get Started on October 13, 2026, the 15th anniversary of International Plain Language Day.