Every block plugin has AI now, but most of them advertise it the same way: enter a prompt and get a finished page. They (almost) never tell you how many prompts it took to get the result they are showing you.
For this post, I decided to try a more transparent approach using the Section Generator in the free Otter Blocks plugin. I built nine sections of a WordPress site with it, but instead of endlessly prompting until I got it “just right,” I relied on a single prompt. Then I took a screenshot of what that prompt produced so I could share it with you here.
Some results were good enough to publish. Others were not. The difference had almost nothing to do with how I worded the prompts and almost everything to do with what I put in them. Let’s take a look.
Before you start: connect your AI provider 🔌
Otter Blocks integrates seamlessly with the WordPress block editor, so accessing the Section Generator is as easy as finding any other standard WordPress block. It also doesn’t rely on AI credits as many other AI builders do, which is another plus.
Here’s how it works: you describe a section, Otter builds it with its own blocks, and then lets you edit everything afterwards. It’s easy-peasy, but you do need to connect to an AI provider before you can start generating sections. The steps are below:
- From the WordPress dashboard (i.e., the wp-admin), go to Settings → Connectors.
- Connect a supported AI provider.
- Return to Otter Blocks → Settings → AI and save your provider settings.
That’s it. Once the provider is connected, open the block editor, add the AI Block, and you’re ready to generate your first section.
If you need additional guidance on the setup, check out Otter Blocks’ AI documentation.
Now, let’s look at the sections I created and see how Otter handled each prompt.
Where it starts: the layout arrives, the words do not 🖼️
The first section I created was for testimonials. It was also the shortest prompt I typed during the entire experiment. More than anything else, it shows what the generator can and cannot do for you.
Prompt: A testimonials section with three customer quotes and names.
There were two parts to evaluate here: the layout and the actual text.
The first part came out good. I got back three cards, a quote above the name, plenty of space, and a built-in offset from the section above. In terms of layout design, it was a winner.
The words were a different story:
“The whole experience feels thoughtful from start to finish. We found our rhythm quickly and delivered better work with far less friction.”
It was textbook, generic AI slop.
I asked for a shape, three quotes, and three names, and I got a shape with filler inside it. The tool didn’t fail. It just did exactly what I instructed it to.
Later in the run, I asked for quotes that named a specific outcome, a deadline met or enquiries lifted, and the writing came out noticeably closer to something a client might say.
Either way, one thing doesn’t change: I made up every name, title and quote. For a real business, they’d have to be replaced with genuine customer testimonials before going live.
So, the useful part here was the layout, not the text.
Keep that split in mind for the remaining eight prompts. In almost every one, the layout held up. The words were a mixed bag.
Four sections where I gave it an opinion 💭
This is where the tool truly shined, and it took me most of the run to appreciate it.
A founder’s letter
Prompt: Three unpopular opinions from a wedding photographer about wedding photography. One sentence each, stated as claims not questions.
In hindsight, this was easily the best result of the day.
I got back a small label reading “A NOTE FROM THE FOUNDER,” the headline “Made slowly, with care in every small batch,” three paragraphs that argue for the practice rather than describe the company, and a signature for Mara Ellis, Founder and Soapmaker.
This was one of only two examples (of the nine) I would have used exactly as it was delivered.
Three unpopular opinions
Prompt: Three unpopular opinions from a wedding photographer about wedding photography. One sentence each, stated as claims not questions.
“A short wedding timeline produces more honest photographs than a packed schedule of formalities.”
A photographer could genuinely argue for that, and another might disagree. That’s what makes it an opinion, rather than a vague statement like “the whole experience feels thoughtful.”
The prompt even spelt this out: “stated as claims, not questions.”
Who I am not for
Prompt: Create a two-column section for a freelance copywriter: who I'm a good fit for, and who I'm not. Four honest points each. The second column should actually say no to people.
The headline it wrote, “A good fit matters more than a full calendar,” is better than the brief deserved.
Every point in the black card starts with “No”: no deceptive claims, no rushed projects, no endless revisions. The black card against the light one was its idea, not mine.
It also gave three points on each side, even though I had asked for four. Not a horrendous error, but noticeable.
A note about Mondays
Prompt: I asked for a short note from a restaurant explaining why it closes on Mondays.
Two paragraphs, one explaining the reason and one apologising to the people it inconveniences, signed “With gratitude, Chef Marco.”
A section like this is normally just a line of small print. Give it a reason to explain itself, and it becomes something people might actually read.
That was the real win across these sections: I gave AI the point, and it handled the structure.
Two sections where I only gave it facts 💯
Not every section needs a point of view. Sometimes you just need information organized correctly. The following two prompts tested how well the generator could do exactly that.
Pricing
Prompt: Create a pricing section for an online course. Three plans: Self-Paced $149, Guided Cohort $349, Mentorship $699. Guided Cohort is the popular one.
This came out ready to use.
It gave me back three columns, the middle plan in the theme’s accent color with a “Popular” badge, a headline it wrote itself, and one benefit line per plan that built on the one before.
I provided the prompt, included prices, and AI did the rest.
A size guide
A shirt size table with chest, sleeve and length measurements for five sizes.
Five sizes went in. Three rows came out.
M, L, and XL were grouped into one row instead of getting a row each. The generator had all the information, but I wasn’t specific enough about how I wanted it organized.
Looking back, what I should have done was give each size its own row, with the size and its measurements kept together.
This is a good example of why context matters. The more specific you are about the structure you want, the less the generator has to guess.
Two heroes, two different lessons 🦸🏼🦸🏽♀️
The dentist
Prompt: Create a hero for Bright Lane Dental, a family dentist in Bristol for nervous
patients. Evening and weekend appointments. Buttons: book a consultation, see
our prices.
The layout came back exactly how I requested it: a hero with a headline, supporting copy, two buttons, and an image.
It also picked up the main details from the prompt. Bristol, the family-dentist positioning, nervous patients, evening and weekend appointments, and both requested buttons were all there.
The main problem was the image. I described a dentist for nervous patients, but it chose a close-up of dental instruments in someone’s mouth. It matched the dental subject, but not the audience or the reassuring context I was going for.
A calmer image, such as a dentist talking to a patient, would have been a better match.
The plumber
Prompt: A hero for an emergency plumber in Bristol. Headline, the phone number as the main element, a line saying answered 24 hours. Image on the right of a van parked on a residential street at night.
This was the strongest layout of the run.
My prompt told the generator exactly what should dominate, and it followed that instruction to a T. The phone number was the main element, supported by the headline and 24-hour message, with no unnecessary hero buttons.
The image was the clear miss. I asked for a van parked on a residential street at night and got a truck on an open road in daylight.
The layout followed the brief. The image didn’t. If this were a real project, I’d swap the image, but I’d keep the rest of the section as is.
The pattern behind the good ones 🔎
After completing all nine experiments and reviewing them, I noticed that the good results had one thing in common: I gave them something specific to work with.
Facts. An opinion. A clear priority.
The weaker ones landed the layout, but the insides were metaphorically hollow.
The takeaway here is that Otter will handle the layout the vast majority of the time. To get good results beyond only the layout, you need to provide Otter with substance.
These three things help:
- Give it the facts. Prices, measurements, steps, names.
- Tell it what matters most. Make the priority clear.
- Give it a point of view. Generic prompts produce generic results.
Then check two things: did it return everything you asked for, and is the image actually right?
If things don’t come out exactly the way you envisioned, don’t panic. Every section remains fully editable in WordPress. That means you can change the copy, adjust the layout, or replace any image. In other words, you’re not stuck.
Try it yourself 💡
The best way to understand what the Otter Section Generator can build is to try it on your own site.
If you need a new WordPress section or even a whole page, take it for a spin and turn your prompts into polished layouts.
FREE GUIDE
4 Essential Steps to Speed Up Your WordPress Website
Follow the simple steps in our 4-part mini series and reduce your loading times by 50-80%. 🚀
