AI patterns. Does anyone here use them? Are they already well-developed?

18 Posts Recent Started

Tuesday, May 12, 2026 at 7:34 PM

Hi everyone!
By now, everyone has probably noticed that more and more patterns are popping up on various websites (and here, too) that seem a bit off. A pretty picture, a promising description—but when you read it, you realize the stitch counts don’t add up or the instructions contain sentences that no one would ever actually write.
On the one hand, I think it’s exciting that AI is making its way into our world, and I like to use these myself for inspiration or to brainstorm color combinations. On the other hand, it annoys me when I pay nine euros for a PDF and end up spending three evenings unraveling my work (okay, it’s only happened once, but it’s still frustrating because of the time wasted).
How do you deal with this? Do you have any tricks for spotting AI-generated patterns right away without even looking inside… besides the obvious “photo that’s too good to be true”? I’m not even sure of my own answer right now. Best regards.

mermieswelt

274 Posts Recent Started

Wednesday, May 13, 2026 at 9:05 AM

I’m really sorry to hear that you’ve already had your own bad experiences with this.

It’s actually getting harder and harder to tell what’s real from what’s fake. Here on CP there are clear rules and guidelines, and content generated entirely by AI isn’t allowed. Authors are encouraged to label their use of AI (even if they’ve only used it for the background, for example), and many do just that. Thankfully. Because transparency and trust have become priceless.

We’ve reached a point where people are so unsettled that many now steer clear of something as soon as they have even the slightest suspicion that AI was used in any way for the images or the pattern. But they don’t just boycott the pattern—they boycott the entire designer behind it.

  I don’t have the ultimate tip for spotting AI today either, but I’d like to share a few general points with everyone:

  1. Before buying a pattern, read the reviews and check whether there are also photos from customers who have actually made the pattern. This is especially true for the major platforms, where so many scammers are now running around with their AI junk. Here on Crazypatterns, thanks to the great work of the CP team, you’re safe from scammers, and it’s enough to read the reviews before buying.

  2. Yes, it’s frustrating to sometimes spend a lot of money on a pattern only to end up disappointed. But not every pattern that’s written strangely is automatically AI-generated. Some designers aren’t proficient enough in German to translate their patterns themselves and—especially before the AI era—ran them through an online translator, which resulted in some truly bizarre translations into German, particularly from certain languages. And that brings me to my next point.

  3. Check beforehand when the pattern was uploaded. More and more often—even with big, established designers—I see customers writing in the comments section of a pattern, “This is AI, because the photo is too perfect.” If the pattern has been selling successfully for many years with good reviews, it’s very unlikely that AI was used. The fact is, there are actually many creators who are simply really good at taking photos 😉

  4. If you want to do something about all this AI junk, I invite you to support us real designers here on CP. You can make a difference with small actions:
- Review the patterns you’ve made,
- share pictures of your finished projects in the gallery
- and if you’d like, visit our social media accounts and leave some love there every now and then.

  This way, you can help us regain some visibility amid the flood of AI content.

I hope some of these points are helpful to one or two of you ❤️

veronika-t15

3966 Posts Recent Started

Wednesday, May 13, 2026 at 9:44 AM

Those “too-good-to-be-true” photos are obvious, so I get skeptical right away and take a closer look. But I’m also a bit lazy—to avoid all that hassle as much as possible, I only use patterns from here now. They pay very close attention to detail here, which I think is great, so I rely on them. And I’ve never been fooled yet.

CreatOwl

4492 Posts Recent Started

Wednesday, May 13, 2026 at 10:48 AM

I include a corresponding AI label on my photos.
This applies to the background, while the patterns themselves are shown as they originally appear.

As for the mention of reviews: Unfortunately, not everyone rates the patterns they’ve purchased.
So it’s hard—if not impossible—to tell whether they were satisfied or not.

Schnuckelina

6011 Posts Recent Started

Wednesday, May 13, 2026 at 12:56 PM

I can totally relate to Bianca’s concerns. Many people are feeling increasingly uneasy about buying craft patterns. I feel the same way, except when it comes to CP.

Since I’ve been an author here for 11 years and have been closely following developments, I know that CP handles the topic of AI very carefully and with great sensitivity. That’s why I only buy patterns on CP.

I can read reviews and comments from real people about the product.
I can message the author before buying if something isn’t clear to me.
I can contact the author after my purchase to clarify any questions.
If I get stuck on a problem, I can always contact CP support. There, real people still take care of the problem.

If I can’t find a pattern here and am forced to buy it on another platform, I read everything—about the product and the shop—very carefully. And if what’s written there doesn’t make sense to me, I don’t buy from there. You just can’t rely solely on the pretty picture.

Best regards, Ina

Schnuckelina

6011 Posts Recent Started

Wednesday, May 13, 2026 at 1:03 PM

P.S. Here’s a little “funny” story on the topic:
On a very large platform for patterns, I’d set my sights on a sewing pattern for a children’s dress. One AI-generated image showed a pretty girl wearing exactly the dress I wanted to sew. The second image even showed her moving around in the dress. The reviews included comments like: “The pattern doesn’t include a sewing pattern—just an A4 image of the sewing pattern.” Well, those who can read definitely have the advantage.

Best regards, Ina

ruthk

9969 Posts Recent Started

Thursday, May 14, 2026 at 7:33 PM

At least that was mentioned. Unfortunately, though, there are also countless AI patterns that are incomplete, and there’s no indication of that.

Operator

2855 Posts Recent Started

Thursday, May 14, 2026 at 7:42 PM

Hello,

thanks for bringing up this topic – it’s something many people in the community are currently concerned about.

We stated our position on this in March and set out our stance on AI-generated content in a separate thread. There you’ll also find our basic rule about what is allowed and what isn’t.

👉 AI-generated images & patterns – our stance

In short: The product shown (on Crazypatterns) must correspond 100% to reality and be exactly reproducible as shown in the image. Subsequent editing of lighting, scenery, or background is fine – the key thing is the product itself.

Warm regards
Your Crazypatterns Team

Schnuckelina

6011 Posts Recent Started

Thursday, May 14, 2026 at 8:41 PM

@Ruth,
The note that there was no sewing pattern included with the sewing instructions came from an annoyed buyer. You don’t get a list of everything you’re getting when you buy patterns. Everything is much more detailed and precise here on CP. If I’m missing any information, I can ask via PM beforehand. But sometimes you just fall in love with a design that isn’t (yet) available here on CP. In that case, you just have to take a closer look and read through everything carefully.

Since I’m grateful for such informative reviews, I’ve gotten into the habit of writing reviews myself. That’s the only way to make AI-generated patterns more recognizable to future buyers who might otherwise fall into the same trap. If we all did this, there would be fewer bad purchases.

Best regards, Ina

Tastenkrueppel

13 Posts Recent Started

Wednesday, May 20, 2026 at 9:39 AM

AI: Both a curse and a blessing...

I just read an article in the context of software development whose main point was, in essence, that while AI allows you to generate a lot of code that actually works in a very short time, the big “but” is that sooner or later someone has to review this code—for example, if it needs to be expanded or if it throws errors when combined with other code. And this review is increasingly becoming the specialty of experienced (and therefore expensive) software architects. The crux of the matter is also that writing code accounts for only a fraction of software development; the figure mentioned was 10% or less. So you’re saving time on a process that’s “fast” anyway, while overloading downstream processes that already take a long time.

The fact that AI-generated output always needs to be reviewed isn’t even up for debate. It’s just something that happens. Maybe people in software development are simply more aware of this; I don’t know 🤷🏻‍♂️

On the other hand, AI can be a great help, for example, in creating basic frameworks that humans can then build upon. Based on my own experience, I see a potential practical benefit in Arduino projects (or C++ projects in general), which typically consist of multiple files and often include several external code libraries. Building such a basic structure can be just as tedious and labor-intensive as cutting out an image to replace the background. This can almost certainly be automated reliably if you feed the AI accordingly.

As far as I’m concerned, craft designers are welcome to do exactly the same: just get a little help from AI here and there, carefully check its output, proofread it, make it themselves, and once everything looks good—go for it.

In my view, this doesn’t necessarily have to be labeled as (partially) AI-generated (though for the sake of transparency, I certainly welcome it), primarily for two reasons: AI isn’t really that “artificial” given that it’s trained on human-generated data, and it isn’t really that “intelligent” given that it relies on prompts formulated by humans—without which nothing happens. And if the “end product” makes sense, you can assume anyway that more than enough human editing has gone into it. And that brings us back to the beginning of the text.

I stand by what I’ve said elsewhere: AI itself isn’t the problem—it’s its far too frequent misuse.

Tastenkrueppel

13 Posts Recent Started

Wednesday, May 20, 2026 at 11:34 AM

Oh right, I haven’t even addressed the questions in the headline yet 🙈

I don’t even know if I’ve ever come across AI-generated or AI-assisted patterns before. If so, I wouldn’t really care as long as the result works out. Because that suggests that any AI-generated output wasn’t simply accepted uncritically. It’s worth mentioning that I’m part of the crochet crowd and work exclusively from crochet charts (currently working on some shawls by @KleinesFeines—definitely recommended!), and a crochet chart can certainly be generated based on detailed written patterns. The other way around is probably more difficult.

Is AI itself mature? I’d answer with a resounding “yes,” considering that AI does what you want it to do. The quality of the output then depends on how extensive and/or specialized its “knowledge” is and how precisely the request was formulated. And that’s also one of the limitations of AI that makes it seem immature.

There are AI models that can conjure up the most epic oil paintings in outrageously high quality right before your eyes, but don’t know what an RDstb is.

We cannot currently expect AI to become a universal genius in machine form. The task is far too computationally and memory-intensive for that. And it certainly makes your ears perk up when you’re told just how much energy can be required to process a single request.

Another limitation is an age-old one: AI is and remains a machine that, for example, doesn’t recognize irony—or barely does. While this improves as it gains the ability to contextualize statements, it introduces yet another level of complexity—and with it, even more computational and energy costs. Until then, it’s quite easy to fool it, lie to it, sabotage it, and feed it misinformation.

CreatOwl

4492 Posts Recent Started

Wednesday, May 20, 2026 at 3:07 PM

Misinformation—once it’s out there, it’s hard to counter it with facts.
Otherwise: AI is only as good or bad as the prompt entered by the person using it.

Tastenkrueppel

13 Posts Recent Started

Wednesday, May 20, 2026 at 7:04 PM

Well, just as easily as AI can be fooled, it can also be corrected with verifiable facts.

Here’s an anecdote from my own experience: I told ChatGPT about my plan to grow Vallisneria spiralis and Crassula ovata together in a window box and asked how I could best care for them. Anyone familiar with both plants knows the right solution. In any case, ChatGPT thought the idea was great and threw various care requirements for both plants right at me. Oookeee, let’s continue: I then asked how I could combine one care requirement from one plant with one from the other in a balcony planter. That confused ChatGPT so much that it just threw the same care requirements back at me.

Here’s the problem: Crassula ovata is a succulent, better known in this country as jade plant, baby tree, or money tree. Vallisneria spiralis, on the other hand, is a marsh plant that’s widely used as a low-maintenance aquarium plant (it’s the one with the loooong, narrow leaves that you see in almost every tank at pet stores). So it’s practically impossible to keep both of them alive in the same container long-term. ChatGPT did, however, acknowledge my correction—or rather, the fundamentally different living conditions of the plants.

That was a while ago, though, and things must have improved significantly since then.

At the end of last year, I asked ChatGPT how big a catapult would have to be to throw a table from Braunschweig to Berlin. ChatGPT first began by defining the table and then threw in all sorts of physical calculations. Dimensions, weight, aerodynamics, potential and kinetic energy, with and without wind, ... The key difference here was in the concluding sentence, which essentially read:

“In summary, you’d need a catapult the size of a small apartment building capable of generating the energy of a small atomic bomb to throw a table from Braunschweig to Berlin. The probability that the table would survive this ‘journey’ is practically nonexistent.”

I haven’t double-checked the math, but I think the summary is brilliant 😅

And it illustrates quite well that we actually need to train AI to stop supporting people in all their endeavors at all costs, and instead sometimes just say: forget it.

But to get back to the actual topic: I also asked ChatGPT to create a crochet chart for a two-color granny hexagon with a floral motif. It failed so miserably that one of the results was a “photo” of a circular woven coaster with a rose on it. The coaster was even just a solid beige color 🤷🏻‍♂️

bianca-aguck

18 Posts Recent Started

Monday, May 25, 2026 at 11:37 AM

Thank you so much, everyone! Let’s just try to be as careful as we can. I can’t tell the difference between AI and non-AI anymore. They’ve gotten that good. Every AI still struggles with difficult patterns. We should introduce a “Trust Seal” for genuinely authentic patterns. Best regards.