Update on Similar Refusals

[Raul Ceron]

Hello!

You may have noticed in recent weeks more content being refused as “too similar.” This is part of our ongoing effort to maintain the relevancy and discoverability of content within the Adobe Stock collection. 

We understand that an increase in refusals can be discouraging, but this is an opportunity to refine your portfolio and submit your strongest content in future submissions.  

To learn more about how to submit distinct content that stands out in our collection, please refer to our Learn & Support article.

Thank you for being part of our Contributor community! 

[ngap_3964]

Adobe Stock's high rejection rate for "similar content" likely stems from an overly strict, poorly tuned AI algorithm prioritizing quantity reduction over quality. Internal KPIs may incentivize high rejection rates, while cost-saving automation reduces human oversight. Lowering upload limits (e.g., from 3000 to 300) was avoided to prevent user backlash and revenue loss. Lack of improvement reflects weak feedback loops, competitive market tolerance, and resource prioritization.

[Abambo]

Lowering upload limits (e.g., from 3000 to 300)

By @ngap_3964

They listened to you. They lowered the submission limit. The rest of your rant does not make great sense, however.

[ngap_3964]

Whatever, since they allowed around contributors to upload 3000+ images, I would queue my images in the database first before they solve the issues. If my queue close to 3000(ex : 2900) but  they haven't solved this issue yet, I could try to generate 100 images without checking the quality and just insert generated keywords, titles from time to time. Since the similarity criteria of the algorithm are almost random, for contributors it don't worth the effort to submit the contents after edit/filter.

[ngap_3964]

If my assumptions are correct, what would happen?

1. Reduce the number of uploads
2. Do not upload until this problem solved
3. Upload a lot of images without thinking twice because the algorithm do not have any logic to follow

If conditions 3 is true, that means some/many contributors do not affect by this similar images issues is not because they are "special treated", it is because no matter this new algorithm apply or not, they still doing the same thing--upload many, many similar images without edit/filter.

Upper managements do not care about details, they only care about the chart. Engineers has more interest about the KPI rather than the algorithm actually work or not; Moderators are happy as long as they do not need to review so many images. Forum manager are happy as long as upper management are happy.

[Jill_C]

Number 3 is definitely still happening - thousands of images submitted in a batch, and the Contributor doesn't seem to care if a high percentage of them get rejected because the entire process was automated and little of their time was invested.

[Abambo]

There exists, since "ever", a search functionality for buyers to find similar assets based on an image.. That search functionality can probably also be used for detecting similar submissions. This is based on visual aspects of the image and not on keywords or titles. I doubt that titles and keywords have a big influencee on this.

[ngap_3964]

Based on the search results, I am not surprise that the "similar content" criteria are close to random

[daniellei4510]

Yes, there is definitely some randomness going on, based on my own experience. While I see similarities in style based on these examples, they all look relatively unique to me. Also, how long ago were these submitted? The "too similar" rejection was only introduced a couple of months ago.

[ngap_3964]

The first one is submitted around 22days ago, other images I haven't submitted yet

[ngap_3964]

Some experiments, I submit them within last 7 days, the "find similar" results tell me there are over 37 millions(at least) similar images compare with my images. You guess what? Except of the last one, every image got rejected. I highly doubted the engineers of the adobe stock just filter out the images by random criteria.

[AmandaYuyi]

Still see lots of similar images approved. Below are the most recent transparent PNG approved for the same contributor. Based on similarity rejection experience, such kind of similarity assets will be refused as similar. Can please explain the actual logic or algorithm of similarity works? 

[daniellei4510]

Yes. This is what makes the "too similar" rejection so frustrating when unique content is in fact being rejected.

[Abambo]

Can please explain the actual logic or algorithm of similarity works? 

By @AmandaYuyi

See here: https://community.adobe.com/t5/stock-contributors-discussions/update-on-similar-refusals/m-p/1530945...

Your example should have been refused since ever. 😉 

[Jittiwat C.]

Regarding the duplication of content in Adobe Stock compared to the millions of content in your library, it is very difficult to prevent every creator from creating unique content. I think it is better to let the buyers choose what kind of work they want. If they reject it because of duplication of content, I think it is very difficult to make it unique. Let me ask you, if you were a content creator and created your own content, and your library has millions of works, would you not create works that are duplicates of what you have created before? As for the reasons for rejecting works in many cases, they only speak in general terms and do not go into details. Because I believe that the creators who submit their work have studied the requirements very well and know that the work meets your requirements. But when it is rejected, it is not clear why it was rejected from the reviewer's perspective.

[Jill_C]

You said "Because I believe that the creators who submit their work have studied the requirements very well and know that the work meets your requirements."

I am quite sure that this is often NOT the case. Many new Contributors start uploading content not having read the guidelines or even trying to understand what Adobe is looking for. They also attempt to copy other images that are already in the database. We've seen many such Contributors here in the forum who are unable to identify the flaws in their submissions. 

[Jittiwat C.]

Thanks for your perspective, Jill_C. I understand your point about many new contributors not reading the guidelines carefully or trying to copy existing work. I apologize if my previous comments sounded like I was generalizing too much. I agree that this is certainly a factor in rejections.

However, the bigger problem I and other creators face (even those who have studied the guidelines carefully) is the sheer difficulty of creating original content when there are already millions of assets in the library.

Moreover, the reasons given for rejections are often very general. They do not specify which of the existing submissions we submitted are similar to which existing work, or are deemed too similar in any particular way (e.g. composition, content, style, etc.). This lack of specific feedback makes it extremely difficult for creators to understand why their work was rejected and how they can improve future submissions. This can be a major frustration for many of contributors

[ngap_3964]

The critieria is likely random, you could check my previous post, there are some experiments with the "find similar" functions of adobe stock. However, reject the images based on random criteria or allowed similar images flooded the database, neither of them are good solutions for adobe stock. For contributors, I guess the best strategy for us are

1. Do not submit before this issue fix
2. Reduce the number of submit
3. Generate the images with the same prompts without checking and perform batch submit, with the same title and keywords
4. Generate the images with the same prompts and filter out the images with obvious AI flaws by eyes, submit with the same title and keywords

Adobe stock is just a small portion of their business, I would bet most of the competentce engineers are in the photoshop team

[Abambo]

Adobe stock is just a small portion of their business, I would bet most of the competentce engineers are in the photoshop team

By @ngap_3964

That's not how companies of that size work and function.

[Abambo]

Moreover, the reasons given for rejections are often very general. They do not specify which of the existing submissions we submitted are similar to which existing work, or are deemed too similar in any particular way (e.g. composition, content, style, etc.). This lack of specific feedback makes it extremely difficult for creators to understand why their work was rejected and how they can improve future submissions. This can be a major frustration for many of contributors

By @Jittiwat C.

Rejection reasons are not often very general, they are always very general. The moderator ticks Reason one to six and that's it. Very fast. The text is the same everywhere and you get it in your language, may that be Chinese, Spanish or English. And it even changes, when the text changes. I have refusals from 8 years ago that make reference to not ticking the "generative AI" flag. 

As for the "similar" refusal, we do not know how that gets constructed, but for sure, Adobe uses an image search tool that returns results in a certain manner. The nearest you can get is the built-in image search tool from the stock site. I do not think that they are using something fundamentally different. 

[ngap_3964]

More detailed tests, 05/10/2025

Keywords

anime girl autumn cat ears， 40 results

anime girl autumn reading, 195 results

anime girl autumn slime, 6 results

The results of keywords and the title search show us this is a niche market, you should be safe from "similar content" issue, right? Unfortunately, "find similar" give you another conclusions

Every picture has more than 30 millions "similar images"

You may ask, "if we could find the image with low search results by keywords and find similar, would it avoid the issue of "similar content"? The answer is no

I really don't get it how their algorithm works, it looks like totally random, unless for anime style images. Maybe old contributors who has more than thousands images would have higher success rate? Looks like they just randomly filter out 80%~90% and tell the managers "We get the job done, clean and fast"

[AmandaYuyi]

The current situation is even if you do research and submit unique content, they are often rejected as similarity. However, when checked the most recent approved content in AS portal, still can see newly approved assets are similar to millions of existing assets. So, doing research carefully cannot totally prevent from similarity rejection. The similarity rejection algorithm is totally random. 

[ngap_3964]

Looks like we either accept the similar contents issues or stop uploading until it fixed. I prefer to wait until the issue fix, before that would try another platform.

[teresao58254354]

I've had my share of rejected "similar images" lately. Also seems that editorial content (no people, mostly natural science museums, landmarks, and public art) have been arbitrarily rejected more often as well. Some batches are accepted, some from the location aren't accepted as they have an "illustrative editorial issue."

Meanwhile, I have seven photos I submitted six months ago that are awaiting review. And there are only three similar images. What's the deal Adobe?

I've been uploading to Adobe since 2016 and have weathered the changes you've made (editorial photos can no longer have recognizable people in them).  At some point, submitting to you won't be worth the aggravation.

[Jill_C]

I have also been a Contributor since 2016, and was getting ~95% of my submissions accepted until a few months ago. Now, the reject rate has soared to 40%, mainly for "similars" and "quality issues". I also have some images that waited between 4 and 11 months to be reviewed. I have nearly halted my uploads for now, because it's just not worth the time required to edit, submit, title and keyword just to earn an arbitrary, unjustified rejection.

[AmandaYuyi]

When seeing some of the most recent approved assets in AS portal, can especially feel the Arbitrary and unjustified rejection is random.