"Similar content” rejection with zero actual similarity – how is this decided?

[Vikky1904]

Why does Adobe Stock keep rejecting new AI submissions as “Similar content” when there is clearly no similar image in my portfolio or on the platform?

I carefully checked using the Visual Search and manual keyword browsing — and found nothing even close in style, color, composition, or subject.

:question_mark: So here’s my direct question:

• Where can we see what Adobe considers “already existing” in the library?

• Where is the list of “overused themes or visuals”?

• How are we supposed to compare before uploading, if these resources are not public?

I’m not talking about random gradients or textures. My content is well-styled, curated, and market-friendly. But I get rejected with “similar” without any actual references.

If Adobe is moving toward exclusivity — fine, but authors need clarity. We are trying to work with the system, not against it.

Please provide real examples or guidelines, not generic advice.
Thank you.

[jie_8284]

If similar, means, just similar color schemes, but different concepts, then I'm happy:

Left: pixie girl. Right: "similar images" according to Adobe.

• Adobe Algorithm: Looks at pixel-level similarities (color, light, composition).

• Human: Interprets context, story, and emotion.

[Vikky1904]

If the Adobe Stock library already holds over 800 million assets,
maybe it’s time to start cleaning out the archive from 1949? 😂
I mean, who’s really looking for those outdated gradients and "Word 2007 clipart" backgrounds in 2025?

Trends evolve. Visual culture moves on. And honestly, there’s no such thing as “identical images” when we’re talking about hand-styled compositions and modern aesthetics.

But if the review system is actually flagging images based on keywords instead of actual visuals —
then yeah, that’s a total f̶a̶i̶l̶ fiasco, my dudes. 😂

You use the word “notebook” and suddenly your work is “similar”?
That’s not image review — that’s SEO roulette.

And the funniest part?
At this point, contributors know the algorithm better than the moderators.
We reverse-engineer every step while someone just slaps “similar” on a submission and moves on.

Maybe instead of saying “Similar content already exists,” just be honest and say:
“We didn’t feel like checking.”
Would be more transparent, at least. 🙃

[jie_8284]

Similar shifts completely by just changing the colors. Here it seems to focus more on the pose.

bias

[Vikky1904]

Next time I’ll try submitting a fan... made out of pencils.
Let’s see if the algorithm handles THAT level of creativity. 

Maybe I’ll call it “Zen Breeze: Stationery Edition”. Might get accepted.
Or not. Because someone else once uploaded “Sharp Objects” and now all pencils are banned.

[daniellei4510]

"Because someone else once uploaded “Sharp Objects” and now all pencils are banned."

I just have five colored pencils images accepted, so maybe it's my fault. 🙂

[Vikky1904]

Five pencils? That’s five too many. The system is fragile.
Next stop — “Banned because it reminds someone of a ruler.”😭😄

[DaveFX]

"If the Adobe Stock library already holds over 800 million assets, maybe it’s time to start cleaning out the archive from 1949? " ....

personally, just my opinion, that a photo taken in "1949" has more value than a fake AI photo made today ... so i can understand why AI images/photo are hardly accepted ....

[daniellei4510]

I, too, am concerned that "similar" images are being compared to older photos, even from the Fotolio days,, which was founded, I believe, in November of 2004. I've never had experience with Fotolia, so I don't know if submissions were moderated. But moderated or not, in an effort to grow its database, I suspect the standards were not as high as they are today.

[Jill_C]

Storage is cheap, and I don't foresee Adobe implementing a program to purge old, unsold images. Adobe acquired the database from Fotolia in 2014 and it grew steadily, though not dramatically, until 2022 when they started accepting AI images. Now in just 3.5 years, AI is nearly half of all assets and still growing rapidly. 80-90% of the assets accepted each week are AI, so those will soon eclipse traditional stock images. 

I think Adobe should segregate and establish a separate database and unique reviewing methods for AI and go back to the old way of reviewing traditional non-AI images. It was reliable and understandable. Now the rejects for similars and quality issues, which are apparently handled by an algorithm, seem random and often don't make sense. Contributors of traditional images have been shunted aside and penalized by Adobe. 

[Abambo]

Contributors of traditional images have been shunted aside and penalized by Adobe. 

By @Jill_C

Despite the bad quality of the generative AI assets accepted to the database, it seems that generative AI is a big business for Adobe, or else, they would limit generative AI submissions. And until now, they did not consider developing the sticky exclude generative AI filter, so I suppose that their statistics show an advantage of not having this. 

[Jill_C]

Yes, I'm sure that they've prioritized AI because they see some upside commercial opportunity; however based on the "Undiscovered Content" filter on stock.adobe.com it seems that 356 million out of the 374 million AI assets in the database have NEVER been sold - a paltry 5% have been licensed. That's a lot of cost in terms of acquisition, storage, Moderator expense, etc. for what appears to be a modest gain.

If I do the same filtering on non-AI assets, the result is that 349 million out of 427 million have been sold - 18%. Perhaps AI will catch up eventually, but it has a long way to go. Meanwhile, ~24% of my non-AI images have sold, some quite regularly...

[Nancy OShea]

it seems that 356 million out of the 374 million AI assets in the database have NEVER been sold - a paltry 5% have been licensed.

By @Jill_C

==========

Stock is a numbers game. And 18.7 million sales is nothing to sneeze at.  Even by the most conservative profit margins, AI brings in more than enough to keep the lights on. 

[Jill_C]

Yes that's true, but for the individual AI Contributor knowing that, on average, only 5% of your contributions might be sold, isn't very motivating...

[daniellei4510]

I'm at 21%, averaging .78 cents per sale. But since the similarities issue especially, I rarely submit more than 8 to 10 assets per week on average. I have around 6 assets under review at the moment, and I'm good with that. Creating AI is expensive, and new AI contributors especially need to find a balance between profit and expense. Submitting to stock "because it's fun and I'm just happy if someone buys my stuff" is a feeling that can dissipate real fast when one is spending $20 a month for the service and realizing a dollar or two in the same amount of time...especially if it's on top of anything over a 50% rejection rate due to quality issues and similarities. To be honest, if I hadn't gotten into the game almost three years ago (which I wouldn't have done anyway without my prior experience as a photographer and graphic designer), I wouldn't have gotten involved in producing AI stock today.

[Jill_C]

21% sale rate seems quite impressive for AI! Nevertheless, as we've learned and counseled others, the AI stock business is certainly not a "get rich scheme". 

[zakir32208234z54q]

If you think it is related to keywords and possibly titles, so I'll say, there are billions of images with tags and titles here. So it will match one or the other. Where can we find so many unique tags and titles? Shall we create new and unique words? We work hard to create quality images. But It is being rejected as similar content, for which we can't find any specific reason, can't find any similar image. What type algorithom Adobe using!!

"Similar Content theory is total fiasco for us".

[daniellei4510]

It's just a theory. And even if correct, I never said it was the best way to go about weeding out similar content. But I have seen at least some evidence personally that putting common keywords last, rather than first, appears to help get assets accepted. Sometimes. And no, one wouldn't want to create new and unique words. But there are synonyms to common words. Of course, one can't overlook the obvious: if a contributor gets TOO creative with synonyms, what is a likelihood that those synonyms will be used in search phrases? Probably not very. As such, it's probably also important to still use the common ones somewhere in the list, preferably near the end. Again. Just a theory.

[zakir32208234z54q]

Not satisfactory, synonyms also very few. But we are in fear to upload new assets, they treat "similar content" as spam, and they can block or permanently close our account! We are captive. Nothing to do!!

[daniellei4510]

Well...there is something you can do. You can diversify and submit your rejections (or even your accepted assets) to other sites. We are hardly being held captive.

[Abambo]

We work hard to create quality images.

By @zakir32208234z54q

Who is we? Generative AI contributors submit to much in a short time, and a lot if not most of it id bad quality. The similar issue has probably been triggered because of those contributors who automatically contributed assets, automatically generated with very little variation. We've seen our share of them here.

But It is being rejected as similar content, for which we can't find any specific reason, can't find any similar image. What type algorithom Adobe using!!

"Similar Content theory is total fiasco for us".

By @zakir32208234z54q

My guess is that Adobe won't tell people what triggers the refusal so that they can't fool the test. 

[Neo-Arts]

It is something that Adobe need to consider.

[Abambo]

It is something that Adobe need to consider.
By @Neo-Arts

As long as Adobe is happy with what they are doing, they won't change. May be (or hopefully) they are tweaking their algorithm for the better (, silently). 

[jacquelingphoto2017]

Hi @Vikky1904 ,

We understand your frustration. I still maintain that the similar refusal is in reference to what you already submitted. Otherwise, 90% of us would end up with all similar refusals. With that said. Similar refusals are wide. It's more than having two images looking alike. If the subjects are close, it comes under similar, and only a limited number of such is allowed. I could better help you if I have access to your public profile and a few of the files refused.

Another thing, you are not addressing Adobe here, but fellow contributors. To address Adobe, you must use the "Contact us" at the bottom of your contributor page. 

Regards

Jacquelin

[Nancy OShea]

Most humans agree that AI is deeply flawed. Unfortunately, mega corporations are embracing it more & more because they think AI is more efficient than human labor.  And for some things, it is, provided it's adequately supervised by humans for accuracy.  

As an example, https://bsky.app/profile/nancy-oshea.bsky.social/post/3lv2kns6jns2g

If enough people send protests to Stock Contributor Support about it's failed AI moderation tool, maybe they'll be compelled to take a closer look at it.

• https://contributor.stock.adobe.com/contact

Until then, we have to suck it up or take our products to competing services. That's all we can do. 

[Vikky1904]

Let me sum up what I’ve learned today from two kind contributors:

1. If you upload pencils and a notebook, and then upload brushes and paper, the AI might flag it as “similar content.” Why? Because the layout is close enough. Doesn’t matter that a buyer looking for a pencil won’t want a brush — AI sees a shape, not logic.

⠀

2. “Similar content” isn’t just about visual duplicates — it also includes anything that feels too... cousin-like. So if your first-ever pencil image gets refused for similarity, don’t worry — it’s probably similar to your own invisible past.

⠀

3. Moderators can’t tell you why something was rejected — because they don’t know either. It’s AI. But you should still be thankful and smile, because it’s less competition, right?

⠀

4. And finally: if enough people complain, maybe Adobe will do something. Until then, “suck it up or move to another platform.” Beautiful customer service. Chef’s kiss.

---

Conclusion as a creator:

AI moderation is here. It’s flawed, it’s lazy, and it doesn’t care what’s inside your work — only how similar it thinks it looks.

And if you dare to ask why — someone will smile, call it “common,” and move on.

Thanks for the wisdom. I’ll go upload a rock and pray it doesn’t remind Adobe of a potato.