I've encountered this myself when using Canva's Background Remover on my iPad. I wasn't generating an image from scratch or changing the subject. I was simply removing the background from an existing photo.
My first workaround was to take a screenshot of the finished image and upload that instead. It worked as an alternative, but it wasn't ideal because I wanted to preserve the original image quality.
Then I was using my laptop, created the image in Canva web version and uploaded the image from the Instagram desktop app instead. It was then that I realised that I could export the image directly and upload it to Instagram without encountering the same problem of it being identified with the AI tag.
What Is Happening With Canva Background Remover?
Users have reported that images edited with Canva's Background Remover are sometimes labelled as AI content on Instagram. And the issue is inconsistent, it does not appear to affect every user or every upload, and the exact cause is not confirmed in every case.
Canva's Background Remover help page explains the feature and offers guidance for users who encounter problems. The tool uses AI to identify a subject and separate it from its background.
That raises a question worth exploring: does using an AI-powered editing feature mean the resulting image should be treated as AI-generated?
For me, the answer depends on what the tool actually does to the image.
Background Removal Is Not a New Concept
Before automated background removers became widely accessible, we could use Photoshop to select a subject, carefully trace its edges and remove the background manually. The process required more time and editing skills, but the objective was the same.
Canva has made that task much easier by automating much of the work. So instead of spending time selecting individual areas, we can often remove a background with a single click.
Technically, the feature uses AI, but I see it as an automated version of an established editing technique. The software identifies the subject and does the tedious work for us; it does not necessarily create anything new.
This is where I think the distinction between assistive and generative AI matters. Generative AI can create new imagery or synthesise visual details, while an assistive editing tool can automate a task that previously required manual effort.
Of course, background removal can sometimes introduce small artefacts around edges. Other AI-powered editing features may also change the image more substantially. But those are different situations from simply isolating an existing subject, and even when you manually do it there are imperfections.
So Why Does Instagram Apply an AI Label?
Instagram does not necessarily determine an image's history by examining its appearance alone. Meta has described using industry-standard indicators associated with AI-generated content, including certain metadata and Content Credentials.
Metadata is information attached to a digital file. Depending on the format and the software used, it can contain details about an image's origin, editing history or processing.
If Instagram detects a recognised AI-related signal, that information may influence whether an AI label appears. However, Meta's detection system is not fully transparent, and it would be inaccurate to assume that metadata explains every reported case involving Canva.
Different export workflows can also preserve or change file information. That is one possible explanation for why two versions of an image might be handled differently, but it has not been confirmed as the cause of my experience.
The important point is that detecting an AI-related signal does not necessarily tell the whole story of how an image was created.
Two Workarounds I Tried
1. Taking a Screenshot on My iPad
When I first encountered the issue, I took a screenshot of my finished Canva image and uploaded that to Instagram instead of the original.
A screenshot creates a new image file. Thus, it may not retain the original export's embedded metadata or editing history, which could make a difference if a particular metadata signal contributed to the label.
The downside is image quality. Screenshots can have different dimensions and resolution from the original export, and unwanted interface elements or cropping can become a problem.
While it was a workaround I could use, it was not one I wanted to rely on whenever I prepared an image for social media.
2. Exporting Through Canva's Desktop App
This approach worked better for me. After experiencing the issue on my iPad, I exported the image using Canva's web version and uploaded the resulting file to Instagram via the desktop app. And it was at that moment that I realised, I did not encounter the same AI labelling problem.
Now, I can use the exported image directly without taking an extra screenshot, which made the process much more convenient. The only downside is I have to start up my laptop.
Is the AI Label Fair?
This is where I question the way AI labelling is being applied.
If I take a photograph of a mechanical keyboard for a review and remove the background to create a cleaner composition, the photograph still documents a real product. I chose the subject, took the picture and decided how to present it.
An AI Content label may lead viewers to assume that the image or the product itself was generated, even when the edit was limited to background removal. That can create a misleading impression of the creator's work.
I am not suggesting that assistive AI should never be disclosed. Rather, I think there is a difference between acknowledging that a tool uses AI and labelling the resulting image in a way that implies it was generated by AI.
Platforms could make these labels more useful by distinguishing routine AI-assisted editing from substantial generative changes. They should also provide clearer explanations and accessible ways for users to report incorrect labels.
Transparency is valuable, but it works best when the information helps people understand what actually happened to an image.
My Takeaway
What bothers me is that a routine editing task can apparently create confusion about the nature of the finished image. Having to find workarounds creates unnecessary work and wastes time... Background removal existed long before today's generative AI tools, and automating that task does not automatically make the original photograph AI-generated.
I would rather see platforms explain what their labels mean and distinguish between different kinds of AI use, instead of leaving creators to guess why an ordinary edit has been flagged.
Posts
0 Comments