What I have seen on my own sites
This starts from my own observation, not a study. I publish a lot of images across my sites: diagrams, social graphics, covers and illustrations. Over time one pattern became hard to ignore.
"AI-generated images are not getting ranked in Google Images. The images made digitally using Figma and similar tools are getting indexed, except for the AI ones, no matter the posting frequency or engagement." Lohith Lalesh
That is what I have seen on my own sites, with my own mix of pages and topics. It is not a controlled experiment, and I am not claiming Google penalises every AI image everywhere. Other causes are possible.
What Google has said officially
Google has not published any rule that demotes AI-generated images. Its February 2023 position on AI content is that "appropriate use of AI or automation is not against our guidelines". Its current generative AI guidance warns that generating many pages without adding value "may violate Google's spam policy on scaled content abuse", and the spam policies cover images as well as pages.
Asked in 2025 whether AI images hurt SEO, Gary Illyes of Google said, as reported by Search Engine Journal, "AI generated image doesn't impact the SEO. Not direct". Note the "not direct". That leaves room for indirect effects, such as an image simply being less useful than the alternatives.
Where Google is firm is labelling. Google Merchant Center requires that "all AI-generated images must contain the IPTC DigitalSourceType trainedAlgorithmicMedia tag" and tells merchants not to remove it. Google Images also reads that field, alongside creator, credit line and copyright notice. If an image carries C2PA Content Credentials from a qualifying signer, Google may show whether it was made or edited with AI in the "About this image" feature. Google DeepMind's SynthID adds an invisible watermark to images from Google's own generative tools. Google has not said any of these labels affect ranking.
Do not strip the metadata
Some guides suggest deleting AI metadata so Google cannot tell. I would not. Merchant Center forbids it for product images, and Google's guidance encourages the labelling. If you want an image to perform better, change the image, not its label.
What other people have reported
I am not the only one. In February 2025, Search Engine Roundtable covered a site owner whose Google Images traffic dropped in late January while web traffic held steady, on a site using AI images. Others with AI image sites told him they saw drops on the same date, and SEO consultant Kane Jamison said Google had "multiple Google Images updates since the fall that appear to be AI-image related".
The same article has the counterexample: Barry Schwartz uses AI hero images on almost all his posts and saw no drop. Google did not comment. The evidence is anecdotal and mixed, and I found no published controlled test.
Why it might happen: confirmed facts and theories
What Google has documented, and what is my reasoning:
SEO and AI search · Dubai
Possible explanations, and how solid each one is
| Explanation | What we actually know | Status |
|---|---|---|
| Google reads AI labels | Google extracts IPTC Digital Source Type and C2PA data and may show it in About this image | Confirmed it reads them; no evidence they affect ranking |
| Images need context | Google uses surrounding text, captions, filenames and alt text | Confirmed |
| Quality and relevance matter | Google favours sharp, relevant images on pages about the same subject | Confirmed |
| AI images look alike | Many AI images share a style and composition, so they add little new | Theory (mine and others) |
| Different use, different result | My Figma graphics show something specific; AI images are often decorative | Theory: could explain my pattern with no AI-specific signal |
| Google is cleaning up AI images in some results | Reported by practitioners after early 2025 traffic drops | Unconfirmed. Google has said nothing |
I take the fifth row most seriously. My Figma graphics usually explain something nobody else has drawn. My AI images often just set a mood. A generic AI image of "a team in a Dubai office" competes with countless similar images; a diagram of a specific process has far less competition. Either way, the fix is the same.
How to test this on your own site
Before you replace anything, measure. Search Console has no report listing every image Google has indexed, so you piece it together. Compare similar pages; don't try to prove a universal rule.
SEO and AI search · Dubai
A simple check you can run in an afternoon
Open the Image search type
In Search Console Performance, set Search type to Image and a long date range. This shows traffic from the Google Images tab only.
Sort pages by impressions
Open the Pages tab and note which pages earn image impressions, and whether each page's image is AI-generated, handmade or a photo.
Inspect a few URLs
Use URL Inspection on pages with and without image impressions. Check the page is indexed and that View crawled page shows a real img tag.
Search Google Images directly
Run a site: search for your domain in the Images tab. It is a rough sample.
Compare like with like
Pair pages of similar topic and age, one with an AI image and one with a handmade graphic, both with good filenames, alt text and captions. If only the AI images are missing, that is a signal worth acting on.
Change one thing and wait
Replace the AI image on a few pages only and watch the Image report for several weeks. Changing filenames, alt text and images at once tells you nothing.
Making any image indexable
Most images I see on UAE business sites fail on basics long before AI is the issue. From Google's image guidance (my technical SEO checklist covers the rest of the site):
- Use a real <img> element with a src. Google says it doesn't index CSS background images.
- Name files descriptively, like marina-apartment-floor-plan.webp, not IMG00023.jpg. Translate filenames for Arabic pages.
- Write alt text that describes the image in context. Google calls it the most important image metadata, and warns that keyword stuffing can be treated as spam.
- Place the image near the text that explains it, and add a caption when it helps.
- Use supported formats (JPEG, PNG, WebP, AVIF, SVG, GIF), keep files small, and lazy-load so images load when they enter the viewport: Google does not scroll or click.
- List important images in an image sitemap, especially if they sit on a CDN.
- Add creator, credit and copyright through IPTC metadata or structured data, and keep any AI label your tool adds.
When to use AI images, and when not to
Many Dubai businesses sell something people want to see for real: an apartment, a dish, a clinic, a fit-out. AI images are the wrong tool for anything that represents the product. A buyer who finds a generated version of your building stops trusting the rest of the page.
Side by side
Where each kind of image belongs
Use real photos or handmade designs
- Your product, property, food or premises
- Diagrams, process charts and data you made
- Anything a customer will judge you on
AI is fine (labelled)
- Background textures and abstract headers
- Concept sketches you then redraw or edit
- Decorative images that are not the product
The middle ground is where I spend most of my time: AI to explore an idea, then the final piece built in Figma with my own layout, real data and the brand's colours. That is a new, original image, and IPTC has an honest label for mixed work (compositeWithTrainedAlgorithmicMedia) if AI elements remain. It is the same principle as what Google's policies say about AI content: the tool is fine, missing value is the problem.
Generic images are interchangeable, and interchangeable content struggles in any search system. That is the theme of why companies are bringing human writers back and of information gain. Auditing images and pages is part of my SEO work in Dubai.