Home › AI & Tools › AI-Generated Images for Websites: When They Work and When
AI-Generated Images for Websites: When They Work and When They Give You Away
AI-generated images are a fast-selling and easy-to-use option. The apps are new, so the ethics and liability rules are still evolving, and non-designers notice that the outputs are useful enough. Use AI images to come up with a 404 screen, or placeholder portrait for a headshot on a page that is two weeks out of the client review cycle. If you are ready to publish, it may not be obvious that an image was AI-generated just by eye.
In this article
But handling AI-generated imagery wrong — trying to disguise it as real, presenting it as a person, place, or experience, or not labeling it properly — undercuts whatever tells a designer to spend time on a generated image in the first place: artistic honesty and client trust. That's how media belongs on the web.
When Generated Imagery is Acceptable
AI pictures should be used sparingly, and clearly indicate they have no documentary value. Unless they come with a direct license for unlimited commercial use, each image includes some functions, terms, datasets, outputs, and training artifacts — most commonly the aesthetic, proportions, framing, placement, settings, backgrounds, props, clothing, and performances generated by the shown or underlying persons, places, property, performances and misc. Just like photography licensing and profit sharing, the tool statement is often buried in the generated file or export options and the law around the contract between the designer and the generator is not stable.
What practical rules are there to create AI images while cutting copyright and liability for the designer?
Conceptual, illustrative, and non-representational art is ok. For Commonwealth University, that means “non-representational background, ethereal placeholders, icons, or spot illustrations. Abstract is OK, but specific features of their campus or campus life worked up in AI call for a different workflow, and require thinking about terms, local legal ownership and licensing rules, and permission from the underlying sources of images — or rough ideas and policy around those. The state of scholarship may change quickly on showing the campus as AI-generated, or showing characteristics associated with the campus or people associated with the campus.
Where It Breaks Trust
With AI, you have to ask: without context or disclosure, can this image be mistaken for a representation, for portraiture, or for a documentation of some context or thing the web designer is putting next to it, where it is placed on the page? Representation is already a tricky issue with AI.
So, generally, a time of using AI for images is to make an illustration that has the surface detail that a photograph or a still from video has, but none of the expectation that it shows what was there, that it is not imaginary, that it shows a real person, group of people, or identity of conventions and service, or that it is related to the tradition of accuracy, reporting, representation or artistic value related to photographs and video.
McGill University, for example, prohibits AI images in things like course materials, alumni profiles, biographical sketches or textbooks “under all circumstances.”
The reason for that kind of specific guideline goes to fundamental questions of credibility and trust in putting a photograph or video on the university page. Putting an illustration on a webpage is very different from skills and questions around posing, getting permission, negotiating fees, and disclosing the source of copyright. Most of the principles, workflows, and checks concern the old-school production of the document. A photographer may be a consultant and the aspects of photography blend into a hand-shadow. But a reader would not know the photo's origin without the person’s fees, permissions, terms of service, and subsequent usage arrangements. And the arrangement around the image may be very different than for text on the same page, or for diagrams, or the general look and feel and trademarking or copyrighting around the site.
Disclosure and Labeling
So the UI has to make it clear what the image is. The website is putting the image, with confidence, as advertising or description; on behalf of the owner, the reader makes decisions; the image makes a very strong impression on the website that may change the way the reader perceives what is being presented. What is the long-term sticking point for AI in this way?
Google says it’s a disclaimer. The institution sets it clearly and outright. The image may be a portrait or cover image, or a background Or someone may get the idea to use generated avatars as placeholder content while another designer puts the finishing touches on the real head shot. A visible label clearly states the image is AI-generated. This allows the designer/institution to present the image as conceptual rather than realistic, documentary or representational. — as much a label as disclaimer.
APSU's guidelines point to good practices to follow when disclosing AI images. The caption for the image should read “Image generated by AI.” This should be placed either below or beside the image. Additionally, for alt text, it is not necessary to repeat the disclosure. Instead, the alt text should accurately describe the visual content of the image to ensure accessibility. This highlights APSU's commitment to creating inclusive and informative content across their digital platforms.
This disclosure helps to set expectations and avoid misunderstandings. When a website visitor sees something that seems to be a photo on a university page representing facts, experiences, a location, or named people, that can raise expectations and raise a donor or publisher's "opinion" of a place or the website. Alerting the reader that this is original content, artificial art, and made up in this context can make sense to what really is new here.
Undisclosed AI art is also tricky for a website to describe, give access to, or comply with copyright. Until the terms of service are explored and expanded, and the liability for the trackbacks to training datasets is sorted out, most universities and affiliations have the same kind of AI guidelines which apply to other entities.
Policy and Compliance
The policy guidelines caution moving forward on generated images as if they are published as photographs, legally.
For example, using AI to create depictions of named faculty, staff, students or others is a breach of the considerations that in general lead to the close, detailed and extensive guidelines for use of photography in the usual college programs.
But when AI-generated images are used responsibly, as conceptual or illustrative elements, they can enhance the visual appeal of websites and digital content.
The Missing Craft Details
Some guidelines call for sticking to some minimal specific rules or caveats, firmware, or details more often like web performance sources would specify. This is why Commonwealth University has a guide to making it clear when AI-generated images are conceptual in nature and not intended to represent real individuals or experiences.
Aspect ratio, export size, and file format are specifically relevant for images used on university websites. This advice helps avoid performance issues and ensure optimal display.
Icon sets, spot illustrations, diagrams, and placeholder art are increasing in websites, but documentation around when to do what is still rare, unlike general uses of imagery made to seem to be back from cameras or screens.
Cleaning up before shipping, old-school style, means making sure there are no visual artifacts disturbing the design balance in an illustrative AI image, including personal traits that should not be visible, inadvertent watermarks, and other deal-breakers and OPS considerations.
Users of AI image tools should be well-versed in the terms and conditions of the AI system they are using. This includes commercial use, proper attribution, and any notable prohibitions on portraying real individuals or well-known entities.
Disclosure and Branding, For Identity and Trust
Running straight to an AI tool to dress up a page is convenient, and AI does a decent job for the designer in bringing the artistic, layout, color, shape, and graphic material. Like stock, AI is limited as far as editorial goals, story, and branding as a way to establish a theme but used right, stock and AI can fulfill this role.
So in branding a logo or brand, fragile images are worth the effort to label the source. Depicting actual products, trademarks, logos or other identifiable aspects of named places, shows identifiable images and legal characters in stricter or specific ways may require permissions not commonly achieved in AI.