PAN reader, Serban Enache Owner, Co-founder & CEO of creative stock photo agency Dreamstime, has penned an interesting article around AI which he has allowed us to reprint it here – enjoy the digest.
Stock photography vs AI-generated content – a preamble
Stock photography emerged mainly from the need for visual content when individuals and businesses could not photograph their own for financial or time-related reasons. Photographer-owned content was amassed by agencies and promoted for “stock,” allowing clients to search an existing database (first offline, then online) and select the exact images they were looking for. Stock photography content conveniently removed time concerns or friction between the client, the agency, and the photographer.
The transition from film to digital sped things up. The time required to find the right content was always crucial, just as it was essential to be safe after using this content. Due to quality demands, images soon began to be enhanced via Photoshop, and people were concerned about how honest the final photo was and how close it was to depict the actual subject (whether it was manipulating objects, people, or news events in editorial photographs). The rise of the still ongoing ‘Authenticity’ trend in stock photography engendered various styles, filters, and techniques, often resulting in purposely flawed imagery that captures lifelikeness. More than a handful of important campaigns are shot on film, some with harsh on-camera flash, others with visible actual or simulated film grain, some with poor lighting, slightly out of focus, anything that can make a statement of realness.
The introduction of generative AI visual content echoes the same concerns but with fewer answers and less safety regarding the ethical, creative, and legal sides.
The Ethical and the Legal
There are cases when you need to post images of your products. You post your imagery if you can photograph the products (on your own, hiring a photographer, getting the props, balancing the budget, and so on). If you can’t do that or don’t need it, you either go to the manufacturer or a stock photography website and select a photo as close as possible to what you need. Whether you manipulate it afterward with additional software, the authors are compensated for their work. Why do they need that? Simply because they become your supplier, you acquired something from them (and they have their costs: education, hardware, software, time, travel, props, models, etc). The copyright law protects them.
This does not happen for AI, where models are trained on billions of images, and the authors have yet to consent to this training. Think about a powerful software tool such as Photoshop accessing all images on the internet in the background, including yours, the reader. Most artists and content creators giving feedback to Dreamstime on this matter do not or would not consent to have their work used for AI training; not knowing the usage, not being compensated, and not being transparent would be their most common comments. That is why many AI platforms got sued by agencies or photographers, and there are famous cases where AI replicated almost exactly the original image. To this date, AI platforms lack the compensation structure stock photography has, leading to legal challenges due to allegedly unauthorized replications of original photos.
Plus, there is a big question: Have the people in the replicated or trained-on image agreed to this ulterior usage? In stock photography, all models will sign a model release document.
Whether it is safe or wise to use AI vs stock photography relies on many facts that the designer may need to learn. Stock agencies employ hundreds of criteria in selecting their images to be able to indemnify their clients (something that doesn’t happen for generative AI platforms). Choosing AI images without expertise in stock photography risks exposing oneself to significant legal and ethical challenges, as opposed to the organized process of selecting images from established libraries, where you can also speak to a human about your content selection.
It is not only about using the original images without consent; it is about featuring people or models without their consent, featuring trademarked properties without consent, infringing authorship, privacy, or commercial rights of others.
Where to, Creativity?
Quality commercial photos require years of experience and resources invested in quality equipment and know-how. The AI only takes advantage of the results by training AI systems for better output. The lack of innovation and human imagination will eventually limit artistic manifestations and freedom of expression. AI-generated images often need more cultural context nuances; they often look like mass-produced, unnatural, and stylistically inconsistent visuals. Take our last image assignment, for instance, where we had tons of AI-generated content that ultimately missed out on creative diversity.
Creativity or the lack of
It is easier to wake up once and produce one hundred perfect sunrises than to wake up early one hundred times to produce one. But when we give up on photographed or drawn imagery, we declare that everything has already been said and done, and we can now remix the old ideas repeatedly. This is the opposite way of progress. Innovation may come a bit slow indeed, but HDR images, intentional camera movement, focus stacking, light painting, tilt-shift, and all the other techniques and composition rules that seem easy have now been discovered by real people over quite several years. AI is only taking advantage of these innovations for now but is not innovating anything. It is simply a tool. What will happen to innovation when most creatives will no longer discover new techniques or innovate in any way? We loved how Freddy Mercury used to sing, and his unique style, among thousands of different styles, made it even more enjoyable. But an influx of AI-generated Freddie Mercury songs will only dilute his unique style over and over again.
Authenticity
We have seen it thousands of times: AI images automatically generate fake content, which is a big concern, especially in the case of journalism or historical documentaries. Reality is not about perfect colors, perfect outlines, or utopian-looking landscapes. While AI may replicate techniques, real individuals with imaging culture are essential for authentically looking campaigns. AI imagery may prove valuable for sensitive topics such as substance abuse, where obtaining permission from real models is difficult; however, conveying authenticity through AI remains challenging, given societal tendencies to skip or scroll over perceived artificial content. Remember that exaggerated or dishonest “photoshopping” is still perceived as a frowned-upon practice.
Stock photography vs AI-generated content – A SEO perspective
Would you use AI-generated or stock photography for your website? Considering the SEO implications, the answer is simple: authenticity, uniqueness, aesthetic value, and safety are pivotal for rankings. Original images often hold more value in SEO due to their relevance and ability to enhance user engagement and brand trust.
Alt Text and Metadata – these indicators, especially metadata, contain details related to the originality and author of the image and the equipment used to take the photo. Although these values can be edited, an image’s authenticity remains essential in a real photo, potentially positively affecting SEO.
User Engagement – It is known that this is a ranking factor for search engines, and natural and impactful images can influence values. AI-generated images may resonate differently, especially if they appear less authentic or relevant to the user. AI-generated images might only sometimes perfectly align with the specific nuances or the exact context of the content, which could impact their effectiveness in supporting SEO.
Brand Perception and Trust – Original images can enhance brand perception by showcasing genuine products, services, or concepts. This authenticity builds user trust, indirectly benefiting SEO through increased engagement and sharing. AI-generated images convey a different level of trust or distrust, especially if users perceive them as artificial or less authentic. Or even fake, when AI images will fake certain products or people endorsing them.
Copyright and Legal Issues – The original images, owned or licensed, can be used without the risk of generating legal issues. AI-generated images can become complicated, especially concerning the original data used to train the AI. Copyright issues can negatively impact SEO if they lead to legal challenges or content removal.
As we have often heard lately, AI is about progress and advancement; rejecting it entirely is not the answer. Dreamstime is accepting AI-generated content under specific terms. However, this type of content has little to no regulatory system for now. What we have today are, at best, in-house or ad-hoc compensation systems for using copyrighted material to generate or train AIs or the publicly expressed intention to work with artists for mutually agreeable solutions. Before using miraculous tools that promise perfect visuals, just a few prompts away, legally carefree and financially practical, consider their benefits or lack for your website, company, brand, business, and yourself. Hoping for future regulations, more transparency, and ethical practices that will empower content creators and give more control over their works and how these are being used AI-wise, we welcome AI technology as a tool.And just as with any tool, building AI platforms or using them requires education and utmost responsibility.
Thanks Serban, great write up! Keep up with Dreamstime via their blog posts here.
View all Dreamstime news on PAN