In the fast-evolving world of generative AI, a new survey from Artificial Analysis offers a timely glimpse into adoption trends - just before OpenAI's Sora 2 redefined video creation. Polling 1,000 individuals and organizations worldwide, the study (conducted prior to Sora 2's September 2025 rollout) reveals robust enthusiasm for image generation, a more cautious embrace of video tools, and clear preferences shaping the market. As AI continues to blur the lines between creativity and computation, these insights highlight where the technology stands today and what might come next.
Widespread Adoption, with Images Leading the Charge
The survey underscores generative AI's penetration into everyday workflows. A staggering 89% of individual users reported leveraging AI for image generation, compared to a more modest 57% among organizations.
This gap suggests consumers are experimenting freely, while businesses are integrating tools more selectively, perhaps due to compliance or scalability concerns.
Video generation, while promising, lags behind: Only 62% of users and 32% of organizations are actively using it. The disparity isn't surprising - video demands more computational power and refinement, and pre-Sora 2 tools often struggled with coherence and length. Still, the numbers signal growing interest, especially as hardware improves and models like Google's Veo and Kling mature.
Top Models Dominating the Landscape
When it comes to favorites, the field is led by familiar giants. For image generation, Google Gemini tops the list at 74% usage, followed closely by OpenAI's DALL-E integrations at 64%, and Black Forest Labs' FLUX at 37%. Gemini's edge likely stems from its seamless integration into Google's ecosystem, making it accessible for quick sketches or professional edits.
Video tools show a similar concentration: Google's Veo claims 69% of the share, outpacing Kling AI at 48% and MiniMax's Hailuo at 35%. Veo's strengths in realistic motion and prompt adherence appear to resonate, positioning Google as a frontrunner in both modalities. Notably absent from the top spots? Early video experiments from Meta or Runway, hinting at a preference for polished, enterprise-ready outputs.
Key Criteria: Quality First, Cost Close Behind
Respondents were unequivocal on what drives their choices: Quality reigns supreme, cited by 82% as the primary factor. Metrics like photorealism, prompt fidelity, and artifact-free results are non-negotiable, especially for creative professionals.
Cost emerges as a close second, particularly for API-driven workflows. Organizations using APIs for bulk generation (e.g., in marketing or product design) emphasized predictable pricing, with many opting for tiered plans to avoid surprises. For casual users, free tiers in apps suffice, but scalability tips the scales toward paid options when volume ramps up.
Apps vs. APIs: Where the Action Happens
Accessibility is key, and the survey reveals a clear divide in delivery methods. Individual users overwhelmingly favor applications (86%), drawn to intuitive interfaces that require no coding. Here, ChatGPT dominates at 69% - its conversational style makes ideation feel effortless - while Google Gemini follows at 64%, thanks to its mobile-first design.
Organizations, however, split evenly at 50/50 between apps and APIs. The API preference suits automated pipelines, like auto-generating thumbnails for e-commerce. Among apps, Midjourney stands out for business users at around 40% adoption, likely due to its pioneering paid model with commercial rights - essential for IP-sensitive industries. Adobe's Firefly suite also shines, integrating seamlessly into creative suites like Photoshop, and capturing 35% of organizational use.
ROI Realities: Quick Wins, But Training Lags
The financial upside is encouraging: 65% of organizations expect to recoup investments in generative AI within a year, with 34% already reporting net profits. Gains come from time savings - think slashing design cycles from days to hours - and novel applications, like personalized ad visuals boosting engagement by 20-30%.
Yet, implementation reveals gaps. While 43% have woven AI tools into core processes (e.g., prototyping or content ideation), only 33% invest in staff training or new roles. This "plug-and-play" mindset risks underutilization; without upskilling, teams may stick to surface-level uses, missing deeper efficiencies like custom model fine-tuning.
Also read:
- NVIDIA Leads Open-Source AI Revolution as Chinese Labs Close In
- Finland Turns Mining Farm Heat into Home Heating Solution
- Anthropic Uncovers Alarming Vulnerability in Language Model Training
Looking Ahead: From Hype to Habit
This Artificial Analysis survey captures generative AI at a pivotal moment - pre-Sora 2, when tools were powerful but not yet transformative. With images nearly ubiquitous and videos gaining ground, the stage is set for broader disruption in media, education, and beyond. As costs drop and quality soars (hello, Sora 2's minute-long clips with audio), expect video adoption to catch up fast.
For now, the message is clear: Prioritize quality, embrace apps for speed, and don't skimp on training to unlock true ROI. In a world where anyone can conjure visuals from words, the real advantage lies in smart integration.
