Observed Signal · Jun 2, 2025 · Technical Release · Source: OnlineMarketing.de · Impact: 2/5 · Sentiment: Neutral
ChatGPT Tests A/B Image Visuals
OpenAI is testing an A/B feedback feature for image generation inside ChatGPT’s desktop version. The feature lets users choose between two automatically generated image variants, using their preference to inform model training and potentially improve output quality and transparency. The test is in an early phase and currently visible only to a selected subset of users; OnlineMarketing.de did not have access at publication. The article frames this as part of a broader shift toward participatory, multimodal AI, where user feedback helps steer how ChatGPT learns and refines visual outputs. It also notes ongoing debates about training data provenance and consent, citing Meta’s EU data-use practices for AI training as a related context for data-ethics discussions.
OpenAI feature in early testing with moderate industry impact.
Track OpenAI Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- OpenAI is testing an A/B feedback feature for image generation in ChatGPT's desktop version.
- The feature enables users to select between two automatically generated image variants to influence model training.
- The test is in an early phase and visible only to selected users, not yet broadly available.
- The article describes this as part of a shift toward participatory, multimodal AI where user input informs training.
- The piece references Meta's EU AI training data practices as context for debates on data provenance.
Connected Companies & Entities
4 Entities mappedRelated Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
OpenAI Unveils Enhanced ChatGPT Images Update
OpenAI released an enhanced ChatGPT Images update that makes edits more precise, speeds up generation by up to four times, and preserves essential details such as lighting, composition, and likeness across edits. The update eliminates the need to reset images and provides concrete prompting tips, with Sam Altman illustrating concepts through examples. The expanded GPT-Image-1.5 model is available in ChatGPT for all users and via API, with costs for creating and editing images reduced by about 20%. A new in-dialog image workspace adds a dedicated area for styles and inspirations, and ChatGPT guides users with prompts to test the improved tool. Adobe’s Power Tools—Photoshop, Express, and Acrobat—are integrated into ChatGPT free at launch, while native Figma integration supports generating diagrams, presentations, and marketing assets from ongoing conversations. OpenAI cautions that multilingual content and very complex scenes may still misrepresent, with current image quality around 70%, underscoring the need for manual review. Access for Business/Enterprise comes later.
OpenAI launches ChatGPT Images 2.0, better text generation
OpenAI announced ChatGPT Images 2.0, a new image-generation model that significantly improves rendering of small text, iconography, UI elements and dense compositions at up to 2K resolution. The company says Images 2.0 includes “thinking capabilities” that let it search the web, produce multiple images from one prompt, and double-check outputs, enabling uses such as marketing assets in multiple sizes and multi‑paneled comics. OpenAI did not disclose the underlying architecture; researchers contrast diffusion approaches with autoregressive mechanisms for improved text generation in images. Images 2.0 reportedly handles non‑Latin scripts (Japanese, Korean, Hindi, Bengali) better, has a knowledge cutoff of December 2025, and will be available to all ChatGPT and Codex users (with paid tiers offering higher-quality outputs). OpenAI will also offer a gpt-image-2 API with pricing based on quality and resolution. The article was published April 21, 2026.
OpenAI Launches ChatGPT Images 2.5 with New Editing Features
OpenAI has launched ChatGPT Images 2.5, a next-generation image model with enhanced detail, realistic lighting, and 50% lower latency, along with two globally available shopping features: a 'Try On' tool for virtual clothing trials and a similar-item search from screenshots. Users can upload selfies or full-body photos to visualize products, and find purchasable items from screenshots or celebrity outfits. Independent testing showed high realism for simple items but inaccuracies for complex garments, including altered body proportions and design distortions. Safety measures include C2PA metadata and watermarking. Developers can access two API models: Flare (general) and Sunburst (premium). These moves position OpenAI against Google and Pinterest in the shopping arena.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.