Welcome to Prompt & Circumstance – your weekly deep dive into the evolving world of Generative AI and its impact on marketing and communications.   

Like this newsletter? Share this with colleagues who would find it useful. They’ll thank you later 😊. 

The Studio Ghibli Backlash, Claude’s Catch-Up, and ChatGPT’s Enterprise Play

This week’s edition is a Gen AI feature roundup! From image generation to search and data files, we’re breaking down the latest model releases and upgrades and why they matter. 

And yes, we had to try the ChatGPT image generator trend ourselves. We used ChatGPT’s new image creator to illustrate this newsletter.

Let’s get into it :). 

Image Generation Goes Viral, but at What Cost? 

In case you somehow missed it, OpenAI released its new image gerator, a significant improvement over its previous iteration, DALLE-3. Users can now type a prompt and click the new “create image” button, generating visuals without leaving the chat. It’s fast, intuitive, and supports inpainting, which means you can edit your photo further and re-render with new instructions, all in one seamless flow. The result? A smoother, more creative user experience with better control over visual output and a faster path from idea to image.

Since it’s release, timelines have been flooded with AI-generated portraits, illustrations, and concept art. From animation aesthetics to hyper-polished LinkedIn headshots. But, one style dominated the feed: Studio Ghibli. 

The Ghibli-style boom turned vacation selfies, political ads, and popular movie scenes into dreamy, pastel-toned animations. While the trend captured the internet’s imagination, it also resurfaced a long-running tension around copyright and creator rights and ethics.

Some artists and fans raised concerns about replicating Studio Ghibli’s iconic look. Its co-founder Hayao Miyazaki, known for Spirited Away, has famously called AI-generated art “an insult to life itself”. That quote made the rounds online, even as the majority of users leaned into the fun. It’s a reminder that as Gen AI gets more powerful and more playful, the conversation around creative ownership is only getting louder. 

Critics noted that the use of a deeply human, hand-crafted animation style as a mass-produced meme filter was more than ironic. It felt to many like a distortion of everything Ghibli stands for. “Imagine being Miyazaki… and seeing it get sloppified by linear algebra”, one post read. Others pointed to the environmental hypocrisy of generating Ghibli-style content using energy-intensive models when the studio’s films stress harmony with nature.   

Despite the controversy, the feature has fueled record user growth. OpenAI added over one million new users in a single hour following the update. App downloads spiked, weekly active users crossed 150 million, and in-app purchases hit all-time highs. The viral appeal is clear, but so is the tension between accessibility and artistic integrity. 

For brand teams, the lesson here is caution  

AI visuals are undeniably powerful. But borrowing visual styles without regard for their cultural or ethical weight can backfire. If you're using AI-generated imagery to tap into nostalgic or beloved aesthetics, it's worth asking whether this is an homage or appropriation. 

Claude (Finally) Adds Web Search 

After months of user requests and, with many giving up hope, Claude, Anthropic’s Generative AI, now supports web search. But there’s a catch, the feature is currently only available to paid users in the U.S, which limits its accessibility for now. 

For those who do have access, web search can be toggled on in settings where Claude will then pull in recent sources and display citations directly in its responses. For teams tracking fast-moving news or competitive insights, it’s a welcome addition. 

That said, many have pointed out how overdue this is. ChatGPT introduced browsing nearly a year ago, and tools like Perplexity have built their identity around real-time search. Luckily, some Claude users have praised its clean user interface and citation accuracy, while others question whether the limited rollout is enough to make up lost ground. 

Where Claude may lag in timing, it hopes to win on trust

According to Anthropic, the delay was intentional. The company has taken a more conservative approach to feature releases, citing safety and reliability as top priorities, especially when surfacing real-time information. Unlike other models that pull from the web more freely, Claude emphasizes source clarity and gives users direct links, helping mitigate misinformation and offering a safer way to navigate the often-messy intersection of web and AI. 

This trust-centric approach reflects a much bigger play. The company recently raised $3.5 billion and is positioning itself as an all-purpose assistant for enterprises. Voice features are in the works, and the company has upgraded developer tools and APIs for business-scale adoption. But the introduction of these features still lags OpenAI. 

Until now, many in the marketing and comms world have viewed Claude as the best writer among the leading Gen AI models, praised for its stylistic polish, clarity, and tone. But without real-time search, its usefulness for content rooted in current events or recent data was limited. This new update may change that.  

It may be enough to get teams to take another look at Anthropic’s offerings, especially as the company simplifies model selection, via 3.7 Sonnet, and continues refining its Claude model lineup. 

So, is Claude worth using over other tools?  

That likely depends on your priorities. If transparency and citations are top of your list (and you’re in the U.S. on a paid plan), Claude may be worth a try, particularly given its strengths in writing and newly added browsing capabilities.  

However, ChatGPT still leads the pack when it comes to overall features and capabilities. From multimodal input and memory to advanced reasoning tools and broader global availability, it remains the most well-rounded platform for marketing and comms teams who need versatility. Tools like Perplexity also continue to stand out for real-time search and research-driven use cases. 

As for what’s next, Claude says it will bring web search to free users and additional countries "soon". But considering how long it took to roll this out, "soon" might be a bit optimistic.

ChatGPT Knowledge Management

Aside from the new image generator update, OpenAI has also started rolling out one of its most requested enterprise features: internal knowledge integration for ChatGPT Teams. On March 27, OpenAI announced that users can now connect ChatGPT to their company’s Google Drive, allowing the model to search, reference, and synthesize internal documents securely and in real time. 

Open AI is positioning this capability as a foundational step in turning ChatGPT into the AI hub for enterprise users. The model not only finds relevant files, but it learns your organization’s shorthand, from acronyms and project nicknames to document structures and roles. It also respects user permissions, customizing responses based on who’s asking. 

Of course, data privacy remains a key concern 

ChatGPT Team chat’s are not used to traing OpenAI’s models, and responses remain private and permission-aware, meaning the model will only accesses files the user is authorized to see. Still, what we hear consistently from marketing and communcations teams is they don’t trust ChatGPT with their data and are wary of sharing information.

Despite these concerns, OpenAI is already planning next steps. Connectors for tools like Notion, Asana, Salesforce, and other platforms are on the roadmap. The goal is to build a broad set of integrations across the enterprise tech stack, from project management to CRMs to data platforms, cementing ChatGPT as a knowledge layer that sits on top of your business tools. 

Gemini vs. Qwen: The Model Wars Continue 

The latest model updates from Google and Alibaba are less about flash and more about strategy. Google’s Gemini 2.5 Pro emphasizes reasoning, scale, and integration. Built with a million-token context window, it delivers strong results across technical benchmarks. This includes impressive performance on reasoning and coding tasks, and internal evaluations, suggesting it can generate responses indistinguishable from humans in certain conversational settings. Already available in Google AI Studio and the Gemini app, the model reflects Google’s ongoing effort to make Gemini central to its productivity suite. 

Meanwhile, Alibaba’s Qwen2.5-Omni-7B takes a different approach. The model is multimodal, open source, and optimized for deployment on devices with limited processing power, like mobile phones, tablets, and edge hardware. Despite its compact size at just 7 billion parameters, it can process text, images, audio, and video in real time, making it a standout in the race to deliver high-performance AI in low-resource environments. It’s part of a larger pattern in China. After the rise of DeepSeek, tech companies are racing to release open models that support localized, cost-effective AI development.   

The divergence is important.  

Gemini is infrastructure. It is deeply integrated with Google’s tools and aimed at scale and enterprise use.  

Qwen is reach. It is lightweight, accessible, and adaptable to different environments.  

Both signal that the AI ecosystem is continuing to fragment. This again reminds us that marketing and comms teams will be unlikely to use one model from one company, but rather multiple models based on the use case and processing power required.  

That’s all for this week, folks! 

Whether it’s image generators, search upgrades, or model megastructures, the AI landscape is evolving faster than your last brand refresh.

If you haven’t subscribed yet, now’s the time. Sign up, share with a colleague—because in the world of AI, sharing is caring.