TL;DR
- A Google executive asked Gemini users on X what surprised them that the app still couldn’t work well.
- Workspace reliability, better tool calls, and better chat organization topped the list of comments.
- Google says many fixes are being made, but the guardrails around celebrity likeness remain as they are.
The AI race is moving so fast right now that it can be hard to keep up. OpenAI is making headlines today with the launch of GPT-5.6 and ChatGPT workBut this week a Google executive is taking a more feedback-focused approach and taking to social media to ask his audience a blunt question: What are Geminis still bad at?
Josh Woodward, whose X profile includes him as vice president of Google, Google Labs, Gemini App and Google AI Studio, asked users yesterday What surprised them was that Gemini still couldn’t get it right and “should have figured it out a long time ago.” Asking the Internet anything can be a bit of a Pandora’s box, and the Internet is duly obliged. After more than 1,700 responses, Woodward has now shared the top 10 requests for the feedbackalong with Google’s position in each one.
Does Gemini need Siri-style voice customization?
88 votes
The clear number one complaint was that Google Workspace integrations need to work more reliably. That’s a big deal, given that Google continues to pitch Gemini as something that can work in apps like Gmail, Docs, Slides, Keep, and Tasks, especially with its newest version. Gemini Spark agent. Woodward said Google agrees on this point, adding that some improvements were recently introduced to Gemini Spark, but that reliability should be “much better across the application.” It sounds encouraging, but it’s also the most diplomatic response.
Second on the list was the most trusted tool call, which Woodward also said Google “totally” agrees with. He didn’t give a precise timeline, but said users should expect “notable gains soon.” Third was better organization of projects and folders for chats. Google apparently sees Notebooks as a step in the right direction, but Woodward said they are “not enough” and the company will reconsider the experience.
Deep Research also attracted attention, with users requesting to export reports to NotebookLM and switch between Deep Research mode and Flash or Pro models in the same chat. Woodward said he was “not aware of these weak points,” which is an honest answer and probably the purpose of the exercise. He called both suggestions good ideas and said Google would “figure something out.” Removing the Nano Banana watermark is also being discussed, although Woodward said Google has to navigate AI rules that vary by country.
Several other solutions seem to be more squarely on Google’s to-do list. Google is apparently working to expand MCPs and custom skills after rolling out initial support in Gemini Spark, while doing message editing and prioritizing voice dictation accuracy – Woodwood plans to dig deeper into specific user responses on that front. As for the mobile app scrolling bugs, Woodward confirmed that they were being fixed.
The only request that is not receiving movement for now is to save railings with images of celebrities. Woodward said Google plans to keep them as is, so deepfake creators may prefer to stick with other models for now.
Thank you for being part of our community. Read our Comment Policy before publishing.





