Google invited blunt feedback from Gemini users. The response poured in fast. More than 1,700 replies landed after a top executive asked what the AI still handled poorly. What emerged was a clear list of pain points. Reliability in Workspace apps topped it. Tool calling came next. Chat organization followed close behind.
The executive, Josh Woodward, vice president for Google Labs, the Gemini app and Google AI Studio, put the question directly on X. He wanted to know what should have been fixed long ago. Users answered with votes and specifics. Android Authority captured the top requests and Google’s replies in a detailed report. https://www.androidauthority.com/top-gemini-users-improvement-requests-3685987/
Reliability across Gmail, Docs, Slides, Keep and Tasks drew the loudest complaints. One hundred nine votes went to that single issue. Users described inconsistent performance when Gemini Spark, the agent’s Workspace integration, tried to pull context or complete actions. Woodward acknowledged the problem. He said reliability needs to be much better across the entire app. Some improvements have already shipped. Others remain in progress.
Tool calling reliability ranked second. Users reported frequent failures when Gemini attempted to invoke external functions or APIs. The model would sometimes pick the wrong tool. Or it would call none at all. Google agreed strongly with the feedback. Noticeable gains should arrive soon, according to the company. That promise matters. Developers and power users rely on accurate tool use for practical workflows.
But the conversation didn’t stop at the top three. Deeper into the replies, users called for better ways to export Deep Research reports straight into NotebookLM. They wanted the ability to switch between Deep Research mode and lighter models like Flash or Pro without losing the thread of a single chat. Woodward admitted he wasn’t aware of those specific pain points. The suggestions struck him as good ideas worth exploring.
Organization of chats drew repeated mentions. Notebooks offer a start. They fall short for users who manage dozens of long-running projects. Google said it will rethink the entire approach. That admission signals openness. Yet it also reveals how much ground still needs covering.
Other requests piled up. Users asked for message editing after sending. Voice dictation accuracy needs work. Mobile app scrolling bugs frustrated many. Google noted that some scrolling fixes have already deployed. Message editing sits in the queue. Voice improvements rank as a priority. The team plans to dig further into the replies for additional signals.
One topic produced mixed reactions. Some wondered why Gemini lacks Siri-style voice customization. A quick poll in the thread showed 44 percent in favor. Fifty-six percent said no need. Gemini already offers voice options. The debate highlighted differing expectations about personalization.
Then there are the guardrails that won’t budge. Users requested removal of the Nano Banana watermark. That internal code name refers to restrictions on generating images of celebrities. Google explained the limitation stems from navigating AI rules that vary by country. The policy stays in place. No change expected.
Expansion of MCPs and Custom Skills also appeared on the list. Early support for Spark encouraged users. They now want more flexibility and power. The request aligns with broader industry moves toward customizable agents.
These complaints didn’t appear in isolation. Recent coverage shows similar patterns. Android Central reported in May that users slammed new usage limits on Gemini Advanced. Single prompts, especially video generation, burned through quotas faster than expected. Google listened. Josh Woodward announced adjustments. The company now caps the quota a single prompt can consume. Failed requests no longer count against limits. Detailed usage breakdowns will arrive to clarify consumption. https://www.androidcentral.com/apps-software/ai/complaints-worked-google-is-already-addressing-geminis-new-usage-limits
That quick response suggests Google treats feedback seriously when volume spikes. Yet the volume keeps rising. TrustRadius synthesized eight recent reviews as of late May 2026. Users praised speed and integration with Gmail threads, Docs, Sheets and Drive. They valued multimodal capabilities and free-tier access. But context loss in long conversations frustrated them. Mathematical errors appeared too often. Hallucinations persisted on simple tasks. Code generation felt basic. Image and video instructions sometimes failed. One reviewer preferred Claude for depth on specialized topics. https://www.trustradius.com/products/google-gemini/reviews/all
Trustpilot reviews echo the pattern. Some called video creation useless, citing artifacts and outright failures. Others described the April 2026 update as having lobotomized capabilities. Aggregate sentiment hovers in the low-to-mid range. Speed earns praise. Reliability draws fire.
Google has pushed forward on the product side. Its May 19, 2026, I/O announcements highlighted AI agents inside Search. Users can now invoke agents simply by asking questions. A redesigned intelligent search box marks the biggest interface change in more than 25 years. Advanced model capabilities power the shift. Elizabeth Reid, vice president of Search, described the effort as reimagining how people ask questions and navigate information with generative AI. https://blog.google/products-and-platforms/products/search/search-io-2026/
Those Search upgrades tie directly to Gemini’s underlying models. Yet they don’t automatically solve Workspace integration headaches or chat organization woes. The gap between flashy announcements and day-to-day friction remains wide.
Recent X posts reinforce the divide. One user ran the same market research prompt across Claude, Grok, ChatGPT and Gemini. Claude outperformed the rest by a wide margin. Gemini once led in that category. Now it trails. Another developer switched from a slow Claude session to Gemini 3.5 Flash High and finished the task in five minutes instead of twenty-five. Inconsistent performance across models and tasks fuels the frustration.
Google’s own Workspace privacy hub notes that in-product feedback is voluntary. The company uses it for aggregate analysis, bug fixes and safety improvements. It does not train models on that data. Transparency helps. But users want faster translation from feedback into shipped code.
So where does that leave Gemini? Strong in ecosystem integration. Fast on many tasks. Yet repeatedly criticized for context retention, accuracy on calculations, tool reliability and long-term chat management. The executive outreach on X produced candid replies and some concrete commitments. Noticeable gains on tool calling. Rethinking of Notebooks. Fixes for scrolling and voice.
Those steps matter. They show willingness to listen. But the list of open items stretches longer than the addressed ones. Celebrity guardrails stay locked. Watermark policies remain. Deep Research export and model switching still need solutions.
Competitors haven’t stood still. Claude gains ground on depth. OpenAI iterates on agentic features. Users notice. They vote with their prompts and subscriptions. Gemini Advanced commands a premium price. Some reviewers say the paid experience underdelivers compared with free tiers of rivals.
Google holds advantages. Billions of users already live inside its apps. Multimodal generation continues to improve. Free access to capable models gives it reach. The question is whether execution on the basics can catch up to the vision.
Woodward’s thread offered a window into that tension. He shared the top 10 requests. He responded to many directly. Some answers signaled progress. Others revealed blind spots. The volume of replies, over 1,700, proves users care enough to speak up. That engagement is an asset if Google acts on it.
Future updates will test that commitment. Will chat organization improve beyond Notebooks? Can tool calling become dependable enough for production use? Does context survive across long sessions and model switches? The answers will shape whether Gemini solidifies as a daily driver or remains a promising but inconsistent option.
For now the feedback loop spins. Users complain. Google responds. Releases ship. New issues surface. The cycle feels familiar across the industry. What sets this moment apart is the directness of the ask and the public nature of the replies. No filtered surveys. No anonymized data. Just an executive posing a tough question on X and watching the answers flood in.
Industry insiders watch closely. Product teams at every AI shop run similar exercises. The difference lies in speed of iteration and willingness to prioritize user pain over new feature announcements. Gemini has the foundation. It needs the follow-through.
WebProNews is an iEntry Publication