Google AI Studio seems to be a very underlooked AI product with many great features. Today we’ll focus on a single feature that appears to be unique amongst LLMs: the ability to share your screen with the AI live, and have the AI directly read screen content and provide feedback live based on what’s on the screen. As far as I know, Google AI Studio is the only tool offering direct screenshare + LLM integration, and it’s pretty amazing. 

Of course here at OpenMoves we’re primarily interested in using these tools for improving, optimizing, and automating marketing tasks, so I’ve tested a number of experimental use cases on how to leverage screensharing with Google AI Studio for Google Ads Optimization. 

Test #1: Analyzing Search Queries

TEST

• First I shared the client’s website with GAI, prompting it to understand the business
• Then I shared the client’s search query report in Google Ads (500 rows)
• Then I asked GAI to propose negative keywords based on what it saw

OUTCOME

• GAI was able to do the task
• Output and results were worse than what I would have got with a copy/paste into an LLM
• Some output was way off, while others were largely on track 
• Conclusion: this actually does work in concept, but with low horsepower

Test #2: Analyzing Landing Page for CRO

TEST

• First, shared a website with GAI screenshare, asking it to remember the page
• Then I asked for 10 CRO suggestions based on the page content
• I specifically emphasized not to give general suggestions, only specific ideas 

OUTCOME

• As before, GAI was able to view and recall the page
• Several of the suggestions were specific and relevant
• Others however clearly were just boilerplate general CRO ideas 
• Conclusion: Works in concept but any LLM could have done about the same

Test #3: Getting Help with Interfaces 

TEST

• Can GAI help users inexperienced with a particular web interface figure it out? 
• I tried asking it to explain the GAI interface itself
• Then I tried asking it to explain how to make a Google Ads report

OUTCOME

• Really effective! This is a unique value especially since GAI talks to you
• You can click around an interface and ask Google questions on how to use it
• It appeared to give good instructions on my test Google Ads report
• Conclusion: No appeal for experts but for new users to any software it could be great 

Test #4: Open-Ended Optimization Suggestions 

TEST

• For this test I played dumb and just asked GAI to make optimization suggestions
• I brought it to Google Ads homepage and asked it what to do to improve performance 
• I talked it through several steps and prompts to see what ideas it would propose 

OUTCOME

• The prompting did work – it was able to self-direct on ideas, telling me where to click
• It made some mistakes interpreting the interface, but was able to correct them
• The suggestions were so-so but probably could have been improved with context 
• Conclusion: Not quite a killer use case yet, but I can see how it could be in the future 

My view overall: I would compare the screensharing in Google AI Studio in somewhat of the same category as ChatGPT Operator. Actually GAI screenshare is probably rather more polished and production ready than Operator, but it is still nascent, and in many cases you’d be better off just text prompting an LLM with whatever you’re trying to screenshare. 

That said, assuming the tool can get smarter and smoother, the voice chat can get slicker, and the underlying engine smarter, this tool can become a powerhouse. I can see it acting as a live companion to almost any computer task, being available to troubleshoot, offer analysis, a second opinion, and more. In fact I would say it’s a preview into a future of work where an AI assistant is watching your screen all the time, constantly awaiting a voice prompt to assist whatever the user is doing.