I Tried Google’s Conversation Feature for AI Mode and I’m Not Sure Who It’s For
If you’ve spent any time searching on Google in the last month or so, you may have noticed that Google is pushing a brand new feature: AI mode . AI mode essentially brings web search to Google search using an AI chatbot. (If you’ve ever used ChatGPT or Gemini to search for information, you might be familiar with this.) AI mode takes your query and runs multiple searches at once, combining the results into one final “report.”
Google’s main idea for AI Mode is to provide fast but comprehensive results for complex, multi-part searches. Let’s say you’re planning a vacation to California. Instead of doing one hotel search, another theme park search, and several restaurant searches, you could theoretically tell AI Mode everything you need for your trip, and it would theoretically give you those results. Of course, we’re talking about AI, so there’s plenty of room for error. At this point: Related AI reviews , AI Mode cites its sources, so you can follow along and make sure it’s not hallucinating any of its answers .
It’s a technology that comes with a few caveats: Are the results I’m getting accurate? Am I taking traffic away from the sites that are feeding the AI information in the first place? Google doesn’t really want to focus on that. Instead, it wants to make AI search even easier to use — if you’re comfortable interacting with your search engine, that is.
Google is now rolling out a new feature called “Real-time Search” for AI mode. Once the feature is active on your side, you can turn it on from AI mode using a new button that displays a waveform with a star. Once open, you can start asking your questions out loud. The AI will take what you say as if you were typing a search, and when you are done, it will process that search and speak the results out loud one by one. You will be able to see the sources containing those results in the middle of the screen, which includes both articles and videos. You can scroll through all the sources and click on any if you want to read or watch more.
In theory, this should work. But in practice, I’m not a fan.
My experience using Search Live in AI mode
For context, I only see this feature in the Google app on my end, not in AI mode on desktop. When I tried the feature this morning, it worked pretty much as expected. I threw out a few random queries from top to bottom (like “tell me about George Washington’s military career”), and the AI followed suit. It did a decent job of offering the opportunity for follow-up questions, listing key facts about my query, and asking if I wanted to learn more about each one individually.
The immediate problem, however, was that, unlike a typical Google search, whenever my attention wandered, I missed information. There’s no pause button, and while you can interrupt the AI by speaking, it’s not easy to get it to repeat itself. Google has a new feature that reads out an AI-generated “podcast” of your search results, complete with a pause button. Search Live is n’t that. If you missed information, you missed it, and your best bet is to simply repeat your search to hear it on top. However, you can click a “transcript” button if you’d rather just see a text version of the bot’s response, but then maybe you should have just relied on text search to begin with.
Still, the live source feed is useful — sometimes. It’s useful to be able to scroll through the list while the AI is talking, to see where the information is coming from. I asked it how we knew the moon landing was real, and it provided (mostly) official, credible sources like NASA, the BBC, and various universities. However, it cited a random Facebook post, and while I appreciate that post confirming that the moon landing wasn’t faked, it worries me that the AI is taking some dude’s mini-essay with 13 likes as gospel. One can only imagine the consequences of using any random (potentially misinformed) Facebook post as a “source” for an AI-generated result — especially if the user asking the question doesn’t bother to follow those sources.
Who is this for?
After playing around with this feature this morning, I was left a little confused. Who is this for? Sure, it’s new, but it’s not particularly convenient. I get that it’s “futuristic” to ask technical questions and have them answered out loud, but I don’t think it’s well suited for search.
Honestly, I’m more sympathetic to voice modes for chatbots like ChatGPT or Gemini . The idea there is to actually have a conversation — even if I can hardly imagine wanting to have a conversation with ChatGPT. You can bounce ideas back and forth between parties. But with Search, you’re presumably trying to get information on a topic, and I think it’s much easier to absorb that information when it’s presented in front of you than when it’s quickly read out, without a live transcript. Sure, you can leave voice mode for that transcript, but it’s not particularly convenient.
It seems like this feature isn’t quite ready yet. Part of the problem is that Search Live doesn’t have the conversational intelligence of other chatbots’ voice modes, making these conversations a bit clunky. If you missed something and could ask the AI to go back 10 seconds or so, for example, that would be helpful.
But I think the worst thing about this particular way of searching the web is the accuracy issue. It’s hard enough to use AI when you have to fact-check everything it says — why would I rely on an audio feed of information, especially if I’m in a situation where I can’t follow the sources? When you agree to use this feature, Google even includes a disclaimer, as many AI companies do, that says, “Real-time Search for AI mode is experimental and may make mistakes.” In other words, it’s not a reliable tool for finding information. If I ask Search Live an important question hands-off and it spits out made-up facts, I may take those made-up facts at face value. The information I’ve just “learned” isn’t just useless — it could be actively harmful.
I don’t like a future where we all rely on Google to spit out misinformation that we accept as truth. Whether it does this via text or “conversation,” it makes no difference to me.