中共辽宁省交通厅党组关于巡视整改情况的通报

Keith Shaw: The arrival of ChatGPT and other generative AI tools — with the ability to get answers quickly and without being flooded with ads — has really disrupted search engines. So what does this mean for the search giants moving forward?

We’re going to dive into the mess around AI and search on this episode of Today in Tech. Keith: Hi, everybody.

Welcome to Today in Tech. I'm Keith Shaw. Joining me on the show today is Thais Castello Branco. She is the Head of Marketing and Strategy at Exa and also an expert in the world of search. Welcome to the show, Thais. Thais Castello Branco: Thank you for having me.

I'm super excited. Keith: All right, so let’s set the table and talk about how, since the arrival of ChatGPT and other AI tools, search engines have been adopting new features like AI Overviews. These changes are disrupting the traditional way people search online.

For example, during Google I/O this past May, Google announced its new AI Mode, which takes this transformation even further. So, as you look at the landscape, is traditional search as we knew it dead? Or is it just going through a transition?

Most people I talk to have changed the way they search. Some have completely abandoned Google and now go directly to ChatGPT or other AI tools.

Thais: Yeah, I think Google is definitely going through a tough phase — probably even internally, right? I would love to be a fly on the wall in a room between the head of Gemini and the head of Google Ads.

I think we're going through a retraining phase of consumer behavior, so we’re still very much in a transition period.

The average person who’s not deeply plugged into the AI world is still going to Google — or they’re using ChatGPT in a Google-like way, typing basic, keyword-heavy prompts instead of embracing a more semantic, conversational approach. That transition is going to take time.

But if Google doesn’t adapt, and traditional search doesn’t evolve, I think they’ll lose significant market share — and they definitely don’t want that. More plugged-in consumers are already migrating their search behavior to platforms like ChatGPT, Perplexity, Claude, and others. Keith: Yeah, I find myself using a mix.

Sometimes I’ll go straight to ChatGPT if I want a deeper conversation or I'm doing research and appreciate the back-and-forth. But if I’m just looking for a company website, I’ll still use Google — it’s fast, it gives me the link, and I’m there.

Of course, I still have to scroll past the sponsored links to find the right one. Also, if I'm not at my desk and I’m using my phone, I tend to go to Google, especially with the default Safari-Google relationship on iPhones.

Maybe that’s the kind of laziness Google is relying on. Ten years ago, Google was my default for everything. But I’ve definitely noticed a shift. Some of my friends have gone all-in on AI — they avoid Google entirely because of the ads. Are you seeing that too? Thais: Definitely.

And I think you nailed it: humans are inherently lazy. We're always in a rush, trying to minimize effort. Google was built for that world — where you type three words and get what you're looking for.

Thais: Google’s UI is still optimized for quick, lazy searches. For things like company websites or trending celebrity queries — “Taylor Swift boyfriend,” for example — or anything native like shopping or Google Maps, Google will remain dominant for a while.

AI platforms still have a long way to go in offering those same native experiences. Another edge Google has is speed — latency. It's extremely fast. So if you already know what you want, and it’s simple, you’ll often go to Google just because it’s quick.

Until AI platforms catch up in UI customization and speed, consumers will still default to Google for many use cases.

Keith: And it’s not just Google being disrupted — it’s also the companies that rely on Google to surface their websites and messages. In the publishing industry, for the past 15–20 years, SEO has been our lifeline for traffic.

I’ve sat through so many SEO seminars, learned how to rewrite headlines, follow best practices — all based on advice that often feels like guesswork. Are companies asking you the same question: “If everyone moves away from Google, how will people find us?”

Thais: Yeah, and I think we’re migrating from SEO to something new: GEO — Generative Engine Optimization. Let me explain by talking about how Exa fits into this evolving ecosystem. Traditional search engines like Google and Bing are keyword-based.

You type a few words, and the engine looks for those exact words on a webpage. That’s how the SEO world evolved — long, repetitive blog posts stuffed with keywords. Exa, on the other hand, is designed for AI. It’s built to understand the full semantic meaning of a page.

As we move into a world where people ask an AI a question, and the AI fetches information for them, search engines need to optimize for AI — not humans.

Thais: GEO is about understanding how AI apps like ChatGPT generate responses. There are two ways to influence that. First, you can affect the model weights — what the AI has memorized from its training data.

But that’s long-term and harder to control, especially with model cutoffs like ChatGPT’s 2023 limit. Second, you can influence the AI’s live search. For example, ChatGPT uses Bing.

So instead of optimizing only for Google, it may be smarter to focus on Bing SEO if you want your content to appear in AI-generated answers. AI tends to prefer content that is short, clear, structured, and semantically rich.

Instead of keyword-stuffed fluff, AI likes clean titles, Q&A formats, and data-packed summaries. GEO is about optimizing for this new behavior.

Keith: So are headlines going to get longer — like full sentences just to trigger AI? Thais: Possibly, but not in a bad way. Titles will need to be more meaningful and truly relevant to the content. You can’t “trick” AI the way you could with keyword-based SEO.

The title has to match the content semantically because AI prefers quoting full passages. The more precise and data-driven the passage, the more likely AI is to surface it.

Keith: I’m glad you’re calling it GEO instead of something like AIEO — that would be a nightmare acronym. Thais: Exactly. I’m really trying to make “GEO” happen.

Keith: Let’s talk about how searching is changing. For example, I used to Google how long to grill chicken, but I’d get six different answers from six different recipe blogs. Now, I want a consensus — I just want the best, most accurate answer.

Is that how AI is going to change the game? Thais: Yes, absolutely.

When you ask something like that, ChatGPT or Claude will search the web — Bing, Google, Exa — and pull top results. Then it extracts and processes the information, synthesizing a consensus answer for you.

For factual, time-sensitive things or subjective queries like recipes, it doesn’t rely solely on its model weights (the information it memorized during training). It pulls fresh data from the web. This gives you a pre-processed, balanced answer so you don’t have to sift through a dozen links.

Keith: And I’m still doing this on my phone, maybe while standing at the grill. You also mentioned retraining users to be more conversational in their searches. Is this a generational shift? Are Gen Z users naturally better at this than older generations?

Thais: Younger consumers will definitely grow up knowing how to talk to AI. Voice will also become more prominent, making it easier to be conversational. People are still lazy typists, so better voice interaction will help. Even older users — like my parents — are learning.

They might start with a Google-style prompt, but as they go back and forth with the AI, they learn to add context, nuance, and detail. It's like training an employee — you need to clearly communicate your expectations.

Keith: I’ve seen those generational usage reports. As a Gen Xer, I read that Gen Z uses AI as a life coach or therapist. That blew my mind.

Thais: I saw a stat that said over 70% of Gen Z users treat AI as an advisor — someone to help with life decisions or offer guidance. It’s a totally different mindset, but it makes sense for them.

Thais: That’s why search becomes even more critical. AI isn’t perfect. It doesn’t have all knowledge memorized. It needs grounding — access to trustworthy, real-time information. Search engines provide that connection to reality. Otherwise, AI can hallucinate or mislead.

It’s like having someone with a high IQ but no medical training operate on you — you want them to be informed. Intelligence without grounded knowledge is dangerous. Search fills that knowledge gap.

Keith: Another trend I’m seeing is that people are searching more on TikTok, Reddit, or YouTube — especially Gen Z. They don’t even use Google anymore. Is that something you’re noticing too? Thais: Absolutely.

Gen Z uses TikTok more than Google for everything from trip planning to news. These platforms are now competing with Google for search traffic and time spent. They’re trying to incorporate AI and expand their search capabilities — adding native search plus AI-generated summaries.

Whether it’s TikTok, Reddit, or Instagram, they’re fighting for the same search moments that used to belong to Google.

Keith: One thing AI platforms have going for them is the lack of advertising. Google feels overloaded with ads — I usually skip the top sponsored results. But if ChatGPT starts adding ads, does that change the trust dynamic?

Thais: That’s a huge challenge. AI has gained trust by being neutral — its recommendations feel objective. If ads get introduced, users will start wondering if the suggestions are paid for. Even with transparency and labeling, the perception changes.

It’s a fine line to walk, and I’m not sure how AI companies will solve it. It’s also a huge issue for Google. Gemini may eventually take market share from Google Ads and traditional search. But how do you introduce monetization without losing trust?

AI is expensive to run, so they’ll need to figure it out.

Keith: So back to the original point — what can companies do now if they’ve relied on SEO traffic for years? It’s not just about swapping SEO for GEO, right? Thais: Right.

First, diversify your channels. If you’re a restaurant, don’t just focus on Google SEO. Use TikTok, optimize your OpenTable or Yelp presence — meet your customers where they are. Second, if SEO has been a big part of your strategy, shift your efforts.

If you’ve got an SEO team or agency, have them focus on Bing, on long-tail keywords, on structured content that AI prefers. Adapt your strategy to show up in AI-generated responses. That’s how users will increasingly discover your brand.

Keith: And what about customer engagement? Should businesses focus more on retaining existing customers rather than just chasing new ones? Thais: Definitely.

It's about building a growth loop. Retention and community management are crucial. Whether you’re a tech company or a restaurant, think about loyalty programs, exclusives, and word-of-mouth incentives. Tech companies have talked about product-led growth and virality for years.

Other industries need to adopt similar tactics — keep your customers coming back and turn them into advocates.

Keith: That was a little bit of inside baseball, but helpful. Switching gears slightly — as generative AI evolves into more agentic AI, what role does search play? Will agents search the web, or will they just “know” everything?

Thais: Oh no, AI agents will definitely be searching — probably more than humans do. That’s exactly what we’re building for at Exa. Our thesis is that AI agents will dominate web search. We’re creating a search engine optimized specifically for AI agents — not humans.

Our customers are building AI agents and features that need access to real-time, reliable web information.

Thais: Agents need to be grounded. Whether they’re pulling from your internal files — Google Drive, Slack, Notion — or the open web, they’ll constantly need to search to stay updated and relevant. That’s how they take informed action.

Keith: What are the pros and cons of having agents search the web — especially if they’re using traditional engines like Google?

Thais: It’s not really about ads or UI for agents. It’s about the structure of the data. Google and Bing are designed for humans — so the results are often indirect.

For example, if you search for “cool AI startups in San Francisco,” Google gives you articles, Reddit threads, and news pieces — not a clean list of companies. An AI agent wants primary, structured data it can quickly process. It needs direct access to the answer.

Keith: That makes sense. For humans, we want summary listicles. But an AI can read all 20 articles instantly. It needs relevance and structure — not fluff. Thais: Exactly.

Search engines for agents should prioritize high-quality, primary sources. That’s how you get accurate and useful AI responses.

Keith: And as agentic AI grows, the issue of trust becomes even more important. At some point, users will just have to trust the answers they’re getting — especially if there’s no easy way to verify why the AI recommended one product over another.

Thais: Yes, and that’s why grounding is so important. Search engines ensure that AI isn’t hallucinating. The AI needs to cite reliable sources and stay current. Otherwise, it’s just guessing — and that’s dangerous when users depend on it for big decisions.

Keith: I haven’t seen official numbers from Google, but do you think we’re heading for an “earthquake” in search disruption? Or will it be more of a slow shift?

Thais: So far, it’s been a slow transition. Google search volumes haven’t dropped significantly — probably because many people still use Google for weather, shopping, sports scores, and other quick queries.

But AI has added a whole new layer of incremental search — people now ask questions they never would have typed into Google. Over time, as apps like ChatGPT, TikTok, or character-based AI apps integrate native search, dependency on Google will decline.

Thais: And much of this will happen without users even realizing they’re “searching.” Companion apps will integrate AI search natively — like you're having a conversation and the app fetches an answer behind the scenes. It feels effortless and natural.

Keith: As long as these apps avoid being intrusive. You mentioned the example where Facebook or Instagram seems to “listen” to conversations and shows you ads immediately — that still feels creepy.

But if an AI companion proactively and helpfully suggests something — without being invasive — that could be useful.

Thais: We’re definitely heading toward more proactive AI. A good example is a company called Klarity. Their assistant sits quietly in the background, monitoring your screen or conversations (with your permission), and then suggests actions or solutions in real time. That proactivity will grow.

But for now, reactive AI — where you initiate the query — is more familiar and less creepy to most users.

Keith: Well, I really appreciate you joining us today, Thais. This is such a massive topic, and I know we’ll be revisiting it again as things continue to evolve. Thanks so much for being on the show. Thais: Of course!

Let me know if you ever want to dive into any other hot topics — I’m happy to help. Keith: Sounds good. All right, that’s going to do it for this week’s show. Be sure to like the video, subscribe to the channel, and drop your thoughts in the comments.

Join us every week for new episodes of Today in Tech. I’m Keith Shaw — thanks for watching. ?