Generative AI Search - Disrupting the Internet ecosystem?

Welcome to this edition of Gen AI Simplified – your go-to newsletter for making sense of the ever-changing world of generative AI. Today, we're diving into the fascinating evolution of search and exploring how AI-generated summaries are reshaping our search engines. We’ll also take a closer look at the search and deep research options available with ChatGPT, Google, and Perplexity. Finally, we’ll share tips on harnessing these technologies effectively, so you can steer clear of any potential blunders. So grab a cup of coffee, get comfy, and let’s dive in!

AI Summaries in Search Engines: Answering in a Snap

Traditional search engines are like that no-nonsense librarian handing you a dusty index, asking you to sift through a list of blue links—it's all keyword matching and passive interactions, leaving you to do the work. You get tons of links, and it's up to you to figure out which one holds the answer. But now, enter Gen AI search, a game-changer that talks to you like a helpful friend. No more scrolling through endless links; Gen AI gives you direct answers or neat summaries with the extra bonus of creating tables, charts, or even code snippets. Just ask, and it'll serve up responses in plain language with personalized insights, transforming the search process from a grueling task to an interactive and efficient experience.

Both Google and Bing have rolled out AI summary answers right on the search results page. Google’s new Search Generative Experience (SGE) uses AI (powered by Google’s Gemini model) to generate an “AI Overview” at the top of your search results. Instead of piecing together info from ten blue links, you might see a concise summary with key points and even follow-up questions. For example, Google’s AI might summarize the history of the “spaghetti western” film genre and show relevant facts with source links included.

Microsoft’s Bing is on a similar track. The new Bing combines its search index with GPT-4’s generative powers to craft a custom answer for your query. If you ask Bing “What is a spaghetti western?”, it won’t just give you a list of links – it will produce a readable blurb about the genre’s origins and famous examples, complete with citations and the usual web results still listed below. In short, search engines have become a bit like that know-it-all friend: ask anything, get a quick answer. The difference is, these AI-generated answers actually show where they got their info (when they’re doing it right) via links you can click for more details.

And then there’s ChatGPT’s “Search” mode, available even to free users and designed for speed. In this mode, the chatbot scours websites to deliver quick, straightforward answers—perfect when you need information in a flash without the deep-dive details. Even Gemini's new models with apps (Gemini 2.0 Flash Thinking Experimental with Apps) combine search with reasoning to provide rapid, clear responses. It’s worth noting that Perplexity.ai was an early pioneer in this space, offering a conversational search experience long before these newer models arrived. Perplexity not only delivers direct answers but also includes inline citations, so you can quickly verify sources and delve deeper if needed. This early integration of search with large language models set the stage for today’s trend, ensuring that users get both fast information and the ability to explore further details as desired.

Fast Answers, but at what cost?

While proponents tout Gen AI search as a revolutionary shift, promising instant answers and a user-friendly experience, a healthy dose of skepticism is warranted. It's true that early reports suggest some users appreciate the quick overviews and summaries, and the idea is that websites will still benefit from follow-up clicks. However, the reality may be far less rosy. Is it truly a 'win-win' when AI delivers the answer upfront, potentially drastically reducing traffic to the very websites that created the information ecosystem? While search engines like to emphasize increased 'clickable sources,' the critical question is whether users, satisfied with the AI-generated summary, will actually click through at the same rate as before, or at all. Furthermore, this 'experimental' AI magic is far from perfect, raising concerns about accuracy, bias, and the depth of understanding sacrificed for the sake of speed. So, while search engines are indeed morphing into something more like AI assistants, we must critically examine if this transformation truly benefits everyone, especially content creators and users seeking in-depth, nuanced information, or if it primarily serves the interests of speed and search engine dominance.

While it's great to have information at your fingertips, the convenience sometimes comes at the expense of accuracy and depth. And speaking of accuracy, let's dive into some real-life AI summary blunders that remind us even the smartest systems can have a not-so-smart moment.

AI Summary Blunders: When the AI Goes Wrong ??

Life with AI summaries isn’t all sunshine and rainbows; sometimes it’s facepalm and cringe. In the past three months, we’ve seen some high-profile AI summary blunders that remind us these systems can and do mess up – often in hilarious or troubling ways.

The biggest oops recently came from Apple’s News app of all places. Apple introduced an AI-driven news summary feature (as part of “Apple Intelligence” in iOS beta versions) that was supposed to condense news headlines into short notifications. Sounds handy, right? Well, it started spewing fake news – literally. In one case, it took a BBC News headline about a man’s arrest and generated a completely false summary claiming the suspect had shot himself, which was nowhere in the actual story. The error was so bad that the BBC filed a formal complaint, and Apple rushed to disable the feature (for now) in a software update.

And that wasn’t a one-off slip. Apple’s AI summary feature went on a hallucinatory bender: it outed tennis legend Rafael Nadal as gay (no, he hasn’t made any such announcement), and even declared the winner of a darts championship hours before the event had started. ?? These were push notifications branded with news outlets’ logos, which made it look like respected publishers were getting things horribly wrong, when in fact it was Apple’s AI invention. After a series of these gaffes, Apple essentially said “OK, let’s put this thing on pause.” They’ve pulled the plug on the AI news summaries until they can hopefully teach it not to fabricate headlines.

It’s not just Apple. Google’s AI summaries have also had their comic fails (albeit a bit earlier). Remember Google’s SGE we talked about? Early testers prodded it with odd questions and got even odder answers. One viral example: Google’s AI endorsed eating rocks – yes, it cheerfully suggested that consuming small stones could be good for you – because it had picked up a satirical article from The Onion and mistook it as fact. In another instance, it proposed using non-toxic glue as a pizza topping (to make the sauce stick better, of course). Needless to say, these “tips” are not FDA-approved. Google had to scramble to make fixes after these outlandish answers went viral, adding guardrails and warning that the system is experimental. Even Microsoft’s Bing AI has had moments of creative imagination (anyone remember the time it called a user “immature” and compared them to Hitler? That was early on in testing – yikes).

The silver lining is that we’re catching these blunders and companies are responding. It’s a bit like raising a toddler AI: it says something absurd, we correct it, and hopefully it learns not to do that again. Google now watermarks some AI answers with disclaimers and sources, and Apple (after some public embarrassment) will likely recalibrate its model or add strict filters. In the meantime, these fails make for great memes and an important caution: don’t blindly trust AI summaries. They can sound convincing while being completely off-base. As one observer quipped, “I, for one, welcome our new AI overlords – but I won’t trust them with the news just yet.”

Going Deeper: ChatGPT, Google Gemini & Perplexity for Research

While Google and Bing race to bake AI into traditional search, other platforms are offering deeper research capabilities that go beyond a quick answer. If basic search is a snack, think of these as a full-course meal for your curiosity.

ChatGPT – OpenAI’s conversational wunderkind – isn’t just a quirky limerick generator or a handy coding companion anymore. With its latest upgrades, it’s evolved into a full-fledged research analyst. OpenAI has introduced a “Deep Research” mode (currently available to ChatGPT Pro users) that can autonomously dive into complex topics, sift through documents, and compile detailed reports complete with summaries and citations. Essentially, it’s like having a research assistant on tap—one that can tackle tasks which might otherwise take human researchers days to compile. For instance, if you ask, “Analyze the key factors in renewable energy adoption in Europe,” Deep Research will generate a mini-report that not only outlines the main points but also backs them up with relevant sources. It even accepts PDFs or data files, meaning it can efficiently chew through research papers or spreadsheets to save you precious time. Depending on the query’s complexity, this process can take anywhere from 5 to 30 minutes.

Google Gemini – Gemini 1.5 Pro with Deep Research is Google’s latest leap in making AI a true research assistant. Available to Gemini Advanced subscribers, it transforms your research queries into comprehensive, multi-step reports by scouring hundreds of websites for up-to-date information. Imagine asking a complex question and, instead of digging through endless search results, receiving a neatly organized report with key insights and source links—all generated by an AI that can process up to 1 million tokens at once. This feature is designed to save you hours of work, whether you’re delving into academic topics, market trends, or intricate coding challenges. Accessible via Google AI Studio by toggling the model to “Gemini 1.5 Pro with Deep Research,” it marks a significant step toward agentic AI that not only answers your questions but actively gathers and synthesizes the information you need.

These deep research models/agents represent a new wave of AI research assistants. They go beyond giving a quick fact – they can compare viewpoints, summarize lengthy texts, and interact through follow-up questions. The best part: you can have a conversation with them, steering the direction of the research. The next time you have a daunting research task, you might consider opening ChatGPT or Gemini instead of performing 10 separate Google searches. Just remember that even these super-assistants have flaws…

Impact on Traditional Search: Changing of the Guard

All these changes raise a big question: what does generative AI search mean for traditional search engines and the web? It’s shaking things up in a way that could rival Google’s revolutionary debut. For starters, the classic approach to SEO—aiming to be the top link on a results page—is facing a major twist. If users start receiving direct answers from AI-generated summaries, the “top result” might be an AI-crafted paragraph instead of a website link. This shift forces websites to find new ways to stay visible, and marketers are already buzzing about “Generative Engine Optimization (GEO)”—optimizing content so that AI summaries pick it up accurately. In this new era, every carefully crafted word might be reinterpreted by the AI, shifting control away from content creators and emphasizing clarity and consistency.

Then there’s the battle for user behavior. We’re seeing a trend toward zero-click searches, where an AI chatbot answers your query without any extra clicks. As one analyst noted, “Googling” is giving way to “GPT-ing” because why wade through a page of ads and links when an AI can just tell you what you need? This shift threatens traditional advertising models that depend on click-throughs; early data even shows some users have nearly abandoned Google for many queries in favor of ad-free, chat-style answers.

To adapt, Google and Microsoft are trying a hybrid approach—delivering AI-generated answers while still keeping traditional results and ads in the mix. New players, like startup search engines with AI, or even giants like Meta and Amazon, see opportunities to disrupt the status quo by offering a better experience. For users, this means search is becoming more conversational and personalized; soon, you might rarely type in a string of keywords and instead ask full sentences, expecting tailored responses. The search landscape is evolving into a dynamic chat with an AI assistant.

That said, traditional search isn’t dead—it’s evolving. There will always be times when browsing a list of results is useful, especially for ambiguous queries or when exploring different perspectives. Routine factual questions (“What’s the capital of X?”, “Weather tomorrow”, “Best Italian restaurant nearby”) are increasingly handled by AI snippets that provide a quick answer along with a prompt like, “Would you like to read more? Here are the sources.” The roles of search engine, encyclopedia, and personal assistant are all converging.

For businesses and content creators, the takeaway is clear: focus on quality, clarity, and authority in your content. AI algorithms are looking for trustworthy text to summarize. If you’re a go-to authority, your information is more likely to be picked up—and even cited—by the AI. If not, you might get drowned out by an AI-crafted meta-answer. The playing field is leveling in some ways: small sites might get featured if they have the right info, even if they’d never have made it to page one before. Yet, the game is more unpredictable than ever. In the past, we optimized for an algorithm; now, we’re optimizing for an algorithmic reader—the AI—that re-presents our content to human users. It’s search inception, and it’s going to be fascinating to watch unfold.

Tips for Deep Research with AI: Staying Smart (and Sane)

With great power (of AI search) comes great responsibility – ours, to use it wisely. Here are some best practices for using generative AI in your research so you get accurate results and avoid being misled by a confident-sounding bot. Follow these tips to harness AI as a helpful ally rather than a misinformation machine:

1. Double-Check with Reliable Sources – Treat AI answers as a starting point, not the gospel truth. Always cross-verify facts with trusted sources. If ChatGPT tells you a statistic or Perplexity summarizes a news story, take a moment to see if that info appears in a reputable outlet or database
2. Ask for Citations (or Use Tools That Provide Them) – Not all AI chatbots cite their sources by default. However, you can often prompt them to: try asking “Can you show sources for that?” Some will oblige. Better yet, use AI search tools that have citations built-in (like Bing in “chat” mode, or Perplexity) when you need high confidence in facts. Why citations? Because they let you verify the information.
3. Be Specific and Clear in Your Query – AI models are language whizzes, but they’re not mind readers. The quality of the answer hugely depends on the quality of your question (this is the art of prompting). Provide context in your query if needed. For example, instead of asking “Python installation issue,” ask “How do I fix a Python installation error on Windows 10 when the PATH variable isn’t recognized?” The detail helps the AI give a more targeted (and correct) answer. Vague questions often lead to vague (or incorrect) answers. If you’re doing deep research, consider breaking the task into parts or explicitly telling the AI your goals (e.g., “Summarize the pros and cons of X, citing at least two sources for each side.”).
4. Use Follow-Up Questions to Clarify – One beauty of AI assistants: you can have a conversation. Don’t hesitate to ask a follow-up if something in the answer was unclear or if you suspect a mistake. For instance, “Wait, are you sure about that population figure? It seems high.” A good AI will double-check itself when prodded. Follow-ups can also drill down: “Can you explain further why that is the case?” or “Show me the part of the document that says that.” This iterative approach helps iron out misunderstandings and usually yields more accurate info than a one-and-done query.
5. Stay Skeptical of Surprising or “Too Perfect” Answers – If an AI summary sounds amazing or perfectly fits what you hoped to hear, that’s exactly when to raise an eyebrow. AI hallucinations often look real – complete with fabricated quotes, stats, even references that on first glance appear legit. OpenAI themselves warn that ChatGPT can “occasionally produce incorrect answers” and even “make up facts” that don’t exist
6. Keep the AI Updated (and Yourself) – Remember that not all AI models have current knowledge. If you’re asking about something very recent (say, events from this week), make sure you’re using a tool that can browse the live web or has been updated recently. ChatGPT’s default knowledge cutoff is 2021, so unless you have access to its browsing or plugins, it might not know who won yesterday’s game or what “Gemini” even is. Tools like Bing, Perplexity, or others that explicitly search the web in real-time are better for up-to-date queries. And for you, stay updated on the AI’s capabilities: new features are coming out constantly. (Today’s tip: if you use ChatGPT, check out if you have the “browse with Bing” option or OpenAI’s new web search beta – super handy for current events.)

By following these tips, you can leverage the strengths of generative AI search while avoiding its pitfalls. Think of AI as your extremely well-read but occasionally absent-minded research assistant: it knows a ton, it works lightning fast, but you (the human) are still the editor-in-chief who needs to fact-check and make final judgments. Together, you can be an awesome team.

Final Thoughts

Generative AI is simplifying search and research in amazing ways – we’re essentially watching the evolution of how humans find information. From instant summaries on Google to in-depth chats with ChatGPT, the landscape is rich and ever-changing. And hey, it’s fun too – who knew we’d live to see the day an AI earnestly advises eating rocks or that we’d have to tell our phones “no, that news alert is just the AI being weird again”? Wild times.

If you enjoyed this newsletter and want to keep up with the fast-moving world of AI (with Gen AI Simplified as your trusty guide), join our community of subscribers! We promise to keep things engaging, semi-formal, and a little witty – just like today.

Plus, this edition and all future editions will also be available on my blog, where I’ll be adding more technical articles related to AI in general

Subscribe now to get “Gen AI Simplified” delivered to your inbox, and be the one in your circle who explains the latest AI trends in plain English (with a joke or two thrown in).

Until next time, stay curious and keep exploring the AI-driven world – just remember to double-check those AI “facts”! ??