From AltaVista to Aitana Lopez: AI Just Ate the Internet (And My ConfigMgr Console)

Christopher Hazlitt
May 23
6 min read

Updated: Jun 5

🌐 The Internet's New Front Door: A Technical Perspective on Google's Overhaul

CNN's Lisa Eadicicco reported on May 23, 2026, that Google is giving its iconic search bar its biggest overhaul in 25 years. This update was unveiled at Google I/O. It features an expandable search field, an AI Mode for back-and-forth conversations, custom visuals, mini-apps generated on the fly, and a universal shopping cart that pulls items from across the web. Depending on your perspective, this is either the most exciting development since broadband or another reason your end users will call you because "Google looks different again."

I will share insights on what this means, drawing from decades of experience and some honest reflections on my own AI interactions.

🕰️ A Brief History of Finding Things on the Internet (1994 → Today)

Younger admins may find it hard to believe, but there was a time when "Google it" was not a verb. Google did not exist. If you needed a Microsoft KB article in 1996, you opened AltaVista. You typed a Boolean expression that resembled a PowerShell pipeline crafted by a lawyer. Then, you waited for your 33.6k modem to deliver 47,000 mostly irrelevant results. The first ten results were often link farms or someone's GeoCities page with an "Under Construction" GIF.

The lineage of search engines is roughly as follows:

1994 — Yahoo!: A hand-curated directory. Humans categorized the web.
1995 — AltaVista: Introduced full-text search and natural-language queries. For three glorious years, it was the internet's brain.
1996 — Lycos, Excite, Ask Jeeves: Ask Jeeves attempted natural-language search. You typed a question, and a cartoon butler pretended to understand it.
1998 — Google Launches: With PageRank, it improved the web's signal-to-noise ratio overnight.
2004 → 2010s — The Keyword Era: Two- and three-word queries became standard. SEO emerged as a profession.
2011 — Siri: Voice search opened the door to conversational queries.
2022 — ChatGPT: This broke the model. Users began asking rather than searching.
2026 — AI Mode: The search box evolves from a lookup field to a workspace.

This evolution spans thirty years, transforming "type three keywords, scroll past the ads, hope" into "describe what you want in a paragraph, and the search engine will assemble an answer."

🔍 What CNN's Eadicicco Actually Reported

Strip away the conference glitter, and the headline is clear: Google is rebuilding its search bar because users have changed how they search. Robby Stein, Google's VP of Search, stated that "people are asking much longer and harder questions that no longer have a clear response anywhere on the internet." Data supports this. Semrush's analysis shows queries of 11+ words climbing from 3.27% to 5.37%, and conversational queries jumping from 5% to 20%.

The redesigned search box does several things:

Expands to fit longer prompts, photos, file attachments, and open Chrome tabs as context.
Bridges into AI Mode without requiring a page refresh. Searches in AI Mode have more than doubled every quarter since launch.
Generates custom visuals, interactive graphics, and mini-apps on the search result page.
Adds Gemini Omni, a model that can create realistic avatars.
Introduces a "universal" shopping cart that consolidates items from different retailers.

Robert Langenback at Eight Oh Two Marketing noted that AI has "trained people how to search differently." This indicates that the customer has changed, and Google is racing to adapt.

🤖 Aitana Lopez and the Rise of the Synthetic Influencer

This section warrants attention. CNN highlights Aitana Lopez, an Instagram personality with nearly 400,000 followers. She attends events, hits the gym, and posts beauty tips. However, she is entirely AI-generated. Other examples include Lil' Miquela, Lu do Magalu, and Granny Spills. There are now awards for the best AI personalities. Nearly 80% of marketers have increased spending on generative AI content in the last year.

The internet has always had fakes. The difference now is scale and polish. A fake Geocities page in 1999 was easy to spot. A fake Instagram influencer in 2026 can deceive for an entire scroll session.

🛒 The AI Shopping Land Grab

Adobe reports that traffic to US retail sites from AI services grew 393% year-over-year in Q1 2026. Google, Amazon, Meta, and OpenAI are introducing AI shopping tools. Amazon has integrated its Rufus assistant into a new "Alexa for Shopping." Google launched the universal cart. Pew Research noted that users are less likely to click links when an AI summary appears in the results.

The strategic question has shifted from "how do I rank #1 in Google?" to "how do I get cited by the model that has transformed Google's homepage?"

💻 What This Looks Like from Inside a Config Manager Console

Now, I will share my perspective. The CNN piece focuses on consumers searching the web, but a similar shift has occurred in IT — louder, weirder, and more humbling than any conference keynote.

For two decades, when something broke in Microsoft Configuration Manager, the diagnostic loop was:

Open the relevant log file (usually one of fourteen with overlapping responsibilities).
Look for the most recent ERROR or <![LOG[ ... ]LOG]!> block.
Paste a fragment into Google.
Land on Sherry Kissinger, Garth Jones, System Center Dudes, or your own TechNet forum post from 2014.

This loop worked for decades. The ConfigMgr community is one of the best online. However, the workflow assumed I had time to read several blog posts to assemble one answer.

Then, Copilot appeared in the Microsoft 365 admin center, suggesting it could "help me manage devices." My first interaction involved asking it to write a WQL collection query for "all Windows 11 workstations missing the November cumulative update." It generated a query referencing SMS_R_System.OperatingSystemNameAndVersion = "Windows 11." While this is a real attribute, it does not contain the build number, which I needed. The collection would have enrolled every Win11 machine, patched or not, into the "missing update" group, including the CEO's laptop.

I cross-checked with Claude, who reads documentation. Claude produced a sub-select against SMS_G_System_QuickFixEngineering, the correct class, with the patch's actual HotFixID in the filter. It warned me that SMS_G_System inventory data depends on hardware inventory cycle freshness. It suggested adding a join to SMS_LastSoftwareScan to exclude stale clients and reminded me to bound the collection with a limiting collection.

I am not saying Copilot is ineffective. It is useful for keyword-shaped questions. However, Claude has been my AI Mode for ConfigMgr work — paragraphs of context in, an actual answer out, with caveats addressed before I asked.

The lesson from the CNN piece — that users have changed how they ask questions — reflects my experience over the last eighteen months. I no longer Google "SCCM client not registering 0x80004005." Instead, I describe the symptom, environment, what I've ruled out, and the deployment topology in a paragraph. Then, I receive an answer that isn't a Reddit thread from 2017 with an ambiguous "nvm fixed it."

🏗️ What This Means for the Rest of Us Building on the Web

I run christopherhazlitt.ca on Wix. I am writing this post by conversing with Claude and publishing through the Wix MCP connector. The post you are reading was generated by AI, drafted into Ricos rich content, and pushed into my blog via API in one conversation. Five years ago, that would have required four separate jobs and a contractor.

However, the open web that Google is now assembling into mini-apps and AI summaries depends on humans being incentivized to write content. Pew's data on click-through collapse is not a footnote; it is the entire equation. If Google answers questions directly, the blogs that taught Google the answers receive less traffic, leading to decreased ad revenue and less motivation to write future posts. The AI improves, but the training data thins. Eventually, the search bar becomes a confidently wrong colleague without a librarian to fact-check it.

This tension is the real story beneath CNN's piece. The search bar evolved because we did. Whether the rest of the web continues to evolve alongside it will depend on whether those who write the content AI summarizes can still afford to do so.

🎯 The Bottom Line

The web's UI is shifting from ten blue links to one synthesized answer with embedded tools. This is the most significant shift since Google itself.
The change originated with users, not Google. ChatGPT trained a generation to ask in paragraphs. The search bar is catching up.
AI personalities, AI shopping, AI agents — every layer of the consumer internet is being rebuilt simultaneously.
For IT admins, the lesson is the same in miniature: stop typing keywords into your tools; start describing problems. The right AI assistant asks clarifying questions before guessing.
Copilot is effective for keyword-shaped questions within the Microsoft tenant. Claude has proven better for paragraph-shaped questions about the environment around the tenant. Use both. Trust neither blindly.
The open web is the foundation on which AI operates. We must be cautious not to pave over it.

Now, if you'll excuse me, I need to ask Claude why my software update group keeps falling out of compliance. I will update you in the next post on whether the answer involved a WSUS metadata mismatch or — as I suspect — a single distribution point in Mississauga that hasn't been content-validated since the Trudeau majority.