On November 16, 2024, Grok received web search capabilities. Yes, really. It took this long for X's AI chatbot to be able to search the web. Meanwhile, ChatGPT had search for weeks, Perplexity has always been a search engine, and Google... is Google.
Better Late Than Never, I Guess
When xAI launched Grok, it could only access X posts. Which was fine if you wanted opinions on the latest controversy, but useless for basically everything else. Want to know what time a restaurant closes? Can't help you. Need to look up a historical fact? Nope. Current weather? Try again.
On October 28, 2024, Grok received image understanding capabilities, and on November 23, 2024, Grok received PDF understanding capabilities. So xAI has been steadily adding features that other chatbots launched with. The web search update brings Grok closer to feature parity with... well, with where ChatGPT was a year ago.
I tried the new web search yesterday, and it works. That's about the most enthusiastic thing I can say about it. It searches the web, finds information, and provides answers. Revolutionary it is not.
The Real Question: Who's Actually Using Grok?
On December 6, 2024, Grok was enabled for free users, but with usage limits. Before that, it was X Premium only. Now it's available to everyone on X, which means its potential user base is massive—but I'm skeptical about how many people actually use it regularly.
Every time I open X (I still can't call it that without wincing), I forget Grok exists. It's not particularly integrated into the experience. There's a button, you can click it, and then you're using a chatbot. Compare that to how aggressively Microsoft pushes Copilot in Edge, or how natural ChatGPT search feels when you need it.
Someone I know who covers social media full-time told me they've never used Grok intentionally. "It's just... there. Like most of X's features nobody asked for."
Grok 2 Is Actually Decent Though
Here's the thing: the underlying model isn't bad. Grok-2 outperforms both Claude 3.5 Sonnet and GPT-4-Turbo on the LMSYS leaderboard in terms of its overall Elo score. When tested anonymously under the name "sus-column-r" (which is very on-brand for xAI), it competed well against the best models available.
Grok-2 and Grok-2 mini are available through xAI's enterprise API, and both models demonstrate significant improvements over Grok-1.5. The technology is solid. The problem is everything else—the positioning, the distribution, the reason anyone would choose it over alternatives.
xAI has the technical chops. They just need to figure out the product part.
The Elon Factor
Let's address the elephant in the room: Grok is inextricably linked to Elon Musk, and that's both a selling point and a liability. Grok was described as having been designed to "answer questions with a bit of wit" and as having "a rebellious streak," modeled after The Hitchhiker's Guide to the Galaxy.
In practice, this meant Grok would sometimes be edgy or snarky in ways that other chatbots avoid. The chatbot had a "fun mode" described by Vice as "incredibly cringey," but this mode was removed in December 2024. Even xAI realized the shtick wasn't working.
The challenge for Grok is that its brand is "the uncensored AI" in a market that's increasingly concerned about AI safety and misinformation. That appeals to a specific audience, but limits its mainstream appeal. Companies aren't going to use an enterprise AI that might say something offensive or politically charged.
xAI's Broader Ambitions
xAI raised $6 billion in November 2024 from investors including Sequoia Capital and Andreessen Horowitz, bringing its total funding this year to $12 billion. That's serious money. More than most AI startups will ever see.
The funding is going toward building out infrastructure, training larger models, and competing with OpenAI and Google. xAI is moving at an extraordinary pace, driven by a small team with the highest talent density, and they've introduced Grok-2, positioning them at the forefront of AI development.
So xAI has the resources, the talent, and the technology. What they need is a clear value proposition for why anyone should use Grok instead of ChatGPT or Claude or Gemini.
Where Grok Could Actually Win
There is one area where Grok has a genuine advantage: real-time information from X. Grok has been continuously improving on the X platform, with new features including a redesigned interface. If you want to know what's trending right now, what people are saying about breaking news, or the vibe around a current event—Grok can tap into X's firehose in ways other AIs can't.
That's valuable. But it's also a narrow use case. And given X's... let's say "complicated" relationship with accurate information, I'm not sure "what X is saying about this" is always the answer you want.
My Honest Assessment
Grok adding web search is a necessary feature update, not a game-changer. It brings the product closer to baseline expectations for a modern chatbot. The underlying technology is competitive, maybe even ahead in some benchmarks.
But I still don't know who Grok is for. X Premium subscribers get it as part of their subscription, which is fine. But would anyone pay for Grok standalone? Would companies choose Grok over established alternatives?
Maybe xAI doesn't care. Maybe Grok is just a feature for X, not a standalone product. Or maybe they're playing a longer game where Grok eventually becomes the AI backbone for more Musk companies—Tesla, SpaceX, Neuralink.
Either way, adding web search is a step in the right direction. Now they just need to figure out why anyone should care.