How do you use Grok on X, and what is it good for?

Grok has been part of X since late last year, but the version released on August 13 is the first many people are trying seriously. xAI calls it an early preview of Grok-2, together with a smaller sibling, Grok-2 mini. Both are in beta, which means features can change from week to week and some accounts see them before others.

What sets Grok apart is not a single benchmark. It is where it lives. Because it sits inside X, it can draw on the platform's posts as they appear, something a general chatbot cannot do on its own. That shapes what it is good for, and what you should be careful about.

How to get to Grok

1. Check your subscription. Grok-2 is available to X Premium and Premium+ subscribers. Free accounts don't see it during the beta.
2. Update the app. xAI asks testers to install the latest version of the X app, since the new interface arrives with the update.
3. Open the Grok tab. xAI has redesigned this space for the release, so it may look different from the Grok you tried before.
4. Pick a model. Grok-2 is the larger model. Grok-2 mini trades a little answer quality for speed, which suits quick questions.
5. Type, or attach an image. xAI describes Grok-2 as understanding both text and pictures, so you can ask it to explain a screenshot or a meme.

When an answer leans on posts from X, ask Grok to show or link the posts it used, and open a few. That is the quickest way to see whether a summary is fair.

What Grok is good for

Following a live conversation. Ask what people on X are saying about a product launch, a storm in your region or a sports result from last night. Grok can gather posts from the last few hours in a way a chatbot trained on older data cannot.

Explaining context. "Why is this hashtag trending?" or "What is this meme referring to?" are natural fits. So is asking for the main arguments on each side of a heated thread.

Everyday writing and coding help. xAI says Grok-2 improves on its predecessor, Grok-1.5, at chat, coding and reasoning, and an early version ranked near the top of the public LMSYS Chatbot Arena leaderboard, where people vote between anonymous answers. That is xAI's claim based on public votes; your own tests on your own tasks matter more.

There is a real catch with the first use. Posts are not facts. A summary of what people are saying can repeat rumors as confidently as news, and TechCrunch notes that earlier versions of Grok slipped up many times when summarizing news. Treat Grok's live summaries as a map of the conversation, then go to the original sources.

Cautions about the image tool

Grok on X can now create images, using a model called FLUX.1 from the start-up Black Forest Labs. In its first days, users quickly showed it would produce pictures of real politicians with few limits, which TechCrunch reports with the US presidential election months away. It is also not clear whether these images carry hidden metadata marking them as AI-made.

Practical rules follow from that. Don't make or share realistic images of real people doing things they never did. Label anything you post as AI-generated, in words, since you can't count on a hidden tag. And when you see a striking image on X this month, remember that it may have taken seconds to make.

A side-by-side test you can run

The fairest way to judge any chatbot is to give it the same job as another one and compare. Here is a prompt that plays to Grok's strength and still works elsewhere:

Summarize what people are saying today about [a local event or product launch].
Give me:
1. The three main points of view, one sentence each.
2. For each one, a link to a post or article that supports it.
3. Anything that looks like an unconfirmed rumor, labelled as such.
Keep it under 200 words.

Paste it into Grok, then into ChatGPT, and score both:

How does Grok-2 compare with ChatGPT?

They are built for different places. ChatGPT is a standalone assistant with a free tier and paid plans, used through its own website and apps. Grok-2 lives inside X, needs an X Premium or Premium+ subscription during the beta, and has direct access to X posts. If your question is about what is happening on X right now, Grok has an advantage by design. If your question is about writing, studying or analyzing a document you upload, both are capable, and xAI's leaderboard results suggest Grok-2 is now in the same league as the leading models.

The bigger differences are in maturity and guardrails. ChatGPT has been refined in public for close to two years, and early Grok users were quick to point out that it refuses the kind of political images Grok will make. Grok-2 is days into a beta, its developer API isn't open yet (xAI says it is coming later this month), and its image tool is visibly looser. For most beginners, the sensible approach is to use each for what it does best: Grok for a quick read on a live conversation, then a careful check of the sources it points to, whichever tool you trust for the rest.