Practical AI Guides

English

简体中文

English

简体中文

Appearance

Topic Research: Find Hot Topics with X Search and Webpage Reading ​

💡 One-sentence summary: Use the Grok Bot to build a reusable AI topic research workflow — search public discussions on X, read webpages for background, and produce an evidence-backed topic list.

What You'll Be Able to Do After This ​

  • A list of trending topics: 5 topics that are heating up, each with 2–3 X post links and mention/quote counts.
  • A background card for each topic: confirmed facts, different sides' claims, and gaps no one has covered yet, all with webpage sources.
  • 3 reusable magic prompts — next week, just swap in a new keyword and run it again.
  • A spot-checking habit: verify links, demand evidence, and tell in a few minutes whether the Bot has written speculation as fact.

Why Use the Grok Bot Instead of a Web-Based AI? ​

CapabilityWeb-Based AIGrok Bot
Search public discussions on X❌ You have to scroll through posts yourself and copy-paste them to it✅ Search posts, mentions, quotes, and trends directly with no connection needed — read-only, no login
Open webpages and read the originals❌ Can only read the snippets you paste in✅ Opens links with a browser on the cloud computer and reads full text
Provide evidence links❌ Often can't give sources you can click and verify✅ You can require a post or webpage link for every conclusion
Run longer research sessions❌ Interrupted as soon as you close the tab✅ Runs on the cloud computer, keeps working even with your laptop closed
Read your X DMs and bookmarks❌ Not possible✅ Possible after connecting the X plugin (requires X API credits)

Your Current Struggle ​

Producing content every week — official accounts, blogs, video scripts — the most draining part is often not the writing, but deciding "what to write."

  • Manually scrolling X: half an hour of scrolling, no hot topics found, time gone.
  • Judging by gut feeling: "seems like a lot of people are discussing it" — how many? Can't say.
  • Web-based AI helping out: gives you a pile of claims, but never links you can click and verify.
  • Scattered results: research lives in chat logs and notes, and next week you start over from scratch.

When to Use This Technique ​

  • Weekly topic selection: run it once before the topic meeting, and walk in with evidence.
  • AI trend chasing: after a breaking event, quickly judge whether the topic is worth chasing and from what angle.
  • Validating vague ideas: when you have a topic in mind, first check whether anyone is discussing it on X and how far the discussion has gone.

Not suitable for: research that requires going through X DMs or bookmarks — that requires connecting the X plugin first, see "Pitfall Reminders."

🎒 Preparation Before You Start ​

  • First complete Your First Bot: you'll know how to create a Bot, assign tasks, and watch it work on the cloud computer.
  • Give the Bot a clear role, e.g., "Topic researcher: only finds hot topics and organizes evidence, doesn't write the body text."
  • No need to connect the X plugin. Grok Bot X search works out of the box: search posts, read posts, view mentions and quotes, count, check trends — all read-only.
  • The Bot works on a persistent cloud computer and keeps running even when you close your laptop; files and browser state are preserved between tasks.

Follow Along ​

Step 1: Have the Bot Use X Search to Find Public Discussions ​

Why: instructions like "help me find hot topics" are too vague, so the Bot can only speak in generalities. State the domain, quantity, and evidence requirements all at once, and the results become verifiable.

Send this to the Bot in the chat box:

Search recent public discussions on X about "AI coding tools," and list 5 topics that are heating up.
For each topic, provide: the main viewpoints, 2–3 representative post links, and approximate mention and quote counts.

Replace "AI coding tools" with your own domain.

You should see: the Bot opens X search in the computer preview on the right, returning viewpoints, post links, and counts topic by topic, rather than a single unsourced summary.

💡 This is the unique value of the Grok Bot

Without connecting any account, the Bot can search X public content: search posts, read mentions, see who quoted, count recent posts, check trends — all read-only, never logging in as you.

Step 2: Have the Bot Read Webpages for Background ​

Why: X posts are mostly opinions and lack facts. Have the Bot open related webpages to match up "what people are saying" with "what actually happened."

From the results of Step 1, pick the 2–3 hottest topics and send this to the Bot:

For the topic [topic name], open related webpages (official blogs, documentation, news reports),
Summarize: 1) confirmed facts; 2) different sides' claims; 3) parts no one has clearly explained yet.
Prioritize the links I give you: [link 1][link 2]; search on your own only if you can't find them.

This uses two efficient-conversation tips from the official documentation: give the Bot the links to fact sources directly, and explicitly require it to provide evidence.

You should see: the Bot opens the links you provided with its browser, and every fact it returns is labeled with its source; for pages that won't open or require login, it explains where it got stuck.

Step 3: Compile an AI Topic Research List with Evidence ​

Why: scattered notes can't be used directly. Consolidating them into a fixed-format list lets you judge and verify item by item.

Send this to the Bot:

Organize the previous research results into a topic list, where each topic includes:
1. Topic title (one sentence);
2. Why it's hot right now — with X post links and mention/quote counts;
3. Existing material — with webpage links;
4. Gaps no one has covered yet.
Every conclusion must come with a source link; conclusions without sources should be marked "to be verified."

You should see: a list of about 5 items, each with all four fields complete and links that open directly.

Step 4: Spot-Check the Evidence and Cut Topics That Don't Hold Up ​

Why: the Bot may also write speculation as fact. Spot-checking is your only reliable line of defense.

  • Randomly open 2–3 links and confirm the posts and webpages actually exist and the content matches.
  • For each topic idea, state in one sentence who it's written for; if you can't, cut it.
  • When you find an unsourced claim, have the Bot add evidence or delete it outright.

For popularity judgments you're unsure about, ask the Bot directly:

What's the basis for this "high popularity" judgment? Which post? Give the mention and citation counts.

You should see: the Bot cites specific posts and mention/citation counts instead of "lots of people are talking about it"; the list gets shorter but every entry holds up, and the cut topics have clear reasons (no source, unclear audience).

📋 Magic Prompt ​

🎯 Trend Scanner ​

Expected result: get 5 heating-up topics in a field in one go, each with post links and popularity counts.

## Role
You are my topic research assistant, responsible for finding trending topics on X.
## Task
Search recent public discussions on X about 【domain keywords】 and identify 5 topics that are heating up.

## Input
### Required
【Domain keywords】: e.g., "AI coding tools"
### Optional
【Excluded terms?】: related words you don't want to see
【Target audience?】: e.g., "beginner developers"

## Output specification
- One paragraph per topic: topic name, main viewpoints, 2–3 representative post links, and approximate mention/citation counts;
- Sort by popularity from high to low.

## Constraints
✅ Base everything only on public posts found via search;
✅ Attach at least 2 clickable post links per topic;
❌ Don't give "lots of people are talking about it" style conclusions without links;
❌ Don't fabricate post content or counts.

📚 Background Digger ​

Expected result: for a single topic, produce a sourced fact background card that exposes gaps nobody has written about.

## Role
You are a fact-checking researcher, responsible for digging out the factual foundation of a topic.
## Task
Open relevant web pages around 【topic name】 and compile a fact background card.

## Input
### Required
【Topic name】
【Known post links】: 1–3 links picked from the trend scan results
### Optional
【Links to read first?】: official blogs, documentation, news reports
【Language preference?】: e.g., "prefer Chinese sources"

## Output specification
- Three sections: confirmed facts (each with a web page link), statements from various parties, and parts no one has explained clearly;
- Label facts and opinions separately.

## Constraints
✅ Open the links I provide first, then search on your own to supplement;
✅ Cite the source web page for every fact;
❌ Don't present one party's claims as confirmed facts;
❌ Don't omit the "no one has explained this clearly" parts — those are often topic gaps.

📝 Topic List Officer ​

Expected result: consolidate the results of the previous two steps into a list you can take straight to a topic-pitch meeting.

## Role
You are a content editor's assistant, responsible for organizing research results into a topic list.
## Task
Based on the previous research results, produce an evidence-backed topic list.

## Input
### Required
【Research results】: the conversation results from the previous two steps
【Content platform】: e.g., "WeChat official account", "blog", "video script"
### Optional
【Number of topics?】: default 5
【Topics already written this week?】: to avoid repetition

## Output specification
- Each topic: title, why it's hot right now (with links and counts), existing material (with links), and gaps;
- End with one line: excluded topics and the reasons.

## Constraints
✅ Attach source links to every conclusion; mark unsourced ones "to be verified";
✅ Titles should make clear in one sentence who they're written for;
❌ Don't output popularity judgments without evidence;
❌ Don't force multiple topics into a single topic idea.

Checkpoint ✅ ​

  • Every topic in the list comes with X post links and web page links; the ones spot-checked all open and the content matches.
  • You can state in one sentence who each topic is written for; the ones you couldn't have been cut.
  • Every popularity judgment is backed by mention/citation counts — no empty phrases like "lots of people are talking about it".
  • Unsourced claims have been backed with evidence, or deleted.

Pitfall Reminders ​

  • X search only covers public content, and is read-only. To read direct messages (DMs), bookmarks, your home timeline, or who liked your posts, you must connect the X plugin from Connect apps and sign in with X (the X plugin requires X API credits); see Connect Apps (Plugins).
  • Don't write instructions too vaguely. "Help me find trending topics" will only get you generic answers; state the domain, quantity, and evidence requirements clearly in one go.
  • Web pages that require login: the Bot will first use login credentials saved in Vault, then ask you to enter them via Secure Form; for sensitive steps like captchas, it will use Take over to hand the computer back to you — you do just that one step, click I'm done, and it continues.
  • Popularity needs numbers. Require the Bot to give mention/citation counts; don't accept adjectives.
  • Whether X search supports precise filtering by time range is not explicitly stated in the official documentation.@@I18N_KEEP_000016@@

Advanced Tips (Optional) ​

  • Save the instructions from step 1–3 as a skill (a reusable instruction set); afterwards, ask the Bot for this skill by name in chat and rerun it before each weekly topic selection.
  • If your topic-selection cadence is fixed, set up a routine for the Bot to run the research automatically on a schedule, leaving results in the conversation; note that each Bot supports at most 50 routines, with a scheduling interval of at least 5 minutes.

Lesson Summary ​

  • A full AI topic research workflow is now working: X search to find discussions → read web pages to fill in facts → an evidence-backed list → spot-check and cut topics.
  • There are just two core habits: give links to the Bot directly as the source of facts, and require evidence for every conclusion.
  • This Grok Bot topic-selection workflow is reusable every week — swap in a new keyword and it's a fresh round of research.

Next Lesson Preview ​