Getting Started with Grok

Use Grok and public X posts to make an evidence-based purchase decision.

Introduction

30 Second Summary

Buying something new, like wireless earbuds, based on online chatter can feel like hearing every opinion at once. The loudest claim can win even when its evidence is thin.

In this project, you will create a research chat in Grok, an AI assistant from xAI, that gathers recent public posts on X, the social platform formerly called Twitter, about a product you're thinking of buying. You will check the sources before producing a Buy, Skip, or Wait verdict for your own use case.

What You'll Build

You open one saved research chat to see current opinions sorted into Pros, Cons, Common complaints, and Unverified before receiving an evidence-limited Buy, Skip, or Wait verdict with a visible confidence level.

By the end of this project, you'll have:

🔎 A live research chat that finds recent public X posts about a product you choose, with links to each post.
🕵️ A source and bias audit that flags possible sponsorships and keeps unchecked claims marked Unverified.
⚖️ A decision-ready verdict that counts repeated feedback and weighs it against how you'll use the product.
💎 Secret Mission: A regional reality check that tests whether your verdict holds in another country.

Are there any prerequisites?

Google Chrome on a computer.
A personal email inbox that you control.
A free Grok account (you'll create it in Step 1). Free usage is limited and depends on your location.

Before We Start

Before We Start...

Checking the evidence behind online opinions stops the loudest claim from deciding what you buy.

Set Up Your Free Grok Account

Your research needs a signed-in chat where each follow-up stays connected to the public X feedback you find.

In this step, you'll set up a Grok Free account. You'll review its model-training preference before returning to an empty chat.

In this step, get ready to:

Create or sign in to a Grok Free account without adding a payment method.
Review your Improve the Model preference.
Prepare an empty signed-in chat for your research prompts.

Create or sign in to your account

A signed-in account keeps your research available for the follow-up prompts later in the project. Any plan screen is low risk because your path stays on Free at $0/month without a payment method.

Press Cmd+Space (macOS) or the Windows key (Windows) to open your search bar. On Linux, press the Super key.
Type Google Chrome into the search field and press Enter.

You'll see a Google Chrome window ready for the account page.

Open the official Grok account page in Google Chrome.
Choose the existing-account sign-in path if you already have a Grok account.

Follow the account prompts with the personal email inbox you control if you need a new account.
Complete the email verification prompt if one appears.
Click Skip in the top-right corner if Grok takes you directly to its welcome screen after verification.
Select Free if a plan choice appears instead.
Continue without adding a payment method.

You're through account setup. Whether Grok opened its welcome screen or showed a plan choice, your Free account can now hold the research chat you build in the next steps.

Can't Reach Your Free Account?

Check your spam folder if the account email has not arrived.
Stop before entering payment details if the Free option is unavailable in your location.

Limited free access depends on location. Help me troubleshoot access to my Grok Free account.

Review the model-training preference

The Improve the Model preference controls your model-training choice on Grok.com. Reviewing it now puts that decision in your hands before the research starts.

Use the account menu in Grok to open Settings.
Select Data Controls.
Find Improve the Model.

Set the preference to your preferred choice.
Confirm that the control reflects your chosen setting.

Your model-training preference now reflects your choice.

Keep Sensitive Details Out

Keep passwords, payment details, private messages, and other sensitive information out of every project prompt.
Use your product model and purchase criteria as the basis for the research.

Return to an empty research chat

One research chat keeps every later follow-up connected to your first product search. Starting it empty gives the project one clean thread.

Open https://grok.com.
Use Grok's control for starting a new chat if an older chat opens.
Leave the message composer empty.
Keep this chat open for the next step.

Check the page for a signed-in chat.
Confirm that the message composer is ready for your first prompt.

You'll see an empty signed-in chat with a message composer ready for your first prompt. That's your research chat ready to go.

Your Grok account is ready to hold the full research trail. Next, you'll choose a product and ask what people on X are saying about it.

Ask About X Feedback

Your Grok Free account is ready, and your research chat is empty.

Now you will use Grok's live search of public X posts to gather recent opinions. A broad request gives you a fast first pass whose evidence quality is still unknown.

In this step, get ready to:

Choose a product and search recent public X posts about it.
Inspect the raw summary for comments, complaints, and source links.
Check one claim against its post and mark gaps as Unverified.

Choose a product and search X

Prompting turns your product choice into a focused request for current feedback. This first prompt sets a 30-day window and asks for links to the underlying posts.

Switch back to your empty research chat.
Choose a product you're thinking of buying. Wireless earbuds make an easy first pick, but any product people post about works, like a phone, a laptop, or a game.
Record its exact name here: your exact product, such as a wireless earbuds model.
Select the message composer at the bottom of the chat.
Copy this prompt. Your product name fills in automatically:

Search public posts on X from the last 30 days about [[PRODUCT_MODEL="your exact product, such as a wireless earbuds model"]]. Tell me what accounts are saying before I buy it, and include links to the X posts you use.

What Does This Prompt Do?

The 30-day window keeps the first pass focused on recent posts.
The request tells Grok to search public posts on X instead of relying on a general product summary.
The link requirement gives you a path back to each underlying post.

The search can take a moment while Grok gathers recent posts. A brief pause means the live search is still working.

Paste the prompt into the message composer and press Enter.
Wait for Grok to finish generating the answer.

You should see your product named in the response, with links to public X posts where Grok found them. The posts Grok picks and its wording will differ from other learners' answers. What matters is that each claim points to a post you can open.

Didn't Get a Usable Response?

Confirm that the completed prompt includes your product's exact name.
Resend the prompt if the response does not mention your product.
Wait before retrying if your limited Free plan access has temporarily paused new requests.

Help me improve a Grok prompt that should find recent public X posts about my chosen product.

What is Grok?

Grok is an AI assistant from xAI that can search public posts on X while it answers you.

A review site shows a few hand-picked reviews. Grok reads many recent posts at once and sums up what people are saying, with links back to the posts it used.

Reach for it when you want a quick read of current public opinion. Don't treat its summary as proof: the links are the evidence, which is why the rest of this project checks them.

Inspect the raw answer

A polished summary can make every claim feel equally solid. Source checking shows which statements have an underlying post you can examine.

Locate one positive comment in the raw answer.
Locate one negative comment in the raw answer.
Identify one complaint that appears more than once.
Find the links supporting Grok's statements.
Scan for posts from brands or retailers.
Scan for claims without a supporting link.

You may see independent experiences, promotions, repeated points, or unsupported claims sitting together. That visible mix is the intended limitation of the broad first prompt.

Check one source and mark uncertainty

A linked post lets you test whether Grok's wording matches the original text. Missing or weak evidence stays visible as Unverified so it cannot quietly influence your purchase later.

Choose one claim from the raw summary.
Open its linked X post in a new tab.
Compare the post text with Grok's wording.
Return to your research chat.
If the post doesn't support the claim, or won't open, send Grok this message with the claim pasted in: Mark this claim as Unverified: <the claim>

Can't Verify the Claim?

Recheck that the link belongs to the claim you selected.
Compare the post's direct wording with the wording in Grok's summary.
Keep the claim under Unverified if the post remains unavailable or fails to support the summary.

Help me decide whether a linked X post supports Grok's summary.

Review the first prompt in the same research chat.
Confirm that the prompt contains your product's exact name.
Confirm that the raw response names your product.
Confirm that available X-post links remain visible.
Confirm that each checked claim either matches its source or carries an Unverified label.

You should see your product named in both the prompt and the raw response. X-post links stay beside the feedback, and Unverified marks each claim you couldn't check.

Your research chat now holds a current X snapshot with visible evidence gaps. Next, you will separate likely independent opinions from posts that may carry a material connection.

Audit Promotional Opinions

Your Grok research chat now holds a raw summary of recent public X posts about your product. That answer can still mix independent experiences with promotions or brand messages.

A source audit adds context to each linked opinion before it influences your purchase. In this step, you will flag possible material connections, like a sponsorship or a free product, without accusing unclear authors of being fake.

In this step, get ready to:

Ask Grok to audit the sources from its raw feedback summary.
Inspect at least two linked posts for visible promotional signals when possible.
Separate possible bias from independent opinions while keeping uncertainty visible.

Ask Grok to audit the sources

A material connection gives an author a financial or brand-related reason to discuss a product. The audit prompt asks Grok to surface those signals beside the original links.

Switch back to your research chat.
Scroll to the raw X-feedback summary in that chat.
Prepare the source audit by copying this prompt:

Audit the X posts you used. Flag any post that shows a possible material connection, including ad, advertisement, sponsored, #ad, #sponsored, a free or discounted product, a brand relationship, or an affiliate commission. Put unclear cases in "Possible bias" rather than calling them fake. Exclude brand and retailer accounts from the independent-opinion count. Keep links beside every flag.

What does this prompt change?

The audit checks each source for visible signs of a material connection.
The Possible bias label preserves unclear cases without making an unsupported accusation.
The independent-opinion count is how many posts come from regular buyers, not from brands or retailers.
The links let you inspect the evidence behind each flag.

Paste the audit prompt into the message composer.

Before you send it, do you think the independent-opinion count will stay the same after Grok checks for promotional signals?

Send the message to run the source audit.

You should see Grok revisit the linked posts from the raw answer and flag visible connections, with links beside each finding. The independent-opinion count often drops once brand, retailer, and sponsored posts come out. Which posts get flagged will differ from other learners' results.

Missing links or source flags?

Confirm that you sent the audit prompt in the same research chat as the raw X-feedback search.
Check whether the original answer included links that Grok can revisit.
Keep any missing or inaccessible source labeled Unverified.

Help me understand why my Grok source audit is missing links or promotional flags.

Inspect the visible signals

A disclosure or affiliate link gives you a reason to weigh an opinion differently. The signal provides context about the relationship behind the post.

What counts as a signal?

Clear signals include a disclosure such as ad, advertisement, sponsored, #ad, or #sponsored. A free or discounted product can also create a material connection.

An affiliate or purchase link can reveal a commission relationship. A visible signal changes how much weight you give the post without proving that its opinion is dishonest.

Choose the first linked post in Grok's audit.
Open that post in a new Google Chrome tab.
Look for a clear disclosure near the opinion.
Check whether the post contains an affiliate or purchase link.
Return to the Grok audit.
Choose a second linked post when another link is available.
Open the second post in a new Google Chrome tab.
Check whether its author represents the brand or a retailer.

Each opened post lets you compare Grok's flag with the source itself. Two source checks give you evidence that the audit is grounded in the linked posts.

Keep uncertainty visible

Some relationships remain unclear after inspection. Your audit needs labels for that uncertainty so weak evidence cannot quietly enter the independent-opinion count.

Return to the audit response in your research chat.
Confirm that brand accounts are excluded from the independent-opinion count.
Confirm that retailer accounts are excluded from the independent-opinion count.
Check that posts with clear disclosures remain visible beside their links.
Ask Grok in the message composer to label every unclear relationship Possible bias if any label is missing.
Ask Grok in the message composer to label every inaccessible or unsupported source Unverified if any label is missing.
Inspect the revised independent-opinion count.

Review the completed source-audit table from top to bottom.

You should see possible material connections labeled beside their links. Unclear relationships should remain under Possible bias, while inaccessible or unsupported sources should remain Unverified.

Good work. Your research chat now separates visible promotional signals from opinions that can count as independent evidence.

Still seeing questionable independent opinions?

Check whether the account name clearly represents the product brand.
Check whether the account profile clearly represents a retailer.
Move any unclear relationship to Possible bias until stronger evidence is available.

Help me review a Grok source audit without making unsupported claims about the authors.

Your linked opinions now carry the context needed for a fair comparison. Next up, you will organize the checked feedback into patterns without letting repeated posts inflate the evidence.

Reach a Purchase Verdict

Your Grok source audit now keeps likely independent opinions separate from promotions. Unclear sources remain visible as Possible bias or Unverified.

The remaining public X posts still need structure before they can support a decision. In this step, you'll count the checked feedback, then turn it into a Buy, Skip, or Wait verdict for your own use case.

In this step, get ready to:

Organize the checked posts into four counted categories.
Ask Grok for an evidence-limited Buy, Skip, or Wait verdict.
Challenge the verdict against the counted summary.

Organize the checked feedback

A structured evidence summary shows where separate posts agree. It also gives weak claims a clear place outside the supported findings.

Click the message composer in your research chat.
Paste this follow-up prompt into the composer:

Now use only the linked posts that were not excluded. Organize the feedback into "Pros", "Cons", "Common complaints", and "Unverified". For each point, show how many distinct linked posts support it. Do not count repeated posts, reposts, or unsupported summaries. Mark a point "Unverified" if I cannot check its source.

How does this prompt organize the evidence?

The Pros section holds positive experiences from the remaining linked posts.
The Cons section holds negative experiences from the remaining linked posts.
The Common complaints section requires support from more than one distinct linked post.
The Unverified section preserves claims whose sources cannot be checked.

Submit the prompt by pressing Enter.

You should see sections named Pros, Cons, Common complaints, and Unverified. Supported points should show distinct linked-post counts.

Missing a Category or Count?

Confirm that you submitted the prompt in your audited research chat.
Send the same prompt again if Grok omitted a requested category.
Ask Grok to add distinct linked-post counts if it summarized the feedback without them.

Get help correcting the summary with this troubleshooting prompt.

Check that positive experiences appear under Pros.
Check that negative experiences appear under Cons.
Confirm that each supported point shows a distinct linked-post count.
Confirm that each item under Common complaints has support from more than one distinct linked post.
Open one example link under Common complaints and check that it describes the same issue. Ask Grok to lower the count if it doesn't.
Move single unsupported claims to Unverified.
Keep claims with inaccessible sources under Unverified.

Ask for an evidence-limited verdict

A use case gives the verdict a concrete situation to judge. Budget stays fixed for this exercise so the evidence can focus on how well the product fits your needs.

Choose one main way you'll use the product. For earbuds, that might be commuting, calls, or exercise.
Record your choice here: how you'll use it, such as commuting.
Assume the product fits your budget for this exercise.

A useful verdict must trace its reasons back to the evidence you already checked. It also needs enough uncertainty to choose Wait when the sources are too thin.

Before you send the prompt, do you think the checked evidence is strong enough for a confident decision?

Continue in your research chat.
Request your purchase verdict by sending this completed prompt:

Based only on the checked evidence in this chat, give me a verdict of "Buy", "Skip", or "Wait" for this main use: [[USE_CASE="how you'll use it, such as commuting"]]. Assume the product fits my budget. Give three reasons, the strongest counterargument, and a confidence level of "Low", "Medium", or "High". If the evidence is too thin, choose "Wait". Remind me that public X posts are opinions, not a substitute for checking specifications, warranty, returns, and local compatibility.

What does this prompt control?

The opening instruction limits Grok to the evidence already checked in this chat.
The selected use case makes the verdict relevant to how you plan to use the product.
The confidence level keeps the strength of the available evidence visible.
The final reminder keeps practical purchase checks beside the social feedback.

Wait for Grok to finish its response.

You'll see a response that begins with Buy, Skip, or Wait, followed by three reasons, the strongest counterargument, and a Low, Medium, or High confidence level. A Low confidence or a Wait means the checked evidence was too thin for a confident call. Your verdict may differ from other learners' for the same product; what must match is that every reason points to evidence in your chat.

Missing Part of the Verdict?

Point out the missing section in the same research chat.
Ask Grok to answer again using only the checked evidence.

Ask for help refining the verdict prompt.

Challenge the verdict

A confident tone can make weak evidence sound settled. Auditing the answer keeps the verdict tied to the linked summary you already verified.

Match each of the three reasons to a checked point in your earlier summary.
Confirm that the counterargument represents the strongest checked evidence against the verdict.
Ask Grok to revise the verdict to Wait or Low confidence if too many supporting links remain uncertain or inaccessible.
Confirm that the response reminds you to check specifications, warranty, returns, and local compatibility.

Your final response should begin with Buy, Skip, or Wait, with three reasons tied to checked points, the strongest counterargument, and a visible confidence level. Public X posts should stay clearly described as opinions.

Strong finish: your purchase verdict is now anchored to checked evidence.

Secret mission

Run a Regional Reality Check

Compare recent public X posts from a second country or region. Apply the same evidence checks. Decide whether the linked regional evidence supports the original Buy, Skip, or Wait verdict.

Clean Up Your Resources

Choose how much of this project you want to keep. Your Grok Free account and saved research chat create no ongoing project charge.

Resources you used:

The saved Grok research chat containing your complete product research. This includes the regional comparison.

✔️ Keep everything running

No action needed. Choose this if you want to keep the research available for your purchase decision.

Leave the research chat in your Grok history.
Recheck the linked posts closer to your purchase date.
Compare any changed evidence with the saved verdict.

✋ Pause - I'll come back to this later

Shut down running processes by closing Google Chrome, but keep your research chat so you can pick up where you left off.

Close the Google Chrome window that contains your research chat.
Visit https://grok.com when you are ready to continue.
Sign in with the same account if prompted.
Select the top-right history icon with three horizontal lines and a magnifying glass.
Choose the saved research chat from your history.

ⓧ Delete - I don't want to use this again

Remove all project resources and start fresh if you ever want to rebuild. For this project, that means deleting your research chat, which can't be undone. Your Grok Free account stays available.

In your research chat, select the top-right history icon with three horizontal lines and a magnifying glass.
Locate your research chat.
Select the trash button beside that chat.

That removes the research chat while keeping your Grok Free account.

Nice Work!

🎉 You did it! You turned recent public X posts into a source-aware purchase decision using Grok. Your research chat now shows the evidence behind a cautious Buy, Skip, or Wait verdict.

You've learned how to:

🔎 Use Grok to gather current product feedback from public X posts. Reach for it when you want a fast read of public opinion, and check its links before you trust the summary.
🕵️ Run a source and bias audit that flags possible sponsorships, keeping unclear cases marked Possible bias or Unverified.
⚖️ Turn checked evidence into a decision-ready verdict with counted Pros, Cons, and Common complaints, a counterargument, and a confidence level.
💎 Secret Mission: Compare feedback across two regions to test whether linked evidence justifies keeping or changing your verdict.

Ready to quiz yourself? 💪