Sentiment Analysis on X

Know what people really think

Sentiment analysis determines whether opinions expressed online are positive, negative, or neutral. For an SMB, this capability is valuable: it lets you measure perception of your brand, products, or industry without investing in expensive research tools. Where a survey commissioned from a vendor requires a budget and several weeks, you get an updated readout in minutes.

X is the platform where opinions are expressed most spontaneously and quickly. Dissatisfied customers share frustrations there, satisfied customers recommend products, and professionals comment on industry trends. Grok, thanks to X Search, can analyze these conversations and give you a clear readout, provided you ask the right thing. That’s exactly what this lesson covers.

Take your brand’s temperature

The basic query must demand more than a global verdict. A “rather positive sentiment” teaches you nothing actionable; what you need are the reasons for satisfaction and dissatisfaction, and identification of the posts that actually weigh in the conversation.

Search X for all mentions of [your company]
or [your product] in the last 30 days.
Analyze the overall sentiment:
- Estimated positive / neutral / negative percentage
- The 3 most frequent topics of satisfaction
- The 3 most frequent topics of dissatisfaction
- The most influential posts (high engagement)

This result only makes sense when compared to a competitor. A 15% negative mention rate is neither good nor bad in absolute terms: it becomes information the day you learn your direct competitor is at 30%, or 5%.

Compare sentiment on X around [your brand]
and [competitor brand] this month.
For each brand: dominant tone, recurring topics,
perceived strengths, perceived weaknesses.

Measure event reception

When you launch a product, participate in a trade show, or publish an announcement, the analysis changes object: it’s no longer about your general image but the reception of a specific moment. Systematically request the questions the audience asked. They are neglected in favor of the positive or negative verdict, yet they constitute free raw material for your next FAQ or product page.

Analyze reactions on X to [your announcement/event]
since [event date].
Identify:
- General reception (positive/mixed/negative)
- Constructive feedback
- Most asked questions
- Influencers who talked about it

Build a time series

An isolated measure is worth almost nothing; the series drives the diagnosis. Three method rules suffice to build it. Always use the same base prompt, otherwise you compare readings taken with two different instruments. Explicitly ask Grok to compare with the previous period, which forces it to qualify the movement rather than the state. Finally, record key indicators in a tracking table, even summary: a five-column spreadsheet works, the main thing is that it survives your memory. Here is the monthly tracking model to reuse as is.

Monthly sentiment analysis - [Month/Year]

Search on X: mentions of [your brand]
from [start date] to [end date].

Provide:
1. Overall sentiment score (scale -10 to +10)
2. Estimated mention volume
3. Top 3 positive topics with examples
4. Top 3 negative topics with examples
5. Change compared to previous month
6. Action recommendations

Expand to the sector

Beyond your brand, also gauge the general sentiment of your sector. An SMB that ignores that its potential customers are worrying about an upcoming regulation prepares its communication off target, and realizes it too late.

What is the dominant sentiment on X regarding
[your industry] in France this quarter?
Identify:
- Main concerns of professionals
- Perceived opportunities
- Regulations or changes causing concern
- Innovations generating enthusiasm

This sector readout helps you adjust your messaging and anticipate customer expectations before they articulate them.

What to do with the results

Faced with proven dissatisfaction, the sequence is always the same and its order counts. Start by identifying the cause: product problem, service problem, or misunderstanding created by your own communication? Then respond publicly on X if the complaint is visible there, because public silence passes for an admission. Correct the problem at the source, otherwise you are only buying time. Finally, communicate about the correction: this is the most often forgotten step, and yet it’s the one that turns an incident into proof of seriousness.

Positive feedback requires the same care with an inverse logic, since it’s about amplifying rather than extinguishing. Thank satisfied customers directly on X, where your response will be seen. Share positive testimonials after requesting permission from their authors. Devote your next posts to topics that generate enthusiasm. And spot potential advocates, those accounts that mention you spontaneously and that you would be wrong to leave in anonymity.

What analysis can’t do

Stay lucid about the nature of the tool. Grok estimates sentiment percentages, it does not calculate them in a statistically rigorous way: never present one of these figures as study data, especially in a commercial document. Reliability depends directly on mention volume, and for a local SMB cited three times in the month the analysis has no value. Sarcasm and irony, particularly widespread on X, regularly skew the classification of a post. So consider the result for what it is, a trend indicator and not an exact measure.

Key takeaways

• Sentiment analysis on X gives a real-time overview of your brand perception
• Use a standardized prompt to track evolution over time
• Compare your sentiment with competitors to position yourself
• React quickly to negative trends and amplify positive ones
• Treat results as trend indicators, not exact measures