GEO sentiment analysis: 5 steps to improve mention quality

What is sentiment analysis? Why it matters for generative engine optimization (GEO)

Sentiment analysis is a method of classifying the emotional tone of text as positive, negative, or neutral. It has long been widely used in social media marketing and customer satisfaction research, and in the AI search era it matters more than ever.

Why does sentiment analysis matter for GEO? Because when AI such as ChatGPT, Gemini, Claude, and Perplexity recommends products or services, its answers reflect how the public regards and feels about a brand. AI does not just look at how many times a brand name appears; it generates answers with an understanding of the context in which the brand is mentioned, that is, whether that context is positive or negative.

For example, if there are many negative reviews online about a sleep supplement brand, saying things like "it didn't work" or "it had side effects," AI may drop that brand from its recommendations or mention it with a caveat. Conversely, if the brand has a strong positive reputation, with reviews like "my sleep quality improved dramatically" or "doctors recommend it," AI will likely recommend it actively.

In other words, sentiment analysis in GEO means understanding and improving the feelings and evaluations that AI associates with your brand. In traditional SEO, keyword rankings mattered most; in the AI search era, "how AI evaluates your brand" decides who wins.

Why GEO needs sentiment analysis

AI reflects brand reputation in its answers

Large language models (LLMs) are trained on huge amounts of text, including review sites, social media posts, news articles, blog posts, and forum posts. From this data, AI implicitly forms an overall reputation score for each brand.

Specifically, the following information influences AI when it generates answers:

• Star ratings and review text on review sites: reviews on Google, Amazon, Tabelog (a major Japanese restaurant review site), @cosme (a major Japanese beauty review site), and similar sites are included in AI training data
• Reputation on social media: the sentiment of posts about the brand on X (formerly Twitter) and Instagram
• Tone of news coverage: how the media reports on the brand
• Expert evaluations: assessments by industry authorities and influencers
• Position in comparison articles: how the brand is rated on comparison sites and in ranking articles

The more positive this information is, the more favorably AI tends to recommend the brand. For details, see how reviews affect AI recommendations, and how to improve them.

The "quality" problem that mention share alone cannot show

In GEO, mention share, which shows how often AI answers mention a brand compared with the other tracked brands, is an important metric. However, a high mention share means little if the "quality" of those mentions is negative. In fact, it can backfire.

For example, if AI answers, "X is inexpensive, but many people complain about its quality," your brand is mentioned, but in a negative context that lowers users' willingness to buy. Sentiment analysis lets you see not only the "quantity" of mentions but also their "quality".

It shows how you differ from competitors

Applying sentiment analysis to competitor brands as well as your own gives you strategic insights such as:

• Which aspects of your brand are rated positively (strengths)
• Which aspects of your brand are rated negatively (weaknesses)
• Which aspects of competitors are rated highly (threats)
• Which areas you can cover where competitors are rated negatively (opportunities)

In this way, sentiment analysis can also serve as the base data for a SWOT analysis in GEO.

How to do sentiment analysis (manual vs. tools)

Method 1: Manual sentiment analysis

First, here is a manual method you can start for free. At an early stage, or for a small business, it can give you enough insight.

Step 1: Ask AI directly

Ask ChatGPT, Gemini, Claude, and Perplexity questions in Japanese, built around keywords related to your brand:

• "What products do you recommend for X (keyword)?"
• "What is the reputation of brand X?"
• "Compare X and Y (a competitor)."

Step 2: Analyze the tone of the answers

Read the AI's answers and decide whether each mention of your brand is positive, negative, or neutral. Look specifically at the following:

• Where your brand appears in the list of recommendations (the higher, the more positive)
• Which adjectives are used (e.g., "popular," "reliable," "use with caution")
• Whether advantages or disadvantages are mentioned more
• Where your brand stands compared with competitors

Step 3: Record the results in a spreadsheet

Ask the same questions regularly and record the results: the date, the LLM, the question, whether your brand was mentioned, and the sentiment (+1, 0, -1).

Limits of manual analysis: it takes time and is easily colored by the analyst's subjectivity. Checking multiple LLMs every day is also not realistic. If you are serious about GEO, consider a dedicated tool.

Method 2: Sentiment analysis with tools

A dedicated tool greatly improves the accuracy and efficiency of the analysis. Its benefits include:

• Automation: run the analysis automatically at a set time every day
• Consistency: uniform criteria keep subjectivity out of the analysis
• Tracking over time: automatically visualize daily, weekly, and monthly trends
• Support for multiple LLMs: monitor several AIs, such as ChatGPT, Gemini, Claude, and Perplexity, at the same time
• Alerts: get notified when sentiment changes sharply

Comparing major sentiment analysis tools

Here we compare sentiment analysis tools you can use for GEO, including both general-purpose sentiment analysis tools and GEO-specific tools.

General-purpose sentiment analysis tools are strong at analyzing reviews and posts on social media and in the news media. However, only a limited number of tools can directly analyze how brands are evaluated within AI search answers (from ChatGPT, Gemini, Claude, and others). For sentiment analysis in GEO, a GEO-specific tool that analyzes the AI search answers themselves is effective.

Sentiment analysis in practice: 5 steps

Step 1: Choose the keywords to analyze

First, choose the keywords (jobs) for sentiment analysis. Points to keep in mind:

• Keywords tied directly to your core services: e.g., "good sleep recommendations," "products for lower back pain relief"
• Cross-industry, job-based keywords: e.g., "what to do about fatigue that won't go away," "how to improve concentration"
• Keywords that include competitor brand names: e.g., "X reviews," "X vs. Y"

For how to think about choosing keywords, see also how to build authority for AI recommendations.

Step 2: Measure your baseline

Before you start improving, measure your current sentiment baseline. Ask each AI questions using the keywords you chose, and record the following:

• Whether your brand is mentioned in each LLM
• The sentiment of each mention (positive/negative/neutral)
• The specific wording and context of each mention
• Your position relative to competitor brands

This baseline becomes the reference point for measuring the effect of your improvements.

Step 3: Classify sentiment and analyze the causes

Sort the data you collected into the following categories and analyze the causes in each:

Step 4: Take action to improve

Based on the results, take specific actions. Here are some examples:

Addressing negative sentiment:

• Identify the causes of negative reviews and improve your product or service
• Respond sincerely to negative reviews
• Announce your improvements in press releases and blog posts
• Expand the content on your official website so that accurate information gets across

Strengthening positive sentiment:

• Thank users who leave favorable reviews
• Feature positive reviews in customer stories on your official website (with permission)
• Ask experts or influencers for third-party evaluations
• Actively highlight your awards and certifications

Step 5: Monitor regularly and keep improving

Sentiment analysis is not a one-off task. It is important to monitor regularly, measure the effect of your actions, and run a PDCA (plan-do-check-act) cycle that feeds into your next actions.

Ideally, check at these intervals:

• Daily: ongoing monitoring of AI answers (automated with a tool)
• Weekly: review sentiment trends and spot anything unusual
• Monthly: create a monthly report and review your improvement actions
• Quarterly: revisit your strategy and add or change keywords

How to use your sentiment analysis results

Improving negative sentiment

If AI answers mention your brand negatively, take the following approaches:

1. Find the root cause and improve the product

The most important thing is to identify the root cause of the negative reputation and actually improve your product or service. Because AI also draws on recent information, its answers will gradually change once the improvements are reflected on the web.

2. Publish authoritative content

On your own blog and in industry media, share how you are addressing problems and examples of improvements. Content that builds authority, such as expert recommendations and third-party certifications, helps improve how AI evaluates you.

3. Improve reviews and word of mouth

Respond sincerely to negative reviews while working to earn new positive ones. It is especially effective to prioritize platforms that are likely to be in AI training data, such as Google reviews and Amazon reviews.

Strengthening positive sentiment

Even if your brand is already mentioned positively, you can strengthen it further.

1. Identify and reinforce the positive factors

Identify which aspects AI rates positively and reinforce them. For example, if your brand is praised as "good value for money," adding more price comparison content can strengthen that perception further.

2. Spread positive mentions

Get positive information into many media outlets in a form AI is likely to cite. Publishing positive information through multiple channels, such as press releases, contributed articles, and interviews, is effective.

3. Create new positive talking points

Keep generating positive information that AI can use as "reasons for recommendation," such as awards, new feature releases, and social contribution activities.

Managing sentiment in the AI search era

Why review management matters

In the AI search era, review management matters more than ever, because AI gives weight to user reviews as "highly reliable third-party information."

Key points in review management:

• Manage reviews on your Google Business Profile: check them regularly and reply to every review
• Monitor review sites: keep a constant watch on industry-specific review sites (such as Tabelog, @cosme, and Kakaku.com, a Japanese price comparison and review site)
• Social listening: track brand mentions on X, Instagram, and TikTok
• Respond quickly to negative reviews: do not leave them unanswered; reply sincerely within 48 hours

For details on how reviews affect AI recommendations, see how reviews affect AI recommendations, and how to improve them.

Improving sentiment through PR

The following PR approaches are effective for improving sentiment in AI search:

1. Coverage in specialist media

Getting articles into specialist industry media makes it more likely that AI will recognize your brand as an "authoritative source." Plan press releases, contributed articles, and interviews.

2. Work with experts and influencers

Positive mentions from industry experts and influencers have a big influence on how AI evaluates a brand. Earn positive third-party evaluations by asking for product reviews, creating joint content, and inviting them to events.

3. Use storytelling

Sharing your brand's philosophy and customer success stories, not just numbers and specs, enriches the context AI can use as "reasons for recommendation." Case studies and customer stories are especially easy for AI to use as material when it writes specific recommendations.

How structured data relates to sentiment

Implementing structured data (Schema.org) on your website makes it easier for AI to understand how your brand is rated. We especially recommend the following:

• Review / AggregateRating: mark up review ratings in a machine-readable format
• Product: describe product information together with its ratings
• Organization: state company information, awards, and certifications clearly
• FAQPage: structure answers to frequently asked questions so AI can cite them easily

Tracking sentiment with AIOPulse

So far we have covered why sentiment analysis matters and how to do it, but doing all this manually and continuously is a heavy burden. With AIOPulse, you can run GEO, including sentiment analysis, efficiently.

AIOPulse's sentiment analysis features

• Monitoring of three major AIs at once: automatically collects and analyzes ChatGPT, Gemini, and Claude answers to Japanese-language prompts every day
• Mention share over time: see the mention share of your brand and your competitors as a time series, updated every day
• Detecting changes: spot sharp rises or drops in mentions of your brand in the daily data
• Analysis by LLM: when sentiment differs from one LLM to another, analyze each one separately
• Competitor comparison reports: compare sentiment with competitor brands to find what sets you apart

Flexible per-site plans

AIOPulse uses simple per-site billing, and you can add keywords as you need them, so you invest only in what you need and avoid wasted costs. You can start small and expand step by step as you confirm the results. Contact us for pricing.

Want to know where your brand stands in AI search today? Contact us to find out how ChatGPT, Gemini, and Claude mention your brand when asked in Japanese, and with what sentiment.

Summary

Here are the key points of sentiment analysis for GEO:

• Sentiment analysis is essential for GEO: AI reflects brand reputation in its answers, so understanding and improving sentiment matters
• Look at the "quality" of mentions, not just the "quantity": even with a high mention share, mentions in a negative context are counterproductive
• Start with manual analysis, then automate with a tool: first get a picture of where you stand manually, then use a GEO-specific tool in earnest
• Work systematically through 5 steps: choose keywords → measure the baseline → classify and analyze → take action → monitor regularly
• Fix the negatives, reinforce the positives: specific actions based on your analysis lead to results
• Review management and PR are key: responding to reviews and authoritative PR work are effective ways to improve sentiment in the AI search era

As AI search grows more influential every day, sentiment analysis is no longer optional; it is a must. If you have not started yet, start today. To learn more about brand management in the AI search era, visit the AIOPulse home page.

Related articles

• GEO in Japan: measure the Japanese answers first: an overview of GEO, the groundwork for sentiment analysis
• How to do competitor analysis for GEO: clarify your position with mention share: compare sentiment with competitors and use it for strategic positioning