Sentiment Analysis in GEO
Tracking “sentiment” in LLM output may not be as obvious, or even as useful as one might suppose. For a start, all of the LLMs have their own overall “personality” programmed into the responses. In addition, if the user asks “what is good about [topic/brand]” the sentiment will be a lot different to “what is good about [topic/brand]”. But this is just the start of a very winding path.
What AI says about you can affect your brand, but getting on top of it can feel like an endless task. To start, we need to know what we’re tracking and why. This post looks at the idea of tracking sentiments; the good the bad and the ugly things that AI can imply about you and where this is coming from. Let’s explore how AI interprets sentiment and why that matters for your brand.
Defining ‘sentiment’ when it comes to AI generated content
When an SEO looks at what an AI model, like a Large Language Model (LLM), says about their brand, they’re essentially performing sentiment analysis on the brand overview the AI provides. This involves evaluating the language, tone, and overall sentiment that AI uses when describing or analyzing the brand.
By examining how the AI portrays the brand—whether it’s positive, negative, or neutral—SEOs can gain valuable insights into how their brand is being perceived, both by AI and potentially by the wider public. This kind of analysis helps SEOs identify any biases, refine brand messaging, and adjust strategies accordingly.
In short, by understanding the brand image that AI reflects, SEOs can make sure their brand is represented in the best possible light, improving brand positioning, and boosting overall engagement.
This is all very well until you realise that LLMs have the tiny, inconsequential little issue that they are, and never will be, human. And sentiments, meaning subjective opinion and emotional reaction is an inherently human trait. The sentiments you pick up on a very miniscule level about whether your partner is mad at you or if you’ve upset your boss at work is something that cannot come easy to a machine. It often not what you say but the way you say it which provides context in human communication – so how can an LLM be accurate when it cannot access implied meaning? Let’s dig a little deeper into how sentiment analysis works when you’re not a human.
Well the answer is opinion mining. This is the process of understanding the subjectivity of a response.
How Opinion Mining works:
- Input Text: The text (e.g., a product review, tweet, or comment) is first processed into tokens (smaller chunks like words or subwords) by the model’s tokenizer.
- Contextual Understanding: The LLM uses its powerful architecture (like transformers) to understand the meaning of the text as a whole. Unlike simpler models, LLMs capture not just individual word meanings, but also how words relate to each other in context, which helps detect subtleties like sarcasm or mixed feelings.
- Sentiment Classification: After processing the text, the LLM classifies the overall sentiment into categories like positive, negative, or neutral. The model might also provide a confidence score to indicate how sure it is about its classification.
- Aspect-Based Analysis (optional): Sometimes, LLMs can go a step further and identify specific aspects (features or topics) mentioned in the text, like “battery life” or “customer service,” and assign sentiment to each aspect separately. This is called aspect-based sentiment analysis.
- Output: The result is a clear classification of the sentiment, often with details about which part of the text expressed which sentiment, and which aspects were involved.
How does this affect your brand?
Well, now that you know how sophisticated opinion mining really is, you may be able to make some assumption on how brand identity is especially effective in AI optimization. It will not take your comments at face value, and the reviews said about your brand can and will be interesting.
| Use Case | Why it Matters | How it’s Useful |
| Reputation Management | Tracks public perception of your brand. | Helps you address issues proactively (e.g., customer service, product problems) before they escalate, protecting your reputation. |
| Customer Feedback & Product Improvement | Provides insights into customer emotions toward your products or services. | Helps identify strengths and weaknesses, enabling improvements based on real customer sentiment. |
| Target Audience Insights | Understands the emotions and attitudes of your audience. | Allows you to adjust your marketing strategy, messaging, and content to better resonate with your target audience. |
| Competitive Advantage | Tracks sentiment towards competitors. | Helps identify areas where you can outperform competitors or find gaps in their offerings, giving you a strategic advantage. |
| Marketing Campaign Effectiveness | Measures how well your campaigns are resonating with your audience. | Tracks sentiment during and after campaigns to determine if adjustments are needed for better engagement and conversion. |
| Crisis Management & Proactive Response | Detects shifts in sentiment before they become a crisis. | Alerts you to negative sentiment spikes, allowing for quick responses to avoid PR disasters. |
| Enhancing Personal Branding (for Individuals) | Helps maintain a positive public image for personal brands or influencers. | Allows individuals to shift their behavior or messaging to improve or maintain positive sentiment. |
| Tracking Sentiment Over Time | Monitors public opinion trends over time. | Measures the effectiveness of long-term strategies and campaigns, helping refine efforts for ongoing success. |
| Content Optimization | Evaluates how content resonates with your audience. | Provides insights on which content pieces generate positive sentiment, allowing you to optimize future content for better engagement. |
Example
Let’s take a look at an example of how sentiment tracking can help in these areas. I’ll create a fake brand to demonstrate it super clearly.
Background:
GlowWell, a clean beauty startup, had built a loyal customer base through ethical sourcing, sustainable packaging, and glowing customer testimonials. Curious about how their brand appeared in Waikay.io, the marketing team ran a report.
The Surprise:
Despite overwhelmingly positive reviews and strong brand values, the AI-generated response was lukewarm. It described GlowWell as a “small brand offering natural products, with mixed customer feedback and limited innovation.” This starkly contrasted with their actual public sentiment—and it puzzled the team.
What Happened?
Digging deeper, they found:
- AI was pulling from outdated or low-authority web sources (e.g., an early review questioning the effectiveness of their first product line).
- Sentiment was skewed by a few critical comments that used strong language, which tipped the model’s classification toward “mixed/negative.”
- The brand had done little to optimize their digital footprint—no recent press, few backlinks, and limited schema markup—meaning AI had sparse, imbalanced data to work with.
The Fix:
GlowWell:
- Launched a brand refresh campaign with updated PR, product pages, and partnerships.
- Created authoritative content (FAQs, behind-the-brand stories, testimonials) optimized for AI indexing.
- Monitored LLM outputs regularly to catch misaligned narratives early.
The Outcome:
Within 2 months, prompts like “describe GlowWell Skincare” began returning descriptions aligned with the brand’s true reputation: “an ethical, fast-growing beauty brand with strong customer loyalty and a commitment to clean ingredients.”
Takeaway:
Even a well-loved brand can be misrepresented by AI if the digital signals it draws from aren’t aligned. Sentiment analysis, paired with SEO and content strategy, is now a key pillar of brand management.
Now although this example isn’t a real case study, the strategy is universal. InLinks (waikay’s sister software) found that AI was saying that SEMrush had bought it. This had skewed online narrative, and it wasn’t easy to overcome. We found that this was being taken from the transcript of a YouTube video in French… Weird right? We worked on removing this content and put out content which recentred the narrative on InLinks being a powerful tool in its own right.
Conclusion
AI-driven sentiment analysis isn’t just a technical feature, it’s becoming a powerful filter through which your brand is seen, interpreted, and judged. Whether the sentiment is right, wrong, or somewhere in between, what AI says about you can shape customer trust, influence purchasing decisions, and even alter competitive positioning.
As we’ve seen, LLMs don’t just regurgitate facts, they build narratives based on the digital footprint your brand leaves behind. If that footprint is outdated, unbalanced, or unclear, you’re leaving your brand reputation to chance. The good news? You have control. By proactively monitoring AI-generated sentiment, optimizing your content for clarity and authority, and correcting misinformation when it appears, you can shape a more accurate, consistent, and favorable brand story both for humans and for the machines they increasingly rely on.
In the era of AI optimization, sentiment isn’t just about feelings. It’s about facts, data, and brand power. Keep an eye on what AI says about you, because chances are, your audience is doing the same.
