Sentiment in AI: how does AI talk about your brand?

Sentiment in AI answers is the tone an AI assistant uses when it talks about your brand: positive, neutral or negative. It is about the words ChatGPT, Gemini, Perplexity or Claude use when they mention you, and the pros and cons they list. Two businesses can both be mentioned, while one is recommended and the other comes with a warning.

Why sentiment matters

A mention on its own is not enough. Look at these two sentences in an answer about gardeners in Apeldoorn:

• "Landscaping company A is known for its craftsmanship and clear quotes."
• "Landscaping company B is an option, although some customers report that appointments overrun."

Both are mentioned, but the reader will probably choose A. AI sums up in a few sentences what can be found about you online. That summary acts as a first impression. For many people it is even the only impression, because they no longer click through.

Where does AI get its tone from?

AI does not invent the sentiment. It summarises what it finds in sources:

1. Reviews. Ratings on your business profile and review sites are an important source. Recurring complaints often come back in the answer. See reviews and AI recommendations.
2. Forums and discussions. Experiences people share on forums and community sites.
3. Media. News stories and articles about your organisation, positive or negative.
4. Your own website. How you describe yourself, although that carries less weight than what others say about you.
5. Training data. Older information the model read during training. A problem from years ago can still echo through because of this.

How to check sentiment yourself

There are tools that give a sentiment score, but you can also get a good picture yourself. Here is how:

Step 1. Ask direct questions about your brand. For example: "What do you think of [business name] in [town]?", "What are the pros and cons of [business name]?" and "Is [business name] reliable?".

Step 2. Ask comparison questions too. "Which is better, [your business] or [competitor]?" This shows how AI positions you against others.

Step 3. Look at the words. For each answer, note the adjectives and the pros and cons mentioned. Do you keep seeing "affordable" or keep seeing "slow"?

Step 4. Give a simple label. Positive, neutral or negative per answer. That is your own sentiment measurement. Do it for all four models, because they differ.

Step 5. Find the source. Does the AI show sources? Click on them. That is how you find where a negative point comes from.

Repeat this every quarter, or sooner if something has happened, such as a negative news story.

How to improve the way AI talks about you

You cannot ask AI to be nicer about you. You can improve the sources AI draws from:

• Fix the real problem. If customers complain about long waiting times, no copy will help. Tackle the problem and show that you have.
• Respond to reviews. A calm, helpful reply to criticism shows that you take customers seriously. Those replies are readable for AI too.
• Ask happy customers for a review. Recent, concrete reviews give a more current picture than a few old complaints.
• Correct inaccurate information. If a fact is wrong, fix it at the source. See AI gives wrong information about your business.
• Tell your own story clearly. A good about page with facts, approach and experience helps AI describe you accurately. Read an about page that builds trust.
• Earn positive mentions elsewhere. An article in the local paper or a mention in a trade magazine. See local press and PR for AI.

Do not expect overnight miracles

Sentiment changes more slowly than visibility. New reviews and articles first have to be found. What is in the training data only changes with a new version of the model. Assistants that search live, such as Perplexity, usually pick up changes sooner.

Frequently asked questions

Does HelloGeoLabs measure sentiment?
Yes. After every measurement we assess the sentences that mention you as positive, neutral or negative, per AI model, and summarise the strengths and weaknesses AI mentions. It remains an interpretation by AI, so also read the answers yourself using the steps above.

Can one bad review define the whole picture?
Usually not. AI looks at patterns. A complaint that keeps coming back carries more weight than a single negative review.

What if AI says something negative that is not true?
Find the source, ask for a correction where possible and put the correct information clearly on your own website.

Start with visibility

Sentiment starts with being mentioned. With the free AI check you see whether ChatGPT, Gemini, Perplexity and Claude mention you for five customer questions. You can then also judge the answers on tone. Want to follow it every week? Take a look at GEO tracking.