A mention is only half the story
An AI answer can mention your brand yet put a buyer off. It may call your prices high, describe an old feature or suggest you only for narrow uses. Counting mentions misses that. SearchSeal reads each saved answer that names your brand and gives it one of four tone labels: positive, neutral, mixed or negative.
Positive means praise or a recommendation without criticism. Negative means criticism or a warning without praise. Mixed means the answer contains both, and SearchSeal keeps a quote for each side. Neutral covers answers that describe you without taking a view. A late or brief mention does not count as negative, and your place in a list never determines tone.
Every label comes with its quote
To make each label checkable, SearchSeal saves the passage behind it. The quote is copied from the saved answer word for word. If the classifier cannot point to a passage that backs its label, the answer stays unscored instead of getting a guess. When an answer is unclear about tone, or attributes a claim to the wrong company, it is left out.
Each scored answer also gets one to three short descriptions of what the answer says about you, such as a pricing claim or a point about service. Descriptions that repeat across answers become themes, split into praised and criticised. The themes reveal repeated claims about your brand that you can check against your product and pages.
What you see on the sentiment page
Four cards lead the page: a tone score from 0 to 100, the share of positive answers, the count of mixed or negative answers and your tone rank among tracked rivals. The score weights positive answers at 100, neutral and mixed at 50 and negative at 0, then averages them across the scored answers in the selected period.
Below the cards, a tone over time chart follows you and your rivals on a rolling seven day window. A table breaks the score down by AI tool, and another sets your brand beside each tracked rival. Each answer lists the AI tool, quote, original question, label and date. Open a row to read the full saved answer. Filter by period, AI tool, region, tone and theme.
SearchSeal currently monitors AI Overview, ChatGPT, Gemini and Perplexity. Sentiment is part of every plan.
Scores wait for enough answers
One response can swing a score based on three answers, so SearchSeal withholds it. The tone score and tone rank appear once there are at least 20 scored answers in the period. A day on the trend chart shows only when its seven day window holds at least five scored answers. Days below that leave a gap rather than a zero, so a quiet week does not look like a collapse.
Rival tone comes from the same answers. When an answer names you and a rival, both are labelled in one pass. A rival that only appears in answers where you are absent has no tone score from SearchSeal. Read the rivals table as context for your score, rather than a complete view of that company’s AI tone.
What sentiment tracking does not do
The labels come from an AI classifier, and a classifier can misread sarcasm, a comparison or a sentence in another language. Check the stored quote before acting on a label. Treat tone as a reading aid that points you to the answers worth opening first, and read those answers yourself before reporting a change.
SearchSeal scores only complete answers that name your brand. It does not score answers where you are absent, and it does not rewrite earlier labels when the method changes, so older answers may show as unscored. The score is not a review rating or a measure of customer satisfaction. It captures how sampled AI answers described you for specific questions, dates and tools.
Turn criticised themes into a work list. If an answer repeats an outdated price or a missing feature, find the page that should correct it. The fix queue can turn that into a ranked task, and later answers show whether the wording moved.