> For the complete documentation index, see [llms.txt](https://trust.impact.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://trust.impact.com/transparency/model-card-social-listening-sentiment-summary.md). # Model Card Social Listening Sentiment Summary This model powers the Social Listening AI Summary feature, which generates a short executive summary of the dominant sentiment around your listener keywords from monitored social posts and their engagement metrics. The summary appears on the Social Listening insights dashboard alongside chart-based sentiment visualizations. | **AI technology** | Google Gemini (via Vertex AI) | | ------------------ | ----------------------------------- | | **Model type** | Large Language Model (LLM) | | **Provided by** | Google (via Google Cloud Vertex AI) | | **Classification** | Generative AI | ## Description This is a specialized generative model that drafts concise, dashboard-ready sentiment summaries for Social Listening. Given a listener's configured keywords and a set of relevant social posts (post text plus engagement metrics such as likes, comments, shares, saves, reach, impressions, and views), it identifies the dominant sentiment related to those keywords (positive, negative, neutral, or polarized) and briefly explains the primary drivers behind it. Social Listening insights gather the keyword-matching posts for the selected listener and time range, request a summary, and show the returned text as the "AI Summary" on the dashboard. The model does not perform open web search; it summarizes only the posts and keywords supplied in the request. The per-post sentiment charts on the same dashboard are produced separately and are not generated by this summary model. ## Intended use **Primary use cases** * Generating a short executive summary of sentiment around Social Listening keywords for a selected time range. * Highlighting the dominant sentiment (positive, negative, neutral, or polarized) and the main engagement-weighted drivers for a dashboard tile. * Helping brand users quickly interpret keyword-related conversation themes without reading every post. * Complementing non-AI chart metrics such as overall sentiment distribution, sentiment by channel, and sentiment over time. **Out-of-scope uses** * Making final brand, partnership, crisis, legal, or enforcement decisions without human review. * Automatically taking actions against partners or creators (for example outreach, suspension, or contractual remedies) based solely on the summary. * Providing legal, regulatory, financial, or reputational counsel. * Serving as the system of record for individual post sentiment (post-level sentiment is handled separately; this model produces an aggregate narrative summary only). * Analyzing content outside the posts and keywords supplied, or performing open web search. * Identifying private individuals or exposing personal data in the summary. The prompt instructs the model not to include personal information. * Generating harmful, unethical, or offensive content. This is strictly prohibited and filtered. ## How we use your data * **Training data.** The underlying Gemini models were pre-trained by Google on large public datasets of text and code. Impact does not fine-tune these models on your impact.com data for this feature. * **Generating your summary.** When Social Listening requests an AI Summary, we provide the model with the listener's keywords and a sample of relevant social posts (text and engagement metrics) for the selected time range. Only the keywords and posts are included in the prompt. The generated summary may be stored with Social Listening insights so it can be shown again without regenerating until a refresh is required. * **Human feedback.** Users can override individual post sentiment labels in Social Listening, separately from this summary. Regenerating the summary after listener or data changes may produce a new narrative based on the posts supplied at that time. ## Limitations & ethical considerations * **Potential for inaccuracy.** Like all LLMs, this model can occasionally misread tone (including sarcasm or emoji-heavy posts), over-weight high-engagement outliers, or give an incomplete picture when keyword matches are thin or ambiguous. Summaries are qualitative, decision-ready drafts, not definitive measurements. Review the underlying posts and chart metrics before acting on a summary. * **Human review required.** The AI Summary does not replace human judgment about brand reputation or partner behavior. You remain responsible for interpreting Social Listening insights and deciding any follow-up. * **Scope and sampling.** The summary is limited to the posts and keywords supplied for the listener and time range. Evidence may be incomplete if few posts mention the keywords, if matches are generic, or if engagement metrics are missing. When evidence is insufficient, the model is instructed to stay neutral or inconclusive. * **Governance.** Use of this AI is governed by our internal AI governance framework and [responsible-by-design practices](/impact.com-ai-trust-center.md#responsible-by-design), which mandate Privacy, Fairness, Transparency, Accountability, Explainability, and Safety. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://trust.impact.com/transparency/model-card-social-listening-sentiment-summary.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.