The Comscore MCP gives you direct access to live audience data through natural language conversation directly within your preferred AI assistant. Use this guide to learn how to structure prompts, interpret results, and troubleshoot data findings.
Comscore MCP connects your AI assistant to the Comscore data available through your Digital, TV, or Social subscription. You can ask questions in everyday language to explore audiences, compare platforms, investigate trends, and assess claims.
Your subscription determines access. Digital, TV, and Social use separate datasets and entity identifiers; tell the assistant which experience you mean, such as a website, app, or TV network. If unsure, ask which dataset fits your question.
Results reflect published measurement periods, rather than real-time audience activity. Some data may be unpublished, suppressed, or outside your access. Ask the assistant to identify missing data and distinguish retrieved figures from calculations and interpretation.
You do not need a perfect prompt or detailed knowledge of Comscore. Start with what you know and ask the assistant to clarify anything missing.
What do you want to learn? What decision will it support? Who will read the output?
Name the platforms or entities, the consumer group, and the geography you want to analyze.
Specify a month, season, event window, or year-over-year comparison. If unsure, ask for the latest available common period.
Name the metrics you need, or ask which would best answer your question. Request a table, summary, or presentation talking points.
Put it together
“Compare TikTok and Instagram for sports fans.”
“We’re preparing an internal media planning recommendation for a sports apparel advertiser. Compare TikTok and Instagram among U.S. adults ages 18–34 who are sports fans, using the latest available common month. Show audience size, reach, and index. Confirm the available audience definition and entity scope before running the analysis, then summarize the main differences and limitations.”
Client support: Identify evidence-supported strengths for a pitch, renewal, or business review. Include findings that limit the case.
Neutral comparison: Evaluate options using consistent criteria, without a preferred winner.
Validation: Test whether a specific claim is fully, partly, or not supported.
Exploration: Discover relevant patterns before choosing a research angle.
Explain the intended use to guide the questions and presentation. Let the evidence determine the conclusion.
For Digital, a domain covers website activity, while a property may include web and app audiences. Ask your assistant to confirm comparable coverage across platforms.
Different months can introduce seasonal effects. Use matching periods across platforms.
A peak month and a period average answer different questions. Label them clearly and include context when highlighting a peak.
“Check whether these results are directly comparable. Flag differences in coverage or timing, and explain how they affect the interpretation.”
Digital entities: Property [P] and domain [M] can represent different coverage. Ask for the selected entity names, IDs, and included assets. Use the scope that fits the question and disclose any differences.
Digital and TV: Do not assume separate dataset audiences can be added to produce deduplicated reach.
Multiple months: An average of monthly audiences is not the unique audience for the full period. Ask how summaries are calculated. Define seasonal windows and apply equivalent peak-selection rules across platforms.
Start with what each metric answers. Ask the assistant to confirm dataset-specific definitions, units, and reference populations.
How many people or households? Confirm the unit and whether the count is unique within the period.
What share of a defined population is reached? Confirm the denominator.
What share of the platform’s audience belongs to the selected group?
How does the selected measure compare with a reference population? 100 is the baseline.
A composition index of 120 means the characteristic is 20% more prevalent than in the reference population. It does not mean 20% more people or a 20-percentage-point difference.
An index near 100 indicates similarity to the reference population. It can coexist with substantial audience scale. Ask about applicable reporting guidance before treating small differences as meaningful.
Explore audience differences: “Which available subsegments are relevant to this question, and how do their patterns differ?”
Understand tradeoffs: “How do audience size, composition, and index change the interpretation?”
Investigate seasonality: “Does this pattern hold across the season and off-season? Show the monthly trend.”
A surprising or unfavorable result may still answer your question. Use follow-ups to understand the finding rather than assuming a preferred outcome.
Request a source note covering the dataset, entities and coverage, audience, market, period, metrics, units, and limitations. Ask how derived results were calculated.
Confirm that the summary matches the supporting figures. Review tool responses or the audit log where your client exposes them; otherwise, ask for source details and verify important figures against your Comscore reporting access.
Choose a prompt and replace the bracketed fields. Ask the assistant to confirm availability when you are unsure.
“Help identify evidence-supported strengths of [platform] for [advertiser category]. This goes to [reader] to support [decision]. Analyze [audience] in [market] during [period]. Highlight relevant differentiation using fair comparisons, and flag findings that limit the case. Provide a summary, supporting table, and source notes.”
“Compare [platforms] on [metrics] for [audience] in [market] during [period]. Use comparable entity coverage and matching periods. Explain the main differences and tradeoffs without a preferred winner. Include source notes and limitations.”
“Assess whether ‘[claim]’ is supported by Comscore data for [market] during [period]. Confirm the relevant entities, audience definition, and metrics. Explain whether the evidence supports the claim fully, partly, or not at all. Suggest more precise wording if needed.”
“Explore the audience profile of [platform] in [market] for [period]. Identify relevant demographic and behavioral patterns without a preset conclusion. Explain the available metrics and reference populations, then suggest useful follow-up questions.”
“Analyze [platforms] among [audience] in [market] across [season/event months] and [comparison months]. Show monthly results using consistent definitions and periods. Explain seasonal differences, label peaks and averages, and describe how any period summaries are calculated.”
Check the dataset, entity coverage, audience definition, market, dates, units, and metric definitions against the reference report.
“Here are my reference figures and settings. Which differences could explain the discrepancy?”
Ask which limitation applies: suppression may reflect reporting or sample requirements; an unpublished period is not yet available; entitlement restrictions concern subscription access.
“Why is this result unavailable? Could an available period, broader audience, or relevant alternative entity answer the question? Explain any change in scope.”
Confirm the setup, then ask what the result establishes and which relevant tradeoffs or audience differences deserve investigation. A baseline index or lower ranking does not by itself establish a lack of business value.
“What does this finding mean for my business question, and what follow-up analysis would help?”
Confirm that the Comscore connector is connected and enabled, that you are signed in to the intended account, and that your subscription includes the requested dataset. If access still fails, contact your Comscore representative with the error message and your request.
If your AI assistant has web access, ask it to read this guide before your first conversation. If it cannot open the URL, upload a PDF or paste the relevant text. Comscore MCP works without the guide.
Read and apply this guide to this conversation: https://mcp.comscore.com/guide
Paste the prompt as your first message. If the assistant cannot access the page, provide the guide as an upload or pasted text.
Add a copy of the guide as a reference file and ask the GPT to use it for Comscore MCP conversations. A saved URL alone may not load the contents.
Reference a locally available copy of the guide in your project’s CLAUDE.md file. Check periodically for updates.
This guide contains prompting tips, rather than research methodology. It does not provide data access and is not required to use Comscore MCP. Review supporting Comscore results before sharing findings externally.