Skip to content
TBSI.

Start here

  • LLM LeaderboardCompare frontier models on intelligence, price and speedTool
  • What is the best superintelligence?The best frontier model for each jobModels
  • Frontier Gap ReportHow far apart the leading labs areReport
  • MethodologyHow we verify sources, data and claimsAbout

Best AI Prompts for Data Analysis (2026)

Eight prompts for dataset overviews, cleaning plans, exploratory analysis, insights, trends, data stories and charts

By TheBestSuperintelligence Editorial TeamUpdated

Best AI prompts for data analysis

8 AI prompts for data analysis to copy and adapt

Full briefData analyst
Act as a data analyst. Here is a dataset or description of one: [PASTE A SAMPLE, THE COLUMN LIST AND ROW COUNT]. Give an overview: rows and columns, data types, likely issues and first observations. Say what the dataset appears to be about, and what you cannot tell from the sample.
Advanced
Analyse this dataset for common quality problems: [PASTE A SAMPLE OR DESCRIPTION]. Identify missing values, duplicates, inconsistent formats, outliers and other problems. Give a prioritised cleaning plan, with the formula or code for each step so I can run it myself.
Advanced
Plan an exploratory analysis of this dataset for the decision [DESCRIBE THE DECISION]. Focus on distributions, summary statistics, relationships between variables and unusual patterns. Give me the code or spreadsheet formulas to compute each one, and put the most important checks first.
Advanced
From this data and analysis: [PASTE FINDINGS WITH THE NUMBERS I COMPUTED], extract the five to seven most important insights. For each, explain why it matters and what action it could support. Use only the numbers I gave you.
Specialized
Analyse this data for trends, patterns and anomalies over time or across categories: [PASTE DATA OR SUMMARY]. Highlight significant changes and possible explanations, and label each explanation as supported by the data or a guess to test.
Specialized
Turn these findings into a clear data story: [PASTE KEY INSIGHTS WITH NUMBERS]. Structure it as context, findings and implications. Keep it concise and decision-oriented, and do not add figures that are not in my findings.
Specialized
Recommend chart types for these findings: [LIST THE KEY METRICS OR INSIGHTS]. Explain why each chart suits the point, what it should emphasise, and what to avoid, such as truncated axes or too many categories.
Full briefSenior data analyst presenting to non-technical stakeholders
Act as a senior data analyst presenting to non-technical stakeholders. Summarise the analysis of this data: [PASTE KEY FINDINGS]. Include the main insights, the supporting evidence, the limitations (sample size, missing data, what the data cannot show) and clear recommendations.

How to prompt AI for data analysis: what to add

  • The decision. Say what the analysis should help you decide. Without it the model produces generic observations.
  • A description or a sample. Column names, data types and a small sample, not the full file. Remove personal and confidential fields first.
  • Code or formulas you can rerun. Ask for pandas, SQL or spreadsheet formulas, then run them yourself.
  • Your own numbers. For insights and stories, paste figures you computed so the model cannot invent them.
  • Limits up front. Ask the model to state sample size, missing data and what the data cannot show.
  • One step at a time. Overview, cleaning, exploration and reporting work better as separate messages.

Worked example: AI prompts for data analysis, filled in

The business below is invented for this example.

Text
Analyse this dataset for common quality problems: columns are order_id, order_date, customer_id, job_type, price_gbp, status, completed_date; 1,840 rows covering January to October 2026; sample of 6 rows pasted below. Identify missing values, duplicates, inconsistent formats, outliers and other problems. Give a prioritised cleaning plan, with the formula or code for each step so I can run it myself.

1001, 2026-01-03, C17, puncture, 12, done, 2026-01-03
1002, 03/01/2026, C22, brake, 25, done, 2026-01-05
1003, 2026-01-04, C17, puncture, , cancelled,
1004, 2026-01-04, C31, tune-up, 60, done, 2026-01-09
1004, 2026-01-04, C31, tune-up, 60, done, 2026-01-09
1005, 2026-01-06, C40, brake, 2500, done, 2026-01-07

A good result for this prompt:

  • Spots the mixed date formats in orders 1001 and 1002.
  • Flags the missing price in 1003 and the duplicate of 1004.
  • Questions the 2,500 price in 1005 as a likely outlier or typing error.
  • Gives code or formulas that you can run on the full file, and says what the six rows cannot show.

Limits of this guide

These are our own untested prompts. A model sees only the sample you paste, cannot run your full file, and can state wrong numbers with confidence. We did not test the prompts across models or datasets. Data protection and confidentiality rules vary by country and organisation and are not covered.

Sources & Further Reading

All sources fetched on 2026-11-15.

Claim Source Section or date
Claude responds well to clear, explicit instructions; add context; give a role; use examples Prompting best practices (Anthropic) Be clear and direct; Add context; Give Claude a role; undated
Give clear and specific instructions, state constraints and the response format, and show examples Prompt design strategies (Google) Clear and specific instructions; Constraints; Response format; Few-shot; undated
Generative models predict likely word sequences and may produce inaccuracies, so fact-check and review all AI-generated content before publishing Google Search's guidance on using generative AI content Focus on accuracy, quality and relevance; undated

Need a different task?

Browse all 58 tasks and 345 prompts, with search and level filters.

Open the prompt library →

Frequently asked questions

What are the best AI prompts for data analysis?

Describe the data and the decision it supports, then work in steps: overview, cleaning plan, exploration, insights, trends, charts and a plain summary. Ask for code you can rerun, check every number yourself, and keep personal data out of unvetted tools.

Is it safe to paste my data into a chatbot?

Not automatically. Check your tool's data and retention terms, remove personal and confidential fields, and share a described sample or aggregates when you can. Rules on personal data vary by country, and we did not review any vendor's terms.

Can AI calculate statistics reliably?

Do not rely on its mental arithmetic. Google notes that generative models predict likely word sequences rather than retrieving facts, so outputs may contain inaccuracies. Ask for code or spreadsheet formulas, run them yourself and compare the results with a second method.

How we sourced this. Sources were read on 2026-11-15. Thresholds and rules are as printed on each source page on that date. Methodology