As of 13 August 2026, AI can clean up your product database.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open a fresh copy of the product database export and save an untouched backup before changing anything.
   2. Gather the current column definitions, naming rules, SKU and barcode formats, category list, unit conventions, price and VAT rules, and several records that a colleague confirms are correct.
   3. Remove unnecessary personal or commercially sensitive data, then paste the export or a representative sample, the rules and the known-good examples into a chatbot with the supplied prompt.
   4. Ask the chatbot to return the cleaned CSV, change log, possible-duplicate list and unresolved-record list separately, without deleting records or inventing values.
   5. Compare the proposed changes against the original export, checking product identifiers, prices, VAT flags, units, supplier references and duplicate decisions with a colleague who knows the catalogue.
   6. Import only the approved changes into a test copy of the database, compare row counts and key totals with the original, then publish the tested version and retain the change log.

   Prompt

   I need to clean a product database for a UK business. I will provide a CSV export, the column definitions, our rules, and examples of records that are correct.
   
   Database export:
   [PASTE CSV OR A REPRESENTATIVE SAMPLE]
   
   Column definitions:
   [LIST EACH COLUMN AND WHAT IT MEANS]
   
   Cleaning rules:
   [PASTE RULES FOR PRODUCT NAMES, SKUs, BAR CODES, CATEGORIES, UNITS, PRICES, VAT FLAGS, SUPPLIERS, DESCRIPTIONS AND DUPLICATES]
   
   Known-good examples:
   [PASTE EXAMPLES]
   
   Produce:
   1. A cleaned CSV using the same columns and row order where possible.
   2. A change log with the original value, proposed value, reason, confidence and row or product identifier.
   3. A separate list of possible duplicates that must not be merged automatically.
   4. A separate list of missing, contradictory or low-confidence fields.
   5. A count of rows processed, rows changed, possible duplicates and unresolved records. Do not invent or infer a value when the source does not support it. Preserve SKUs, bar codes, prices, VAT flags and supplier references exactly unless a supplied rule explicitly permits a change. Do not delete records. State any rule you could not apply. Return the cleaned data in a format I can save as CSV.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.