Confirm AI Glot’s translation scope before giving wording instructions
When filling translation gaps in a product CSV, first confirm which columns and blank entries AI Glot’s plan will process, then set terminology and tone. Writing “do not change SKUs” only as a wording instruction is not enough. Check the scope item by item and compare identifiers after downloading before returning the file to your product import workflow.
In a product spreadsheet, titles need translating while SKUs, prices and existing translations should remain unchanged. When you hand the entire CSV to AI Glot, it is easy to overlook that telling it how to write and confirming which content will be sent for translation are two separate checks.
AI Glot’s official operating guide clearly distinguishes the scope plan from batch wording instructions. The latter change how text is translated, not which text is selected. For this kind of product translation gap-filling, the sensible order is therefore to check the scope, address terminology, then validate a file ready for import. The workflow below is designed from the official documentation; it is not an actual upload or translation test.
Take a small gap-filling sample through the whole process
Imagine a product table containing SKUs, English titles, French titles and prices. The team wants only to fill blank French titles, not rewrite translations already confirmed. Keep the original file first, then select a sample containing blank translations, existing translations, repeated brand terms and special characters. The sample is not meant to prove quality at scale. It is meant to expose problems in scope interpretation and import format.
After uploading, give a complete instruction specifying the source column, target language, the condition that only blanks should be filled, and the columns to leave untouched. AI Glot proposes a plan based on the file and the natural-language request. Do not start just because its description sounds right.
Open the item table and inspect the actual text listed. Have existing French titles been selected? Are SKUs or prices included? Then compare the visual preview with the rules and scope description to check that these views have not interpreted “fill the blanks” as “retranslate everything.” If the scope is wrong, return to the plan and change it rather than trying to repair it through later tone instructions.
Once the scope is right, decide how each term should be translated
The website positions the glossary as a way to keep terminology consistent across rows and batches. For brand names, product ranges and easily confused category terms, establish the approved wording first, then use it for this batch. Tone instructions describe the style of the titles, such as keeping them concise and not adding selling points absent from the source.
Keep a clear distinction here: consistent terminology does not mean the right rows were selected, and fluent language does not prove that fields meant to stay unchanged were left alone. Even an attractive translation misses the task if it overwrites previously confirmed text.
Before starting, check the target language, processing volume, quality options and credit cost again. The official API documentation says creating and planning use no credits; approval reserves credits and starts the work. If you have not understood the scope, refining the plan makes more sense than running the full batch and redoing it later. Use this file’s final plan for the actual cost, rather than estimating from file size.
Downloading is only the start of returning to the business system
The official documentation says the result preserves content outside the translation scope, including structure, order and identifiers. Before importing the file back into the product system, turn that promise into a check: align old and new records by SKU, confirm that prices and existing French titles are unchanged, then check that blank entries were filled as intended.
Next, validate the sample in a test environment you can roll back or through a draft import. Check special characters, long titles and field mappings. If you can only import directly into production, first confirm your backup and recovery method instead of treating “the file opens correctly” as proof of safety.
Finally, have someone who knows the target language read the titles customers will see. AI Glot itself describes its output as a first draft at scale; important customer-facing copy still merits human checking. Expand to the full product table only after scope, language and import all pass. Otherwise, keep the original table and resolve the problems the sample exposed first.
Content note:This article is based on publicly available information and is intended as a reference for evaluating tools. Features, pricing, and related policies may change at any time; please refer to the latest official information.
Original source:ai-glot.com