How to review an AI menu import with human checks

Prepare a source that can be checked

Identify which menu version the import file represents. If a folder contains two PDFs with old and new prices, decide which one to use first. For photographs, make sure product names and prices are readable, edges are complete and page order is clear. A missing page may not stop an import from technically completing, but it can leave the product list incomplete.

Orvunai offers a flow for preparing an import draft from PDF, photographs and text. The draft is a starting point for the business to review and correct. The business remains responsible for knowing whether the import source is current and accurate. Adding a date or version note to the file helps different reviewers refer to the same source.

Check product counts and price matches in the first pass

First compare categories and the product list in each one. Flag missing products, duplicate entries and rows placed in the wrong category. Then check every price alongside its associated product. In particular, a two-column menu can cause a price to be attached to the neighboring product or different size options to be read as one product. Clarify which options the source presents separately.

A fictional drinks menu might contain two different prices for small and large sizes. If the import combines them into one description, correct them according to your system's product organization. Check amounts that use periods and commas too. A price looking familiar is not enough; both its numerical value and currency must be correct.

• All categories visible in the source have been checked.
• Each product name and price have been matched to the same source row.
• Duplicate products and separate size options have been reviewed.
• Missing pages or cropped photographs have been ruled out.

Review descriptions and information sources separately

In the second pass, inspect descriptions, allergen fields and dietary labels. Correctly importing something written in the source is different from a model guessing information absent from it. Do not accept a nutrition value, dietary label or recipe detail inferred only from the product name as a verified business record. Leave uncertain fields empty and ask the kitchen to verify them.

Even a small word change in a product name can affect its meaning. Reopen the source for similar-looking ingredients or foreign-language names. If a description needs rewriting, decide whether to correct the source at the same time; otherwise the next import may restore the old error. This is why the reasons recorded in review notes matter.

Complete approval with the guest view

After the review, choose products from a few different categories and read them in the mobile preview. Do long names, description lines and photographs appear as expected? If an image absent from the source has been attached to the wrong product, remove it or replace it with the correct photograph. Importing is more than moving text; the result should be a menu in which guests understand the right product.

After publication, open the ordinary visitor link and compare the approved sample products again. Verify that the saved draft has actually been published. Finally, keep a short record of the import source, reviewer and unresolved information questions. This prevents you from having to search for the same errors again at the next price or seasonal change.