By Smári Ásmundsson · · 4 min read
Your next big buyer may be a shopping helper, and it cannot read your listing
Your next big buyer may be an agent doing the shopping, and it picks the brand whose product data it can read without guessing. Spend the next quarter on your spec sheet, not your hero image.
A blank field costs you twice. A distributor or retailer item-setup form bounces when the case count, pack size, shelf life or allergen line is empty, and an AI shopping helper reads the same gaps on your product page.
Start with wholesale. An item-setup form is the sheet a distributor or retailer has you fill in before they will list a product. It asks for the case count, the unit size, the pack size, the shelf life and the allergens. If one line is blank, the form comes back. A rejection email may say only "incomplete" and not which fact is missing, so you guess and resend. A shopping helper hits the same wall from the other side. It reads your product page, finds the same blank, and has nothing to compare.
The online shopper is the second case. Some of your future customers may never see your main product photo. They will tell an AI shopping helper what they want, something like "a plant protein, chocolate, under 5 dollars a serving," and the helper will go and compare brands for them. It reads the product page and the shipping terms, then returns four brands. If yours is not one of the four, the shopper never knew you existed.
I do not know how fast this arrives. Some stores may treat these helpers as visitors to block. I would not bet on a date. I would bet on what the helper needs, which is already knowable.
My definition: a product is readable when the page states, in text and without guessing, the size, the count, the ingredients, the price per serving and the shipping cost and time, plus the case count, pack size, shelf life and allergens a buyer's form asks for. If any of those has to be inferred from a photo or a vague sentence, a helper may skip you or guess about your product, and a form will bounce.
This fails quietly. The serving size is in the picture of the label and nowhere in the text. The count says "family size." Shipping shows up at checkout. A human forgives that. A comparison across five brands, or a setup form with a blank line, does not.
In the older days you would find out by doing the shopping yourself: open your listing and four competitors' in separate tabs, build a table of those facts, and mark every cell you could not fill from the page. That is an afternoon's work (illustrative), and the table goes stale.
Today, give this to Claude. Claude is an AI assistant you type to, like a chat window on your computer. You open it at claude.ai in a browser. The text below is the instructions you paste in. Put the web addresses of your product page and four competitors' pages underneath it, yours first. The reading takes about thirty seconds. Getting the five addresses together takes you or whoever handles your software about twenty minutes the first time (illustrative). These are public pages, so no customer data is involved.
You are a careful online shopper comparing five products from the web addresses listed below. The first address is my product. The other four are competitors.
For each product, find these facts using only what is written on the page: size of each unit, number of units or servings, full ingredient list, price, price per serving, shipping cost and delivery time. Where the page gives them, also find case count, pack size, shelf life and allergens.
Then give me three things.
1. A table with one row per product and one column per fact. Where a fact is not stated in text on the page, write "NOT STATED" in the cell. Do not guess and do not calculate anything the page does not support, except price per serving, which you may calculate when price and serving count are both stated.
2. A list of everything you could not determine about MY product, with the reason for each. For example: serving size appears only inside a label image.
3. Draft replacement text for my product page that fixes every item in the second list, written in plain sentences a shopper would read. Then draft the matching rows for a spreadsheet with one row per product and one column per fact, using the same wording.
If you cannot open a page, say so for that address instead of guessing from memory.
The list of gaps is the useful part. Whatever Claude could not determine is what a shopping helper could not determine either, and it is what a buyer's setup form will ask you for. If you already have a rejection email from a distributor or retailer, paste it in under the instructions and ask Claude which missing fact it points at.
The replacement text and spreadsheet rows come back already written. Paste them in.
Illustrative example, made-up brand. A granola company lists "Large bag" and shows the nutrition panel as a photo. Claude returns NOT STATED for size, servings and sugar. The fix is a few sentences and one spreadsheet column, and Claude drafts both.
You end up with a product page and a spreadsheet where every fact is written as text, which is also what a buyer's setup form asks to see.
CPG AI Assessment
Fix the CPG workflow that keeps landing back on your desk.
I map the recurring reporting or operating workflow first, then build only when the first useful fix is clear.