A GEO strategy for an online shop: the first month's four decisions

Why a shop plans by product line, not for the whole catalog

A catalog is too big to change all at once and learn anything from. Rewrite four hundred product pages in a month and, whatever the answers do next, you cannot say which change moved them, or whether the season did. Choose two or three product lines, change those, and leave the rest as they are for a quarter. The lines you did not touch become the comparison that makes the result readable.

The rest of the content strategy hangs from that choice. The questions test the lead lines. The product data that matters first is theirs. The stop list keeps the other lines from creeping into the month. And the judgment compares the lead lines with one line you deliberately left alone.

Which product lines to lead with, and a test for each

Pick lines where you hold a fact nobody else states, can keep the stock, and can afford the work. Put each candidate line through the questions below, with your own numbers; a line that fails two of them waits for a later quarter.

Keep it to two or three. Every line added is another set of pages to rewrite, check and keep true, and the more lines change at once, the harder it is to say which change moved an answer. Write down why each line was chosen: at the quarter's review, those reasons are what the choice is judged against.

The questions each lead line is judged on, plus a control

For each lead line, write two or three questions a buyer would type, each pairing the need with a condition your page meets, and add two or three about the shop itself: shipping to a named country, returns, who runs it. These prompts are the panel you read every month, and they stay fixed, because months only compare when the questions behind them are identical. How to write them by need and condition, and what to record for each answer, is in AI visibility for online shops.

Then add a control: one product line you will not change this quarter, with one or two questions of its own in the panel. If the lead lines move and the control does not, the change is more likely yours. If everything moves together, look at the season, a promotion or the assistants before you credit the pages.

Take the first reading before any page changes. That baseline is the only fair starting point for the quarter's judgment, and a free first pass through the visibility check is one way to have it.

Who owns the product record, and which copies are made from it

Every fact an answer might repeat about a product lives in several copies: the product page, the feed you send to a shopping channel or a marketplace, the structured data, the marketplace listing itself, and the shipping and returns pages. Copies drift when several people edit several of them. Decide which system holds the one record, who owns each kind of fact in it, and which copies are generated from it rather than typed.

In a two-person shop the owner column may hold the same name six times, which is fine as long as it is written down. What matters is that every edit goes through the record and every copy is regenerated from it wherever the platform allows. A price typed into a marketplace by hand is a copy nobody updates when the record changes. Product data for AI covers what to check when the copies are compared; the markup that is generated from the record is in schema.org for online shops.

The stop list for a shop's first month

Write down what the month will not include, so nobody spends it on work the strategy ruled out:

• No rewriting the whole catalog. The lead lines first, the rest after the quarter's reading.
• No manufacturer descriptions pasted as they came. A page says what your shop adds, or it says what every stockist's says.
• No marketplace ratings in the markup, and no reviews bought with discounts or free products. Google's review snippet rules exclude ratings aggregated from other websites and incentivized reviews that do not disclose the incentive.
• No page per city or per country that differs only in the place name; one shipping page states the destinations.
• No stock or urgency lines that are not true at the moment they are read, such as "only two left" shown to every visitor.
• No certifications, origins or materials you cannot document.
• No plugin or special file bought as the fix. Google's page on AI features says no special markup or file is required, and the product pages come first.
• No rewording the panel after a disappointing month.

Judging it: the lead lines against the control

Decide the judgment before the first result, and make it a direction rather than a number: the lead lines named on more of their questions, by more of the four assistants, than at the baseline, while the control line stays about where it was. Add accuracy, because a shop can be named wrongly: the prices, destinations and return terms the answers give for the lead lines should match your pages. Read monthly, judge at the quarter.

Beside the panel, count what reaches your orders. A question on each order about how the buyer found you, assistant referrals read as a floor, and cohorts compared before and after a change are set out in measuring AI recommendations for a shop. The overview for online shops shows where each piece sits in the wider work on your AI visibility.

The shop's strategy on one page

The four decisions fit on one page. Keep it where the people who edit products can see it, and change it only at the quarterly review.

Shop GEO strategy, [shop name], written [date]

Lead lines:        [line 1], because we state [fact]
                   [line 2], because we state [fact]
Control line:      [a line we leave unchanged this quarter]
Panel:             [2 or 3 questions per lead line]
                   [1 or 2 for the control line]
                   [2 or 3 about the shop: shipping, returns, who we are]
Baseline:          taken [date], before any page changed
Product record:    kept in [system]
Owners:            descriptions [name]; prices [name]; stock [name];
                   shipping [name]; returns and identity [name]
Copies generated:  page, feed, markup, [marketplace]
Copies typed:      [list them, and shrink the list]
Not this quarter:  [the stop list]
Judged by:         lead lines named on more questions than at baseline,
                   control line about flat, prices and terms right in answers
Read on:           [day each month]; judged on [date, a quarter on]

It differs from the general one-page strategy where a shop differs from a service business: product lines instead of services, a control line, and a product record that several copies are made from.